Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The flow diagrams depicted in the figures are merely illustrative and not necessarily all of the elements and operations/steps are included or performed in the order described. For example, some operations/steps may be further divided, combined, or partially combined, so that the order of actual execution may be changed according to actual situations.
It is also to be understood that the terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
It should be further understood that the term "and/or" as used in the present specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes such combinations. Referring to fig. 1, an embodiment of a method for adaptively adjusting parameters of an analog light source controller according to an embodiment of the present application includes:
Step 100, acquiring an input voltage signal, an output current signal, an ambient temperature signal and a channel power signal of an analog light source controller, and inputting the signals into a neural network model for parameter mapping to obtain a light source parameter dynamic response matrix;
it can be understood that the execution body of the present application may be an adaptive adjustment device for parameters of an analog light source controller, and may also be a terminal or a server, which is not limited herein. The embodiment of the application is described by taking a server as an execution main body as an example.
Specifically, an input voltage signal acquired by an SMP-03V-BC port is subjected to digital quantity conversion and phase compensation through a 16-bit high-precision analog-to-digital converter, and a voltage input sequence is obtained. Digital quantity conversion is to convert analog signals into computer-processable digital signals, and phase compensation is used to correct phase shift caused by sampling delay and system response so as to ensure accuracy of voltage signal acquisition. When the output current signal is processed, the output current signal is collected through the constant voltage driving circuit, and the current detection and the baseline calibration are carried out by using the Hall current sensor with the 2.5A measuring range, so that a current input sequence is obtained. The Hall current sensor can accurately detect a current signal, can provide higher response speed and lower energy loss, and ensures that the current signal is not influenced by baseline drift in the detection process in the baseline calibration process, so that the accuracy of current detection is improved. Meanwhile, an ambient temperature signal is acquired through the forced cooling controller, and temperature sampling and temperature compensation are carried out by using the NTC thermistor, so that a temperature input sequence is obtained. An NTC thermistor is a temperature sensor whose resistance varies with temperature, and is capable of sensitive sampling of ambient temperature. The temperature compensation process is to ensure that the temperature signal is not interfered by the external environment in the measurement process, so as to obtain accurate ambient temperature data. For the channel power signal, a four-channel power acquisition mode is adopted. And carrying out power detection on the power signal of each channel through a channel isolation operational amplifier, and calculating the load condition of the channel to obtain a power input sequence. The use of the channel isolation operational amplifier can avoid signal interference among the channels and improve the accuracy of signal acquisition and the stability of the system. And carrying out sampling time sequence alignment and feature matrix construction on the voltage input sequence, the current input sequence, the temperature input sequence and the power input sequence to obtain an input feature sequence matrix. The sampling timing alignment ensures time synchronicity of the signals so that the neural network can simultaneously receive input data with the same time stamp, and the feature matrix construction orderly organizes various input signals for subsequent neural network processing. The input feature sequence matrix is input into a first hidden layer of the neural network model, the first hidden layer containing 256 neuron nodes with a dropout mechanism. By carrying out nonlinear characteristic transformation through the ReLU activation function, the introduction of a dropout mechanism can effectively prevent the model from being over-fitted, and the generalization capability of the model is improved. The ReLU activation function performs nonlinear mapping on the input features, so that the network has the capability of processing complex features. The feature distribution is dynamically adjusted by adopting a batch normalization method, the batch normalization can accelerate model training and improve the stability of the model, and the feature distribution is ensured to be stable in the training process by normalizing the data of each batch. The first layer feature map is input into a second hidden layer of the neural network model, the hidden layer containing 128 neuron nodes with residual connections. Through the residual connection mode, the gradient vanishing problem in the deep network is effectively relieved, and the training efficiency of the network is improved. In the second hidden layer, the features are compressed by using a Sigmoid activation function, the Sigmoid activation function can compress the feature values to be in the range of 0 to 1, and the distinguishing capability of the model on the features is enhanced. At the same time, attention mechanisms are introduced in this layer to enhance the key features. The introduction of the attention mechanism enables the model to pay more attention to the characteristics which have important influence on the final output, and improves the understanding capability of the model on input data and the accuracy of output. And (3) inputting the second layer of characteristic mapping into an output layer of the neural network model, wherein the output layer comprises 64 fully-connected neuron nodes, and mapping light source parameters through a linear activation function to obtain a light source parameter dynamic response matrix. the function of the linear activation function is to maintain a linear relationship of the input and output in order to map the features of the previous layer to the final light source parameter output.
Step 200, adopting a principal component analysis algorithm to perform characteristic dimension reduction on the dynamic response matrix of the light source parameters, and constructing a target optimization model;
specifically, the mean value centering operation and the standard deviation normalization operation are performed on the dynamic response matrix of the light source parameters so as to obtain a standardized parameter matrix. The mean value centering operation can eliminate offset in the data, so that the data mean value of each parameter is zero, the stability of subsequent operation is improved, the dimension of each parameter is kept consistent by the standard deviation normalization operation, and the difference in dimension between various input features is eliminated. And performing covariance calculation and eigenvector decomposition on the standardized parameter matrix to obtain an eigenvector set representing the light source characteristic. The characteristic components include a light intensity response function, a color temperature response function, a power consumption response function, and the like, and each of the characteristic components reflects a characteristic of a change in a certain aspect of a light source parameter. The objective of covariance calculation is to obtain correlations between features so that important feature information is preserved in the subsequent dimension reduction, while feature vector decomposition is to map the features into new space so that their changes are easier to interpret and process in the new space. And calculating the characteristic value of each characteristic component based on the characteristic component set, and screening the main component characteristics according to the threshold value of the contribution rate of 95% to obtain a dimension-reducing mapping matrix. The eigenvalues reflect the interpretation ability of the corresponding eigenvalues on the original dataset, while the threshold of 95% contribution is to preserve as much important information as possible, while reducing the data dimension to achieve the purpose of dimension reduction. And performing orthogonal transformation operation on the standardized parameter matrix and the dimension reduction mapping matrix to obtain a low-dimensional feature vector containing the light intensity stability parameter, the spectrum fidelity parameter, the energy consumption efficiency parameter and the temperature balance parameter. The operation of orthogonal transformation maps the features in the high-dimensional space into a lower-dimensional space through a dimension-reduction mapping matrix, and different features are mutually independent in the new low-dimensional space, so that the subsequent parameter optimization process is facilitated. When the target optimization model is built, the deviation from a set threshold is calculated based on the light intensity stability parameter in the low-dimensional feature vector, and a light intensity optimization target item is built. The aim of the light intensity stability parameter is to ensure that the light intensity output by the light source can be kept stable, obvious fluctuation can not occur due to the change of external conditions, and the stability of the light source is ensured by setting a threshold value and controlling the deviation within an acceptable range. And meanwhile, calculating the matching degree with a standard spectrum based on the spectrum fidelity parameter in the low-dimensional feature vector, and constructing a spectrum optimization target item. The spectrum fidelity parameter reflects the difference between the output spectrum of the light source and the ideal standard spectrum, and the aim of constructing the spectrum optimization target is to ensure that the output spectrum of the light source can meet the expected light quality requirement so as to ensure the effect of the light source in practical application. And calculating the ratio of the energy consumption efficiency parameter in the low-dimensional feature vector to the input power to construct an energy consumption optimization target item. The energy consumption efficiency parameter reflects the efficiency of the light source in the energy conversion process, and the aim of constructing the energy consumption optimization target is to improve the energy utilization efficiency of the light source to the maximum extent, reduce unnecessary energy consumption and improve the economical efficiency and the environmental protection performance of the system. For the temperature balance parameter in the low-dimensional feature vector, a temperature balance target item is constructed based on the difference value between the temperature balance parameter and the maximum allowable temperature. The construction of the temperature balance target item aims at ensuring that the temperature of the light source does not exceed a preset safety threshold during the working process, so as to avoid equipment damage or potential safety hazards caused by overhigh temperature. And combining the light intensity optimization target item, the spectrum optimization target item, the energy consumption optimization target item and the temperature balance target item together to form a comprehensive target optimization model. Through the target optimization model, balance is carried out among different optimization targets, and a group of control parameters which are excellent in light intensity stability, spectrum fidelity, energy consumption efficiency, temperature balance and the like are obtained.
