Detailed Description
In order that the above objects, features and advantages of the invention will be readily understood, a more particular description of the invention will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings.
The embodiment of the invention provides an intelligent scheduling system of a coal mine transportation chain based on machine learning, which is shown in a modularized structural diagram of the intelligent scheduling system of the coal mine transportation chain based on machine learning in the figure 1.
The monitoring and scheduling layer is used for acquiring coal mine transportation chain data, generating a predicted value corresponding to the coal mine transportation chain data according to the coal mine transportation chain data, generating a corresponding scheduling instruction according to the predicted value and sending the scheduling instruction to the control layer. The control layer is used for completing real-time dispatching of the coal mine transportation chain according to the dispatching instruction. The data storage layer is used for storing coal mine transportation chain data and carrying out abnormal alarm by analyzing the coal mine transportation chain data. The coal mine transportation chain data comprises historical coal mine transportation chain data, first real-time coal mine transportation chain data and second real-time coal mine transportation chain data, wherein the second real-time coal mine transportation chain data is coal mine transportation chain data corresponding to the later time sequence of the time sequence corresponding to the first real-time coal mine transportation chain data.
When the monitoring and dispatching layer needs to acquire the data information of the second real-time coal mine transportation chain data, the second real-time coal mine transportation chain data is converted into the first real-time coal mine transportation chain data.
Optionally, with continued reference to FIG. 1, the monitoring scheduling layer includes a real-time monitoring module, a machine learning model building module, and a scheduling module.
The real-time monitoring module is used for acquiring historical coal mine transportation chain data and first real-time coal mine transportation chain data, sending the historical coal mine transportation chain data to the machine learning model building module, and sending the first real-time coal mine transportation chain data to the scheduling module. The machine learning model construction module is used for constructing a scheduling model according to historical coal mine transportation chain data, and the scheduling model comprises a prediction model and a machine learning classification model. The scheduling module is used for generating scheduling instructions according to the first real-time coal mine transportation chain data and the scheduling model.
Alternatively, referring to the schematic modular structure of the machine learning model building block shown in fig. 2, the machine learning model building block includes a first data processing sub-block, a prediction model building sub-block, and a machine learning classification model building sub-block.
The first data processing sub-module is used for preprocessing the historical coal mine transportation chain data to generate a historical sample data set, wherein the historical sample data set is data obtained by normalizing the historical coal mine transportation chain data. The prediction model construction submodule is used for constructing a prediction model according to the historical sample data set and a first preset machine learning network architecture, wherein the first preset machine learning network architecture is an ARMA model architecture or a BP neural network model architecture. The machine learning classification model construction submodule is used for constructing a machine learning classification model according to a historical sample data set, a scheduling instruction corresponding to the historical sample data set and a second preset machine learning network architecture, wherein the second preset machine learning network architecture is constructed based on a support vector machine algorithm.
Optionally, the historical coal mine transportation chain data comprises coal mine transportation chain basic data and coal mine transportation chain real-time data. The prediction model construction submodule takes the coal mine transportation chain basic data as model boundary data, takes the coal mine transportation chain real-time data as model training data, and takes the coal mine transportation chain basic data as the total number of mine cars and the total number of transportation tracks, wherein the coal mine transportation chain real-time data as the number of mine cars and the number of transportation tracks in an operation state at historical moment.
Optionally, referring to the schematic block diagram of the scheduling module shown in fig. 3, the scheduling module includes a second data processing sub-module and a scheduling command generating sub-module.
The second data processing sub-module is used for preprocessing the first real-time coal mine transportation chain data to generate a real-time sample data set, wherein the real-time sample data set is data obtained by normalizing the first real-time coal mine transportation chain data. The scheduling command generation submodule is used for generating a first real-time coal mine transportation chain data predicted value according to the real-time sample data set, generating a scheduling command according to the first real-time coal mine transportation chain data predicted value and sending the scheduling command to the control layer, wherein the first real-time coal mine transportation chain data predicted value is a coal mine transportation chain data predicted value corresponding to the later time sequence of the time sequence corresponding to the first real-time coal mine transportation chain data.
Optionally, with continued reference to fig. 1, the data storage layer includes an anomaly alarm module.
The abnormality alarm module is used for judging whether the coal mine transportation chain is abnormal or not according to the second real-time coal mine transportation chain data and the first real-time coal mine transportation chain data predicted value and alarming.
Optionally, referring to the schematic modularized structure of the abnormality alarm module shown in fig. 4, the abnormality alarm module includes a third data processing sub-module, an alarm signal generating sub-module, and an alarm signal displaying sub-module.
The third data processing sub-module is used for preprocessing the first real-time coal mine transportation chain data predicted value to generate a second real-time coal mine transportation chain data predicted value, wherein the second real-time coal mine transportation chain data predicted value is data obtained by performing de-normalization based on the first real-time coal mine transportation chain data predicted value. The alarm signal generation sub-module is used for generating abnormal alarm information according to the second real-time coal mine transportation chain data, the second real-time coal mine transportation chain data predicted value and the preset threshold value, and the alarm signal display sub-module is used for displaying or broadcasting the abnormal alarm information.
Optionally, with continued reference to fig. 1, the data storage layer includes a data storage module.
The data storage module is used for storing historical coal mine transportation chain data, first real-time coal mine transportation chain data, second real-time coal mine transportation chain data, first real-time coal mine transportation chain data predicted values and second real-time coal mine transportation chain data predicted values.
It should be noted that, according to actual needs, a historical sample data set and a real-time sample data set may be stored.
Optionally, the data storage module is further configured to protect data security of the historical coal mine transportation chain data, the first real-time coal mine transportation chain data, the second real-time coal mine transportation chain data, the first real-time coal mine transportation chain data predictive value and the second real-time coal mine transportation chain data predictive value, and trace data sources of the historical coal mine transportation chain data, the first real-time coal mine transportation chain data, the second real-time coal mine transportation chain data, the first real-time coal mine transportation chain data predictive value and the second real-time coal mine transportation chain data predictive value.
Optionally, referring to the schematic block diagram of the data storage module shown in fig. 5, the data storage module further includes a data security and trusted trace back sub-module.
The data security and credibility tracing submodule is used for constructing a blockchain data storage architecture and generating unique tracing codes corresponding to historical coal mine transportation chain data, first real-time coal mine transportation chain data, second real-time coal mine transportation chain data, first real-time coal mine transportation chain data predicted values and second real-time coal mine transportation chain data predicted values.
According to the embodiment of the invention, the intelligent monitoring and scheduling system of the coal mine transportation chain is introduced, so that the technical bottlenecks and management short plates in the traditional coal mine transportation system can be effectively overcome. Firstly, the machine learning model can train a prediction model according to historical operation data, so that accurate prediction of a coal mine transportation chain is realized. The state of the coal mine transportation chain is monitored in real time through the real-time monitoring module in the system, the machine learning module and the scheduling module constructed in the system can automatically analyze the state of the current transportation chain, and the scheduling strategy is dynamically adjusted according to the prediction result, so that common errors and delays in manual scheduling are avoided, and the scheduling efficiency and accuracy are improved.
And secondly, the introduction of advanced algorithms such as a machine learning classification model, an ARMA model, a BP neural network and the like can optimize and schedule a transportation chain according to the complexity and the variability of transportation tasks. The algorithm can automatically generate scheduling instructions according to different transportation requirements and real-time conditions, and continuously optimize the prediction result through continuous learning and training, so that the self-adaptability and the flexibility of the system are improved.
Finally, through real-time data monitoring and big data analysis, the system can monitor the coal mine transportation chain in an omnibearing way and perform fault early warning. Once the system detects abnormal conditions, such as overtime running of a mine car or congestion of a transportation route, the system can timely send out an alarm and automatically adjust a scheduling scheme, so that the time of manual intervention and emergency reaction is reduced, and the safety and stability of coal mine transportation are improved.
In summary, the intelligent monitoring and scheduling system for the coal mine transportation chain provided by the embodiment of the invention is based on machine learning, and the coal mine transportation efficiency and the resource utilization rate are obviously improved through the organic combination of real-time monitoring, data processing, prediction model and automatic scheduling. The system can collect various operation data of a coal mine transportation chain in real time, such as the number of mine cars, transportation time, time consumption of the mine cars and the like, and analyze and predict historical data through a machine learning model (comprising a support vector machine, an ARMA model and a BP neural network), so that an accurate scheduling instruction is generated. Compared with traditional manual scheduling, the system has high automation degree, reduces human intervention and operation errors, and improves the accuracy and response speed of scheduling decisions. Through continuous deep learning and model optimization, the system can dynamically adjust the scheduling strategy according to the change in actual operation, optimize the resource allocation and avoid transportation bottleneck and no-load waste. The fault early warning mechanism can also give an alarm in time when equipment abnormality or transportation problem occurs, and adjust transportation tasks to ensure efficient, safe and stable operation of coal mine transportation. Overall, the scheme greatly improves the intelligent level of the coal mine transportation chain, reduces the operation cost, enhances the flexibility, adaptability and long-term self-optimizing capability of the system, and provides powerful technical support for the automation and fine management of coal mine transportation.