Step 300, grid division is carried out on the parameter space based on the target optimization model, and a steepest descent direction vector and a vertical search direction vector are respectively generated on the divided grid points to obtain a parameter optimizing direction set;
It should be noted that, the target optimization model is input into the multi-parameter grid generator to perform parameter interval division, so as to obtain a search boundary matrix including a light intensity parameter interval, a color temperature parameter interval, a power parameter interval and a temperature parameter interval, so that the parameters are optimized in a constraint manner, and the parameters are prevented from exceeding the physical allowable range. And equally dividing the search boundary matrix, and determining the minimum sampling interval according to the physical constraint of each parameter to obtain an initial grid sampling point matrix. A uniform sampling point grid is constructed in the parameter space so that the search process can cover the entire parameter space, while the minimum sampling interval is set according to the physical characteristics of each parameter to ensure the balance between sampling accuracy and computational overhead. On the basis, calculating the first derivative of the objective function at each sampling point based on the initial grid sampling point matrix to obtain a gradient vector group in the parameter space. The calculation of the first derivative can describe the change trend of the objective function at different parameter points and is the basis for determining the optimal direction. Orthogonalization decomposition operation is carried out on the gradient vector group, and base vectors of the local coordinate system are constructed. The purpose of orthogonalization decomposition is to eliminate the correlation between gradient vectors, so that the basis vectors of the local coordinate system are mutually independent, and the subsequent search direction generation is facilitated. And meanwhile, carrying out unitization processing on the constructed base vector to obtain an orthogonalization base vector set. The length of the basis vectors is normalized to perform equal trade-off optimization operations in different directions, ensuring that searches in each direction have the same importance. And inputting the orthogonalization basis vector set into a steepest descent method optimizer to calculate a negative gradient direction and obtain a steepest descent direction vector. The steepest descent method is an optimization method that searches along the negative gradient direction of the objective function, which is generally the direction in which the objective function value is most rapidly descended, so that the calculation of the steepest descent direction vector can effectively direct the rapid convergence of the parameters toward the optimization target. And carrying out projection operation of an orthogonal complement space on the steepest descent direction vector, generating a search direction orthogonal to the steepest descent direction through Schmidt orthogonalization processing, and obtaining a vertical search direction vector. The purpose of the projection operation of the orthogonal complement space and the Schmidt orthogonalization is to construct other search directions which are not related to the steepest descent direction, thereby avoiding sinking into local optimization and simultaneously increasing the diversity and flexibility of parameter optimization. In this way, the vertical search direction vector can provide a search means independent of the steepest descent direction, so that the whole parameter optimizing process has higher exploration capability. And inputting the steepest descent direction vector and the vertical search direction vector into a linear searcher to determine the optimal combination weight, thereby obtaining the mixed search direction. In this process, the linear searcher finds the combination of directions that makes the objective function drop fastest by performing a weighted combination between the steepest descent direction and the vertical search direction. The introduction of the hybrid search direction helps to overcome some of the drawbacks in the steepest descent method, such as the tendency to sink into the zig-zag path, thereby improving the efficiency of the optimization process. And correcting and updating the searching direction of the mixed searching direction through a conjugate gradient optimizer to obtain a final parameter optimizing direction set. The conjugate gradient optimizer is an improved optimization method, which can avoid the common repeated calculation path problem in the gradient method by introducing the conjugate direction, and quicken the convergence speed. Under the action of the conjugate gradient optimizer, the mixed search direction is corrected, so that the parameter optimizing direction has better global property and high efficiency.
Step 400, carrying out iterative updating on each control parameter by adopting a dynamic step length according to the parameter optimizing direction set to obtain a parameter iterative sequence set;
Specifically, the change trend analysis is carried out on the light intensity parameter, the color temperature parameter, the power parameter and the temperature parameter based on the parameter optimizing direction set, and the initial step size of each control parameter in the current optimizing direction is determined, so that an initial step control matrix is obtained. And carrying out objective function gradient calculation on each parameter in the initial step control matrix, and evaluating the sensitivity of the parameter to the objective optimization model under the current step to obtain a dynamic step factor. Through the step, the dynamic step factor adapts to the changing requirements of different parameters, and flexible step adjustment is realized, so that the dynamic step factor can be quickly converged in the optimizing process, and the missing of a global optimal solution can be avoided. Based on the dynamic step factor, damping attenuation coefficients are applied to each control parameter, certain inhibition is applied to parameter changes, unstable systems caused by overlarge parameter fluctuation are prevented, and step sizes are ensured to be gradually reduced in an iterative process so as to realize accurate convergence. The step convergence speed is adjusted through a feedback control algorithm, so that the step can be reduced in a self-adaptive manner when the optimal solution is approached, and the final convergence accuracy is improved. And after the adjustment of the feedback control algorithm, obtaining a step length updating vector, and describing the step length adjustment amplitude of each parameter in the current optimizing direction. And performing inner product operation on the step length updating vector and the parameter optimizing direction set to obtain candidate parameter updating quantity. The step size update amount is projected to the optimizing direction, thereby determining the actual update amplitude of each control parameter. In this way, the candidate parameter updating quantity can accurately reflect the change condition of each control parameter in the current iteration process, and a specific operand value is provided for subsequent parameter updating. And verifying the validity of the candidate parameter updating quantity, and cutting off the parameter based on hardware limitation of the input voltage range AC100-240V and the output voltage DC24V to obtain valid parameter increment. The purpose of validity verification is to check whether the candidate parameter updating quantity meets the physical and hardware constraint conditions of the system, so that the generated parameters cannot exceed the working range of the equipment in practical application. And the parameter cut-off based on the hardware limit ensures that the update amount of each control parameter works in the power supply and output limit, and avoids system failure or hardware damage caused by improper parameter setting. And linearly combining the effective parameter increment and the current iteration parameter through a dynamic weight distributor with a memory term to obtain a parameter iteration sequence set. The dynamic weight allocator with memory term maintains a portion of historical information during parameter updating to increase stability of parameter adjustment. The linear combination process makes the current parameters not only depend on the new effective parameter increment, but also considers the history parameters of previous iteration, thereby realizing smooth transition in the whole optimizing process and avoiding system oscillation caused by too intense parameter change. Through the process, the final parameter iteration sequence set contains the updating condition of each control parameter in each iteration, and the parameter iteration sequences can gradually approach the optimal solution of the target optimization model, so that the parameter self-adaptive adjustment of the simulated light source controller is realized.
Step 500, performing voltage stabilizing output test on each group of control parameters in the parameter iteration sequence set through a constant voltage controller, and calculating the comprehensive scores of a voltage stability index, an overcurrent protection index, a trigger response index and a temperature balance index by utilizing a multi-objective evaluation function to obtain a plurality of candidate control parameter combinations;
Specifically, the parameter iteration sequence set is input into a constant voltage controller, and the constant voltage controller carries out voltage stabilization treatment on the input voltage AC100-240V through a PWM rectification conversion circuit to obtain a voltage stabilization output sequence. The PWM rectification conversion circuit can rectify and filter an ac voltage to convert the ac voltage into a stable dc voltage for output. And carrying out voltage ripple detection and overcurrent threshold detection on the voltage-stabilized output sequence through a sampling circuit with a 2.5A measuring range to obtain single-channel operation characteristic data. Voltage ripple detection evaluates the smoothness of the output voltage and the quality of the power supply, and low ripple means good voltage regulation performance and helps to reduce interference to the load. The overcurrent threshold detection is used for monitoring whether the output current exceeds a set safety threshold or not so as to ensure that the system can operate in a safe current range and avoid damage to hardware. And inputting the single-channel operation characteristic data into a performance evaluation module, and performing quantitative calculation on four indexes of voltage stability, overcurrent protection, trigger response and temperature balance in the performance evaluation module to obtain a performance index matrix. The voltage stability index is used for quantifying the stability degree of output voltage under different working conditions and reflecting the inhibition capability of the voltage stabilizing controller to input voltage fluctuation, the overcurrent protection index is used for measuring whether the controller can respond timely and take protection measures when encountering an overcurrent condition, the trigger response index is used for evaluating the response speed and accuracy of the system to input signal change, and the temperature balance index is used for evaluating the temperature change condition during the system operation and ensuring that the temperature fluctuates within a reasonable range so as to prevent system faults or performance degradation caused by overheating. Through the quantitative calculation of the indexes, the performance index matrix can comprehensively reflect the performance of the control parameters in various aspects. And carrying out weighted summation on each index in the performance index matrix through a weight distributor to obtain a comprehensive performance score. The weight distributor gives different weights to the system overall performance according to the contribution degree of different indexes so as to ensure that the overall performance score can objectively reflect the overall quality of the control parameters. And arranging the comprehensive performance scores in descending order and setting a screening threshold value to obtain candidate control parameter combinations. The descending order is to select the control parameters with higher comprehensive scores preferentially, so as to ensure that the candidate combination has better performance on various performance indexes. And the setting of the screening threshold value is to eliminate the parameter combination with the comprehensive score not reaching the standard so as to reduce the range of the candidate set.
And 600, respectively inputting a plurality of candidate control parameter combinations into the multi-channel cooperative controller for parallel verification test, and determining an optimal control parameter combination according to the interference suppression degree among the channels.
Specifically, candidate control parameter combinations are respectively input into SMP-03V-BC ports of the four-channel 60W power controller, parallel operation tests are simultaneously carried out on the four channels, the actual operation environment of the light source controller during the multi-channel cooperative work is simulated, and stable control of the candidate control parameter combinations under the complex multi-channel condition is ensured. By collecting parameters such as output voltage, current, power and temperature of each channel, a multi-channel operation data matrix is formed, and the operation characteristics of four channels in parallel operation are reflected. And performing crosstalk detection on the multi-channel operation data matrix through the channel isolation operational amplifier, and evaluating signal interference conditions among the channels, namely judging whether signals of one channel can have adverse effects on other channels. In order to analyze the mutual influence among channels, the load proportion of each channel is calculated based on a dynamic power distribution algorithm, and interference characteristic data among channels is obtained. The dynamic power distribution algorithm dynamically adjusts the power distribution proportion of each channel by analyzing the load characteristic of each channel so as to consider the load condition of each channel in the interference characteristic analysis and provide a more accurate interference evaluation result. And inputting the interference characteristic data among the channels into a compensation controller, and inhibiting the interference among the channels through a zero drift compensation circuit and a temperature compensation network. The zero drift compensation circuit has the function of eliminating zero drift errors in signals, ensuring that the reference voltage of the circuit is stable and unchanged in the measuring and controlling processes, and improving the interference suppression precision. The temperature compensation network is used for compensating performance fluctuation caused by environmental temperature change in the running process of the system, so that the stability of the controller under different temperature conditions is ensured. Through the double compensation of zero drift and temperature, the compensation controller can effectively reduce the mutual interference among all channels and improve the anti-interference capability of the system. And curve correction is carried out on the compensation control parameters through a piecewise linearization circuit, and nonlinear compensation parameters are converted into linear forms so as to facilitate subsequent control calculation and adjustment. And after the linearization correction is completed, delay control is carried out on the compensation control parameters through triggering a delay compensation network, so as to obtain a multichannel cooperative control sequence. The purpose of triggering delay compensation is to adjust signal delay in compensation control, ensure that control signals of all channels can arrive synchronously, realize coordination control among the channels, and avoid mutual interference aggravation caused by signal asynchronization. And performing feature mapping on the multichannel cooperative control sequence and the candidate control parameter combination, and screening out the parameter combination with the strongest anti-interference capability through the minimum interference criterion. And comparing and analyzing the compensated control parameter with the candidate parameter, and finding out an optimal solution based on a specific criterion. The minimum interference criterion is used to evaluate the performance of the parameter combinations in terms of interference suppression, with the aim of finding a parameter combination that minimizes the interference between the channels. Through the process, the optimal control parameters which meet the requirements best are screened from the plurality of candidate control parameter combinations, and the optimal control parameter combinations are obtained.
According to the embodiment of the application, the parameter mapping and the principal component analysis dimension reduction are carried out through a neural network model, the accurate modeling of the complex dynamic characteristics of the simulated light source controller is realized, a reliable mathematical basis is provided for parameter optimization, a multi-objective optimization and grid search strategy is adopted, a plurality of control targets of light intensity stability, spectrum fidelity, energy consumption efficiency and temperature balance are simultaneously considered, the comprehensive performance of the system is obviously improved, a dynamic step self-adaptive adjustment mechanism is introduced, the control parameters can be automatically adjusted according to the change of the working environment, the environmental adaptability of the system is enhanced, inter-channel crosstalk is effectively restrained through a multi-channel cooperative control strategy, the problem of mutual interference during parallel operation of multiple channels is solved, the parameter optimization process is divided into a plurality of links such as feature extraction, target construction, parameter search and performance evaluation based on a hierarchical optimization structure, the convergence efficiency of an optimization algorithm is improved, a complete performance evaluation system is designed, the reliability and the practicability of an optimization result are ensured through the multi-objective evaluation function.
In a specific embodiment, the process of executing step 100 may specifically include the following steps:
The method comprises the steps of carrying out digital quantity conversion and phase compensation on input voltage signals acquired by an SMP-03V-BC port through a 16-bit high-precision analog-to-digital converter to obtain a voltage input sequence, carrying out current detection and baseline calibration on output current signals acquired by a constant voltage drive circuit through a Hall current sensor with a 2.5A range to obtain a current input sequence, carrying out temperature sampling and temperature compensation on environmental temperature signals acquired by a forced cooling controller through an NTC thermistor to obtain a temperature input sequence, carrying out power detection and channel load calculation on channel power signals of four channels through a channel isolation operational amplifier to obtain a power input sequence, and carrying out power detection and channel load calculation on the channel power signals of the four channels;
Carrying out sampling time sequence alignment and feature matrix construction on a voltage input sequence, a current input sequence, a temperature input sequence and a power input sequence to obtain an input feature sequence matrix;
Inputting an input feature sequence matrix into a first hidden layer of a neural network model, wherein the first hidden layer comprises 256 neuron nodes with dropout mechanisms, performing nonlinear feature transformation through a ReLU activation function, and dynamically adjusting feature distribution by adopting a batch normalization method to obtain a first layer feature map;
inputting the first layer of feature map into a second hidden layer of the neural network model, wherein the second hidden layer comprises 128 neuron nodes with residual connection, performing feature compression through a Sigmoid activation function, and performing enhancement processing on key features by using a attention mechanism to obtain a second layer of feature map;
and (3) inputting the second layer of characteristic mapping into an output layer of the neural network model, wherein the output layer comprises 64 fully-connected neuron nodes, and performing light source parameter mapping through a linear activation function to obtain a light source parameter dynamic response matrix.
Specifically, the input voltage signal collected by the SMP-03V-BC port is subjected to digital quantity conversion and phase compensation through a 16-bit high-precision analog-to-digital converter, and a voltage input sequence is obtained. The input voltage signal is a continuous analog signal, and the 16-bit high-precision analog-to-digital converter converts the analog signal into a corresponding digital signal. The high-precision analog-to-digital conversion ensures that details of the sampled signal are preserved, and the resolution of the sampled signal can express finer voltage variation, so that the sampling precision meets the requirements of a control system. Meanwhile, phase compensation is used to compensate for phase lag or lead caused by signal acquisition and system transmission, and the compensation process is described as:
;
Wherein,Representing the compensated voltage input signal,Is an uncompensated voltage signal that is,Is the correction amount for phase compensation. The output current signal collected by the constant voltage driving circuit is processed, and the process depends on a Hall current sensor with a 2.5A range. The hall current sensor is used for realizing current detection and simultaneously carrying out baseline calibration to eliminate baseline drift in the measurement process. The current signal detection is expressed as:
;
Wherein,Is the output current signal after calibration and,Is the current initially measured by the hall sensor,The offset due to sensor characteristics or environmental conditions is eliminated by the calibration process to improve measurement accuracy. Meanwhile, the environmental temperature signal is collected and processed, the environmental temperature signal is collected through the forced cooling controller, and the temperature sampling and the temperature compensation are carried out through the NTC thermistor, so that a temperature input sequence is obtained. The characteristic of the resistance value of the NTC thermistor that varies with temperature is used to obtain temperature information of the surrounding environment, and the purpose of temperature compensation is to ensure that the measured temperature signal truly reflects the monitored ambient temperature in the event of external temperature fluctuations. The description is as follows:
;
Wherein,In order to compensate for the temperature after the compensation,For a temperature signal to be measured directly,Is a temperature compensation coefficient.AndThe ambient temperature and the reference temperature, respectively. And for the channel power signals of four channels, the power detection is carried out through the channel isolation operational amplifier, so that the mutual interference among the channels is avoided. The power per channel is expressed as:
;
Wherein,In order to detect the resulting single channel power,AndChannel voltage and current, respectively. And obtaining load distribution by calculating the load condition of each channel for further power balance and management. And carrying out sampling time sequence alignment and feature matrix construction on the voltage input sequence, the current input sequence, the temperature input sequence and the power input sequence to obtain an input feature sequence matrix. The sample timing alignment ensures that all input signals have consistent time stamps on the same time axis so that each column in the input feature matrix represents a feature at the same point in time. Will input characteristic sequence matrixThe first hidden layer is input to the neural network model, and comprises 256 neuron nodes with dropout mechanisms, and nonlinear feature transformation is performed through a ReLU activation function. The function of the ReLU activation function is to introduce non-linear characteristics that enable the network to handle more complex input features. The expression is as follows:
;
The dropout mechanism prevents network overfitting by randomly setting the output of a portion of the neuron nodes to 0, improving the generalization ability of the model. Meanwhile, a batch normalization method is adopted to dynamically adjust the characteristic distribution, and the expression of batch normalization is as follows:
;
Wherein,AndRepresenting the mean and variance of the small batch input respectively,Is a very small value for avoiding division by 0. The introduction of batch normalization helps to speed up training and improve network stability. The first layer features are input to a second hidden layer of the neural network, which contains 128 neuron nodes with residual connections, and feature compression is performed by a Sigmoid activation function. The Sigmoid activation function compresses the input value between 0 and 1, expressed as:
;
This allows the output eigenvalues to fall within a fixed range, helping to maintain the numerical stability of the model. The introduction of residual connection enables the original input to pass directly to the next layer bypassing some hidden layers, alleviating the gradient vanishing problem in the depth network, the residual connection being expressed as:
;
Wherein,Is an input to which the user is exposed,Is the output of the hidden layer. Meanwhile, the key features are enhanced by using an attention mechanism, and the importance of different features is calculated and weighted and amplified, so that the model can pay more attention to the features with obvious influence on the final output. And inputting the second layer of characteristic mapping into an output layer of the neural network model, wherein the output layer comprises 64 fully-connected neuron nodes, and mapping the light source parameters through a linear activation function to obtain a light source parameter dynamic response matrix. The linear activation function is used to maintain a linear relationship between the input and the output, and is expressed as:
;
The dynamic response matrix of the light source parameters comprises response characteristics of the light source under input conditions such as input voltage, current, temperature, power and the like.
In a specific embodiment, the process of performing step 200 may specifically include the following steps:
performing mean value centering operation and standard deviation normalization operation on the dynamic response matrix of the light source parameters to obtain a standardized parameter matrix;
covariance calculation and eigenvector decomposition are carried out on the standardized parameter matrix, and a characteristic component set for representing a light intensity response function, a color temperature response function, a temperature response function and a power consumption response function is obtained;
Calculating the characteristic value of each characteristic component based on the characteristic component set, and screening the main component characteristics according to the threshold value of the contribution rate of 95% to obtain a dimension-reducing mapping matrix;
performing orthogonal transformation operation on the standardized parameter matrix and the dimension reduction mapping matrix to obtain a low-dimensional feature vector comprising a light intensity stability parameter, a spectrum fidelity parameter, an energy consumption efficiency parameter and a temperature balance parameter;
calculating deviation from a set threshold value based on the light intensity stability parameter in the low-dimensional feature vector, and constructing a light intensity optimization target item; calculating the matching degree between the spectrum fidelity parameter in the low-dimensional feature vector and the standard spectrum, and constructing a spectrum optimization target item, calculating the ratio between the energy consumption efficiency parameter in the low-dimensional feature vector and the input power, and constructing an energy consumption optimization target item;
And combining the light intensity optimization target item, the spectrum optimization target item, the energy consumption optimization target item and the temperature balance target item into a target optimization model.
Specifically, the mean value centering operation and the standard deviation normalization operation are carried out on the dynamic response matrix of the light source parameters, so as to obtain a standardized parameter matrix. The offset of each parameter in the data is eliminated through the mean value centering, so that the mean value of each feature is zero, the subsequent calculation is more stable, and the formula is as follows:
;
Wherein,Is a dynamic response matrix of the parameters of the light source,Is the mean vector for each column in the matrix,Then it is the matrix after centering. And then carrying out standard deviation normalization operation to make the scales of the data consistent so as to eliminate the influence caused by different sizes among features, wherein the formula is as follows:
;
Wherein,Is a vector of the standard deviation of the values,Is normalized parameter matrix. And (3) preprocessing to obtain standardized data of each feature under the same scale. Covariance calculation and eigenvector decomposition are performed on the standardized parameter matrix to obtain a set of eigenvectors characterizing the light intensity response function, the color temperature response function, the temperature response function, and the power consumption response function. Covariance matrixThe calculation formula of (2) is as follows:
;
Wherein,For the number of samples to be taken,Is the transpose of the normalized parameter matrix. Covariance matrixThe linear relationship between the features is described for judging the correlation between the features. Performing eigenvector decomposition on the covariance matrix:
;
Wherein,Is the eigenvector of the covariance matrix,Is the corresponding characteristic value. The eigenvectors represent the principal directions of the covariance matrix, the eigenvalues represent the variances of the directions, and the eigenvalues are combined with the eigenvectors to obtain an eigenvector set for describing the light intensity, the color temperature, the temperature and the power consumption response. And calculating the characteristic value of each characteristic component based on the characteristic component set, and screening the characteristic of the main component according to the threshold value of the contribution rate of 95% to obtain the dimension-reducing mapping matrix. The contribution rate is the duty ratio of the eigenvalues in all eigenvalues, expressed as:
;
Wherein,Is the front selectedThe value of the characteristic is a value of,Is the total number of eigenvalues. When the cumulative contribution rate reaches 95%, selecting the corresponding feature vector as the main component feature to construct a dimension-reducing mapping matrixEach of which is a principal component feature vector. Mapping the standardized parameter matrix and the dimension reduction map matrixPerforming orthogonal transformation operation to obtain low-dimensional eigenvector comprising light intensity stability parameter, spectrum fidelity parameter, energy consumption efficiency parameter and temperature balance parameter:
;
Wherein,Is a low-dimensional feature vector, wherein each element represents different key characteristics of light source control, including a light intensity stability parameter, a spectrum fidelity parameter, an energy consumption efficiency parameter and a temperature balance parameter. The orthogonal transformation maps the original high-dimensional data into a space with lower dimension through the dimension-reducing mapping matrix, and meanwhile, the most important features are reserved, so that the subsequent optimization calculation is simplified. And constructing a target optimization model based on each parameter in the low-dimensional feature vector. Based on the light intensity stability parameter in the low-dimensional feature vectorCalculating the deviation between the light intensity and a set threshold value, and constructing a light intensity optimization target item:
;
Wherein,Is a set target value of the light intensity,Is an optimization target item of light intensity. The target item measures the difference between the actual light intensity stability and the set target, and by minimizing the difference, the light intensity output by the light source is ensured to be stable within the set range. Spectral fidelity parameters in low-dimensional feature vectorsCalculating the matching degree with the standard spectrum to construct a spectrum optimization target item:
;
Wherein,Represents the first of the standard spectrumThe number of components of the composition,Is the number of components of the spectrum,Representing the spectral optimization target term. By minimizing the error between the spectral fidelity parameter and the standard spectrum, the spectral output of the light source is ensured to match the expected standard spectrum, thereby improving the light quality of the light source. Based on energy consumption efficiency parameters in low-dimensional feature vectorsCalculate its and input powerConstructing an energy consumption optimization target item:
;
The energy conversion efficiency is maximized by minimizing negative energy consumption optimization target items, and the light source controller can work under the energy efficiency as high as possible, so that the energy consumption is reduced, and the economical efficiency and the environmental protection performance of the whole system are improved. Temperature balance parameters in low-dimensional feature vectorsCalculating the maximum allowable temperatureConstructing a temperature balance target term:
;
By minimizing the deviation between the temperature parameter and the maximum allowable temperature, the light source controller is ensured to work within a reasonable temperature range, and equipment damage or performance degradation caused by overheating is avoided. Through the steps, the light intensity optimization target item, the spectrum optimization target item, the energy consumption optimization target item and the temperature balance target item are combined to finally obtain a target optimization model:
;
Wherein,The weight coefficients of the respective object items are used to balance the relative importance between the different optimization objectives. By adjusting these weight coefficients, specific performance metrics are optimized according to different application requirements. For example, if the energy efficiency requirement is high in a certain application scene, the increaseTo more tend to improve energy consumption efficiency.
In a specific embodiment, the process of executing step 300 may specifically include the following steps:
inputting the target optimization model into a multi-parameter grid generator for parameter interval division to obtain a search boundary matrix comprising a light intensity parameter interval, a color temperature parameter interval, a power parameter interval and a temperature parameter interval;
equally-spaced division is carried out on the search boundary matrix, the minimum sampling interval is determined according to the physical constraint of each parameter, an initial grid sampling point matrix is obtained, and the first derivative of an objective function at each sampling point is calculated based on the initial grid sampling point matrix, so that a gradient vector group in a parameter space is obtained;
carrying out orthogonalization decomposition operation on the gradient vector group, constructing a base vector of a local coordinate system, and carrying out unitization treatment to obtain an orthogonalization base vector set;
Inputting the orthogonalization basis vector set into a steepest descent method optimizer for negative gradient direction calculation to obtain a steepest descent direction vector;
Carrying out orthogonal complementary space projection operation on the steepest descent direction vector, generating a search direction orthogonal to the steepest descent direction vector through Schmidt orthogonalization processing, and obtaining a vertical search direction vector;
and inputting the steepest descent direction vector and the vertical search direction vector into a linear searcher to determine the optimal combination weight, obtaining a mixed search direction, and correcting and updating the search direction of the mixed search direction through a conjugate gradient optimizer to obtain a parameter optimizing direction set.
Specifically, the target optimization model is input into a multi-parameter grid generator, and the parameter intervals are reasonably divided. The target optimization model contains optimization requirements of light intensity parameters, color temperature parameters, power parameters and temperature parameters, and the multi-parameter grid generator is used for dividing the parameters into different sections to form a search boundary matrix containing parameter ranges. Assume that the intervals of the light intensity parameter, the color temperature parameter, the power parameter and the temperature parameter are respectivelyThe search boundary matrix is expressed as:
;
Wherein the matrixUpper and lower boundaries describing light intensity, color temperature, power and temperature,AndThe lower limit and the upper limit of each parameter are respectively indicated. The value range of the parameter is defined so as to ensure that the subsequent searching process is carried out within a reasonable boundary and avoid exceeding the physical limit of the equipment. And equally dividing the search boundary matrix to construct an initial grid sampling point matrix. When equidistant dividing is performed, the minimum sampling interval is determined according to the physical constraint of each parameter, and the interval is assumed to be for a certain parameterMinimum sampling interval isThe expression for the sample point is expressed as:
;
Wherein,For the number of sampling points,Represent the firstAnd sampling points. After all parameters are equally divided in this way, an initial grid sampling point matrix is obtainedThe matrix contains possible values of the parameters in the parameter space. The first derivative of the objective function at each sampling point is calculated based on the initial grid sampling point matrix to obtain a set of gradient vectors in the parameter space. The first derivative of the objective function describes the change trend of the function at the current point, and the objective function is assumed to beWhereinRepresenting a parameter vector whose gradient vector is represented as:
;
Wherein,Respectively, represent different parameters of the device,Representing the objective function at the pointGradient vector at. Obtaining a gradient vector group by performing first derivative calculation on each sampling point in an initial grid sampling point matrixEach gradient vector describes the direction of the functional change at a particular point in the parameter space. Orthogonalization decomposition operation is carried out on the gradient vector group so as to construct base vectors of the local coordinate system, and unit processing is carried out on the base vectors to obtain an orthogonalization base vector set. Assume the original gradient vector set isThrough orthogonalization decomposition, an orthogonalization base vector group is obtainedWherein each basis vector is mutually orthogonal. The orthogonalization process is realized by a gram-schmitt method, and the recurrence relation is as follows:
;
after the orthogonal basis vectors are obtained, they are unitized to ensure that the length of all basis vectors is 1:
;
Wherein,Is the orthonormal basis vector after unitization,Representing vectorsIs a die length of the die. Through this procedure, an orthogonalized basis vector set is obtained for precisely describing the direction of variation of the parameters in the local coordinate system. And inputting the orthogonalization basis vector set into a steepest descent method optimizer to calculate a negative gradient direction so as to obtain a steepest descent direction vector. The steepest descent method searches along the negative gradient direction of the objective function to most rapidly reduce the value of the objective function. The steepest descent direction is expressed as:
;
Wherein,The steepest descent direction vector is indicated, along which the value of the objective function can be rapidly reduced, directing the parameter to move toward the optimum point. And carrying out orthogonal complementary space projection operation on the steepest descent direction vector so as to obtain other search directions. By orthographically projecting the steepest descent direction vector, a direction not associated therewith can be obtained, avoiding the optimization process from falling into local optimum. In order to ensure that the newly generated direction is orthogonal to the steepest descent direction, a Schmidt orthogonalization process is employed to generate a search direction orthogonal to the steepest descent direction, assuming that the direction isThe expression is:
;
Through this process, a product is obtainedThe direction is orthogonal to the steepest descent direction, so that more exploration dimensions are provided for parameter optimization. And inputting the steepest descent direction vector and the vertical search direction vector into a linear searcher to determine the optimal combination weight, thereby obtaining the mixed search direction. The hybrid search direction is a weighted combination of the steepest descent direction and the vertical search direction, and the formula is:
;
Wherein,AndThe combination weights in the steepest descent direction and the vertical direction are respectively, and an optimal combination weight capable of minimizing the objective function value is found by linear search. And correcting and updating the searching direction of the mixed searching direction through a conjugate gradient optimizer to obtain a final parameter optimizing direction set. Conjugate gradient optimization is an efficient iterative optimization method, and by conjugate of a new direction and a previous search direction, repeated search along the same direction is avoided, and convergence speed and optimization efficiency are improved. Assume that the current search direction isThe updated search direction isThe recurrence relation of the conjugate gradient is:
;
Wherein,As conjugate coefficients, it is generally calculated as:
;
after the mixed search direction is corrected by the conjugate gradient optimizer, a parameter optimizing direction set is finally obtainedThese directions are used to direct the updating of the control parameters towards the optimal values to achieve global optimization of the target optimization model.
In a specific embodiment, the process of performing step 400 may specifically include the following steps:
carrying out change trend analysis on the light intensity parameter, the color temperature parameter, the power parameter and the temperature parameter based on the parameter optimizing direction set to obtain an initial step control matrix;
performing objective function gradient calculation on each parameter in the initial step control matrix to obtain a dynamic step factor;
applying damping attenuation coefficients to each control parameter based on the dynamic step factors, and adjusting the step convergence speed through a feedback control algorithm to obtain a step update vector;
Performing inner product operation on the step length updating vector and the parameter optimizing direction set to obtain candidate parameter updating quantity;
verifying the validity of the candidate parameter updating quantity, and carrying out parameter truncation based on hardware limitation of an input voltage range AC100-240V and an output voltage DC24V to obtain an effective parameter increment;
And linearly combining the effective parameter increment and the current iteration parameter through a dynamic weight distributor with a memory term to obtain a parameter iteration sequence set.
Specifically, based on the parameter optimizing direction set, the change trend analysis is carried out on the light intensity parameter, the color temperature parameter, the power parameter and the temperature parameter, and an initial step control matrix is obtained. Parameter optimizing direction setIn the process,Respectively represent the optimizing directions of the light intensity, the color temperature, the power and the temperature parameters. Analyzing the trend of each parameter in these directions to determine an initial step control matrix. Let each parameter change in the initial step length direction beThe initial step control matrix is expressed as:
;
Wherein the method comprises the steps ofRepresents the firstSample numberStep sizes of the parameters. The initial step sizes of all parameters in different optimizing directions are known through the change trend analysis. And carrying out gradient calculation of an objective function on each parameter in the initial step control matrix to obtain a dynamic step factor. Assuming an objective function ofWhereinRepresenting the light intensity, color temperature, power and temperature parameters, the gradient of the objective function at each step is expressed as:
;
the gradient of each parameter is calculated according to the objective function value of each parameter in the unsynchronized rectangular direction to obtain a dynamic step factorThe expression is:
;
The step length adjustment quantity required by each control parameter in different optimizing directions is represented, and the change trend of the objective function in the current direction and the sensitivity to the step length are reflected. Based on the dynamic step factor, damping attenuation coefficients are applied to each control parameter to reduce instability caused by an oversized step, and the step convergence speed is adjusted through a feedback control algorithm. Damping attenuation coefficientThe function of (2) is to attenuate the step size appropriately at each iteration so that the step size is gradually reduced to ensure final convergence to the optimum point. Assume that the current step size is updated toThe step update after damping decay is expressed as:
;
Wherein the method comprises the steps ofTypically a positive number less than 1. And (3) adjusting the convergence speed of the step length by using a feedback control algorithm, dynamically adjusting the step length by monitoring the change of the objective function value, gradually reducing the step length if the change of the objective function is slow, and otherwise, increasing the step length to improve the efficiency of the optimization process. Through these processes, the final step update vector is obtained. And performing inner product operation on the step length updating vector and the parameter optimizing direction set to obtain candidate parameter updating quantity. Setting the step update vector asThe parameter optimizing direction set isThe candidate parameter update amount is expressed as:
;
Wherein,The parameter updating amplitude in the current optimizing direction is represented, the step length and the optimizing direction are combined through inner product operation, and the updating quantity of each parameter is determined. And verifying the validity of the candidate parameter updating quantity. Considering that the input voltage is in the range ofAnd output voltageTo truncate the candidate parameter update amount to obtain effective parameter increment. The purpose of validity verification is to ensure that all parameters are not updated beyond the safe working range of hardware, thereby ensuring the stability and the safety of the system. The parameter cutoff is expressed as:
;
Wherein,AndParameters respectivelyIs defined as the lower limit and the upper limit of (c),Representing the effective parameter delta after the cutoff. And linearly combining the effective parameter increment and the current iteration parameter through a dynamic weight distributor with a memory term to obtain a parameter iteration sequence set. The dynamic weight distributor with the memory item has the function of retaining a part of historical information so as to improve the stability of parameter updating and prevent excessive fluctuation of parameters in the updating process. Assume that the parameters of the current iteration areThe effective parameter increment isParameters of the next iterationExpressed as:
;
Wherein,Is a memory coefficient with a value range ofFor controlling the weight ratio between the historical parameter and the new parameter increment. When (when)When larger, the representation is more prone to retaining historical information, and whenSmaller, then the parameter delta is indicated as more prone to accepting new updates.
In a specific embodiment, the process of performing step 500 may specifically include the following steps:
inputting the parameter iteration sequence set into a constant voltage controller, and performing voltage stabilization treatment on the input voltage AC100-240V through a PWM rectification conversion circuit to obtain a voltage stabilization output sequence;
Performing voltage ripple detection and overcurrent threshold detection on the voltage-stabilized output sequence through a sampling circuit with a 2.5A measuring range to obtain single-channel operation characteristic data;
Inputting the single-channel operation characteristic data into a performance evaluation module, and carrying out quantization calculation on four indexes of voltage stability, overcurrent protection, trigger response and temperature balance to obtain a performance index matrix;
Each index in the performance index matrix is weighted and summed through a weight distributor to obtain a comprehensive performance score;
And arranging the comprehensive performance scores in descending order and setting a screening threshold value to obtain candidate control parameter combinations.
Specifically, the parameter iteration sequence set is input into a constant voltage controller, and the input voltage is input through a PWM rectification conversion circuitAnd performing voltage stabilization treatment. The range of the input voltage is AC100-240V, and the PWM rectification conversion circuit is used for converting the alternating current input voltage into stable direct current output voltage. PWM (pulse width modulation) technology controls the average value of the output voltage to be kept constant by rapidly switching on and off of the input voltage. The PWM rectification process is expressed as:
;
Wherein,Is the proportionality coefficient of the rectification transformation,Is duty cycle, by adjusting the duty cycleRegulating output voltageTo achieve a stable target value. In the case that the input voltage may fluctuate, the PWM circuit can effectively compensate for these fluctuations, resulting in a regulated output sequence. And performing voltage ripple detection and overcurrent threshold detection on the voltage-stabilized output sequence through a sampling circuit with a 2.5A measuring range to obtain single-channel operation characteristic data. Voltage ripple detection monitors whether there is an irregular small amplitude oscillation in the output voltage, which can interfere with the load of the system, thereby affecting the stability of the system. Assume that the voltage ripple has a magnitude ofThe quantitative calculation formula is as follows:
;
Wherein,AndRepresenting the maximum and minimum values of the output voltage respectively,Representing the ripple amplitude of the output voltage. By detectingJudging whether the voltage stabilizing effect is good or not, and ensuring the output stability of the system. The overcurrent threshold detection can determine whether the output current exceeds a safe working range, and damage to the system due to excessive current is prevented. Assuming that the output current isA safe maximum current threshold ofThe conditions for overcurrent detection are:
;
if the current is outputExceeding the limitThe over-current protection mechanism is triggered to limit the current magnitude. The output current is monitored in real time through a sampling circuit with a 2.5A measuring range, so that the output current is effectively ensured to be in a safe range, and further damage to a control system is avoided. And inputting the single-channel operation characteristic data into a performance evaluation module to quantitatively calculate four indexes of voltage stability, overcurrent protection, trigger response and temperature balance, so as to obtain a performance index matrix. Index of voltage stabilityThe change amplitude of the output voltage under different working conditions is reflected, and the calculation formula is as follows:
;
Wherein,In order to evaluate the time interval of the time,For the target voltage to be the target voltage,Is time ofOutput voltage at time. And calculating the voltage deviation in the whole time interval through integration to obtain a quantization result of the voltage stabilizing effect. Overcurrent protection indexIt is used to quantify the validity of the over-current protection function with a value of 0 or 1, indicating the absence and presence of over-current, respectively. Trigger response indexRepresenting the response speed of the system to the input signal, assuming that the input signal changes in timeOccurs at the moment and the time required for the output of the system to reach steady state isThe trigger response index is calculated as:
;
the index reflects the quick response capability of the system, and a larger value indicates a faster system response. Index of temperature balanceThen it is used to quantify the stability of the system temperature during operation, assuming that the system temperature isThe target temperature isThe temperature balance index is calculated as:
;
by calculating the four indexes, a performance index matrix is obtainedIt is expressed as:
;
Performance index matrixReflecting the specific performance of the voltage stabilizing output sequence on various performance indexes. And carrying out weighted summation on each index in the performance index matrix through a weight distributor to obtain a comprehensive performance score. Set the weight coefficient asThen the overall performance scoreThe calculation is as follows:
;
Wherein the weight coefficientThe method is used for measuring the relative importance of each index in the overall performance, and weights of the indexes are different in different application scenes. For example, in some cases, the voltage stability of the system is more important than the response speed, at which time the settings are made. Combining the quantized results of each index in a weighted summation mode to obtain a comprehensive performance evaluation valueTo comprehensively reflect the overall operational performance of the system. The comprehensive performance scores are arranged in descending order, and a screening threshold is set to obtain candidate control parameter combinations. Assume the aggregate of the composite performance scores isThese scores are arranged in descending order:
;
Setting a screening thresholdScreening out all the comprehensive performance scores greater than or equal toIs expressed as:
;
through the screening process, a group of candidate control parameter combinations are obtained, the combinations perform well on various performance indexes, and the optimization requirement of a system can be met.
In a specific embodiment, the process of executing step 600 may specifically include the following steps:
inputting the candidate control parameter combinations into SMP-03V-BC ports of a four-channel 60W power controller respectively, and simultaneously performing parallel operation test on four channels to obtain a multi-channel operation data matrix;
Performing crosstalk detection on the multi-channel operation data matrix through a channel isolation operational amplifier, and calculating the load proportion of each channel based on a dynamic power distribution algorithm to obtain inter-channel interference characteristic data;
Inputting the inter-channel interference characteristic data into a compensation controller, and performing interference suppression processing through a zero drift compensation circuit and a temperature compensation network to obtain compensation control parameters;
curve correction is carried out on the compensation control parameters through a piecewise linearization circuit, and delay control is carried out through a trigger delay compensation network, so that a multichannel cooperative control sequence is obtained;
And performing feature mapping on the multichannel cooperative control sequence and the candidate control parameter combination, and screening out the parameter combination with the strongest anti-interference capability through a minimum interference criterion to obtain the optimal control parameter combination.
Specifically, candidate control parameter combinations are respectively input into SMP-03V-BC ports of the four-channel 60W power controller, so that parallel operation tests are simultaneously carried out on the four channels, and a multi-channel operation data matrix is obtained. Setting the power control parameter combination of each channel asThese parameters are input to the respective channel power controllers through SMP-03V-BC ports to achieve independent regulation and control of the respective channels. The four-channel parallel operation test aims to simulate the multi-channel collaborative operation environment of the system in actual operation, and ensure that the candidate control parameter combination can provide ideal control effect under complex working conditions. Through testing, the data of the four channels in the running process is recorded to form a multi-channel running data matrixIt is expressed as:
;
Wherein,Represent the firstIndividual point-in-time channelsIs provided. The crosstalk detection is performed on the multi-channel operation data matrix through the channel isolation operational amplifier (operational amplifier), and whether signals of one channel interfere with other channels is analyzed, and especially, the interference can significantly influence the performance of the system under the conditions of high power and parallel operation of the multi-channels. With channelsThe output of (2) isTo the channelGenerated crosstalkBy detecting the interference component in its output:
;
Wherein,Is a channel without interferenceIs provided with an output of (a),Representative channelPaired channelsIs a disturbance level of (a). After crosstalk detection, the load proportion of each channel is calculated based on a dynamic power distribution algorithm so as to obtain interference characteristic data among channels. The dynamic power distribution algorithm dynamically adjusts the distribution of power according to the load condition of each channel so as to reduce the mutual interference among the channels as much as possible and balance the power load of each channel. Assume the firstThe power output of each channel isThe total power of the system isChannelsLoad ratio of (2)Expressed as:
;
calculating the load ratio of all channelsAnd obtaining characteristic data reflecting the load condition of each channel. And inputting the interference characteristic data among the channels into a compensation controller, and performing interference suppression processing through a zero drift compensation circuit and a temperature compensation network. The zero drift compensation circuit can eliminate dc offset generated during operation of the amplifier or sensor, which can affect control accuracy and signal authenticity. Zero drift compensation is expressed as:
;
Wherein,Is an uncompensated output that is output,Is zero drift offset by subtracting the offsetA more accurate output signal is obtained. The temperature compensation network is used to compensate for performance fluctuations due to ambient temperature variations, the temperature compensation being expressed by the following formula:
;
Wherein,Is a temperature compensation coefficient, and the temperature of the liquid crystal display device is controlled by the temperature compensation coefficient,AndThe system can maintain relatively stable output under different temperature conditions through temperature compensation. And after zero drift compensation and temperature compensation treatment, compensation control parameters are obtained. And carrying out curve correction on the compensation control parameters through a piecewise linearization circuit, and carrying out delay control through a trigger delay compensation network to obtain a multichannel cooperative control sequence. The piecewise linearization circuit approximates the nonlinear compensation parameter to be multi-piecewise linear for subsequent control calculation, and the compensated parameter is set asThen by piecewise linearization is expressed as:
;
Wherein,AndRespectively the firstThe slope and intercept of the segment,The range of the interval of the linearization process is shown. By approximating the compensated parameters to a linear form, the control process can be made simpler and predictable. After piecewise linearization, the compensation control parameters are subjected to delay control through a trigger delay compensation network. Triggering delay compensation ensures control signal synchronization between multiple channels and avoids mutual interference aggravation caused by delay desynchronization. Let the delay time beThe delayed output is expressed as:
;
And through delay compensation, signals of all channels arrive synchronously to form a multi-channel cooperative control sequence so as to improve the coordination capacity and control precision of the system. And performing feature mapping on the multichannel cooperative control sequence and the candidate control parameter combination, and screening out the parameter combination with the strongest anti-interference capability through the minimum interference criterion. The feature map is a relationship of the compensation control sequence to the original candidate parameters analyzed to determine which combinations of parameters perform best in suppressing inter-channel interference. Characterised by interferenceThe anti-interference index obtained after feature mapping isThe goal of the minimum interference criterion is to find the minimum interference criterion such thatThe smallest combination of parameters, expressed as:
;
Wherein,Representing candidate control parameter combinations. By countering interference indicatorsAnd (3) screening out the parameter combination with the minimum interference characteristic to obtain the optimal control parameter combination.
The method for adaptively adjusting parameters of an analog light source controller in the embodiment of the present application is described above, and the device 10 for adaptively adjusting parameters of an analog light source controller in the embodiment of the present application is described below, referring to fig. 2, an embodiment of the device 10 for adaptively adjusting parameters of an analog light source controller in the embodiment of the present application includes:
The acquisition module 11 is used for acquiring an input voltage signal, an output current signal, an ambient temperature signal and a channel power signal of the analog light source controller, inputting the signals into the neural network model for parameter mapping, and obtaining a light source parameter dynamic response matrix;
The dimension reduction module 12 is used for carrying out characteristic dimension reduction on the dynamic response matrix of the light source parameters by adopting a principal component analysis algorithm and constructing a target optimization model;
The dividing module 13 is used for carrying out grid division on the parameter space based on the target optimization model, and respectively generating a steepest descent direction vector and a vertical search direction vector on the divided grid points to obtain a parameter optimizing direction set;
The updating module 14 is configured to iteratively update each control parameter by using a dynamic step according to the parameter optimizing direction set, to obtain a parameter iteration sequence set;
The calculation module 15 is configured to perform a voltage stabilizing output test on each group of control parameters in the parameter iteration sequence set through the constant voltage controller, and calculate a comprehensive score of a voltage stability index, an overcurrent protection index, a trigger response index and a temperature balance index by using a multi-objective evaluation function, so as to obtain a plurality of candidate control parameter combinations;
the test module 16 is configured to input a plurality of candidate control parameter combinations into the multi-channel cooperative controller for parallel verification test, and determine an optimal control parameter combination according to the interference suppression degree between the channels.
Through the cooperative cooperation of the components, parameter mapping and principal component analysis dimension reduction are carried out through a neural network model, so that accurate modeling of the complex dynamic characteristics of the simulated light source controller is realized, and a reliable mathematical basis is provided for parameter optimization; the system adopts a multi-objective optimization and grid search strategy, simultaneously gives consideration to a plurality of control targets of light intensity stability, spectrum fidelity, energy consumption efficiency and temperature balance, remarkably improves the comprehensive performance of the system, introduces a dynamic step length self-adaptive adjustment mechanism, enables control parameters to be automatically adjusted according to the change of a working environment, enhances the environmental adaptability of the system, effectively inhibits the inter-channel crosstalk through a multi-channel cooperative control strategy, solves the problem of mutual interference during multi-channel parallel operation, divides the parameter optimization process into a plurality of links such as feature extraction, target construction, parameter search and performance evaluation based on a layered optimization structure, improves the convergence efficiency of an optimization algorithm, designs a complete performance evaluation system, comprehensively evaluates the control parameters through a multi-objective evaluation function, and ensures the reliability and practicability of an optimization result.
Referring to fig. 3, fig. 3 is a schematic block diagram of an electronic device 300 according to an embodiment of the present application, where the electronic device 300 includes a processor 301 and a memory 302, and the processor 301 and the memory 302 are connected through a device bus 303, and the memory 302 may include a nonvolatile storage medium and an internal memory.
The non-volatile storage medium may store a computer program. The computer program comprises program instructions which, when executed by the processor 301, cause the processor 301 to perform any of the above-described methods of adaptively adjusting parameters of a simulated light source controller.
The processor 301 is used to provide computing and control capabilities to support the operation of the overall electronic device 300.
The internal memory provides an environment for the execution of a computer program in a non-volatile storage medium, which when executed by the processor 301, causes the processor 301 to perform any of the above-described methods of adaptively adjusting parameters of an analog light source controller.
It will be appreciated by those skilled in the art that the structure shown in fig. 3 is merely a block diagram of a portion of the structure associated with the present inventive arrangements and is not limiting of the electronic device 300 associated with the present inventive arrangements, and that a particular electronic device 300 may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
It should be appreciated that the Processor 301 may be a central processing unit (Central Processing Unit, CPU), the Processor 301 may also be other general purpose processors, digital signal processors (DIGITAL SIGNAL Processor, DSP), application SPECIFIC INTEGRATED Circuit (ASIC), field-Programmable gate array (Field-Programmable GATE ARRAY, FPGA) or other Programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like. Wherein the general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
It should be noted that, for convenience and brevity of description, the specific working process of the electronic device 300 described above may refer to the corresponding process of the foregoing adaptive adjustment method for parameters of the analog light source controller, which is not described herein.
Embodiments of the present application also provide a computer readable storage medium storing a computer program, where the computer program when executed by one or more processors causes the one or more processors to implement a method for adaptively adjusting parameters of an analog light source controller according to the embodiments of the present application.
The computer readable storage medium may be an internal storage unit of the electronic device 300 of the foregoing embodiment, for example, a hard disk or a memory of the electronic device 300. The computer readable storage medium may also be an external storage device of the electronic device 300, such as a plug-in hard disk, a smart memory card (SMART MEDIA CARD, SMC), a Secure Digital (SD) card, a flash memory card (FLASH CARD), etc. that are provided with the electronic device 300.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, which are not repeated herein.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in whole or in part in the form of a software product stored in a storage medium, comprising several instructions for causing an electronic device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. The storage medium includes a U disk, a removable hard disk, a read-only memory (ROM), a random access memory (random access memory, RAM), a magnetic disk, an optical disk, or other various media capable of storing program codes.
While the application has been described in detail with reference to the foregoing embodiments, it will be understood by those skilled in the art that the foregoing embodiments may be modified or equivalents may be substituted for some of the features thereof, and that the modifications or substitutions do not depart from the spirit and scope of the embodiments of the application.