The embodiment of the invention also provides an intelligent monitoring and dispatching system for the coal mine transportation chain, which is shown in a modularized structure schematic diagram of the intelligent monitoring and dispatching system for the coal mine transportation chain in FIG. 6, and comprises a monitoring and dispatching layer, a control layer and a data storage layer. The monitoring and scheduling layer interacts with the control layer through a CANopen protocol, and interacts with the data storage layer through a database.
The control layer comprises a lower computer, wherein the monitoring and scheduling layer interacts with the lower computer through a CANopen protocol, and the lower computer is a bottom hardware device or a control system for directly controlling the coal mine transportation chain field device to execute actions and collect operation data. The monitoring and scheduling layer comprises a real-time monitoring module, a machine learning model building module and a scheduling module. The machine learning model module comprises a data processing module, a prediction model construction module and a machine learning classification model module, wherein the real-time monitoring module is used for acquiring real-time running state data of a monitored coal mine transportation chain and transmitting the real-time running state data of the coal mine transportation chain to the machine learning model module connected with the real-time monitoring module through a CANopen protocol, the data processing module carries out data cleaning, feature preprocessing, data integration, data change detection and data dimension reduction on the real-time running state data of the coal mine transportation chain transmitted by the real-time monitoring module to form a sample data set, the sample data set and data feature information are transmitted to the prediction model construction module and the machine learning classification model construction module, the prediction model construction module carries out machine learning training according to the received sample data set to construct a prediction model (for example, an ARMA model or a BP neural network model), the machine learning classification model carries out machine learning training according to the received sample data set and a scheduling command label, the scheduling model is formed by the prediction model and the machine learning classification model after training is transmitted to the scheduling module, the scheduling module carries out real-time data transmission of the coal mine transportation chain to the scheduling model through input to the scheduling model, the scheduling module outputs corresponding scheduling command to the scheduling command of the scheduling chain according to the scheduling command of the scheduling control layer, and the real-time transmission module completes the scheduling command of the scheduling command to the scheduling module.
Illustratively, the machine learning classification model is constructed in the following manner:
The machine learning classification model is trained through a deep learning method to achieve matching of coal mine transportation chain data and scheduling instructions. First, the system collects real-time operating state data of the coal mine transportation chain from the real-time monitoring module and performs preprocessing. In the data processing module, cleaning, feature preprocessing, data integration, data change detection and dimension reduction processing are carried out on the data to generate a sample data set. And transmitting the processed sample data set to a machine learning classification model.
The first step of the training process is normalization processing, and the process is used for eliminating scale differences of different features, so that the feature values are in a similar range, and the excessive influence of certain features on the training process due to larger values is avoided. The normalized data will be used as input data for the machine-learned classification model. The normalization method uses one of min-max scaling (min-maxscaling) or normalization (z-score).
The machine learning classification model is trained using a support vector machine (Support Vector Machine, SVM) algorithm. The SVM algorithm aims to maximize the separation between different classes by constructing a hyperplane to achieve classification of data. Specifically, the SVM algorithm may find an optimal separation hyperplane based on the training sample set, so that the sample data is correctly classified, i.e., different coal mine transportation chain data are classified into corresponding scheduling instruction labels.
In the training process, the model can judge whether the classification model meets the expected conditions according to the set evaluation indexes such as accuracy, recall rate and the like. If the performance of the model meets the requirement, the training process is completed, the model is used for actual dispatching tasks, and if the performance of the model does not meet the condition, iterative training is carried out again by adjusting the super parameters of the model, such as kernel functions, C values, gamma values and the like, so as to optimize the performance of the model. Through the process, a machine learning classification model is trained, so that the machine learning classification model is used for scheduling instruction generation of a coal mine transportation chain.
The machine learning classification model selects a Support Vector Machine (SVM) algorithm, and fully considers the characteristics and practical application requirements of coal mine transportation chain data. The data generated by the coal mine transportation chain has the characteristics of high dimensionality, nonlinearity, complexity and variability. When the SVM algorithm processes high-dimensional data, the nonlinear problem in the low-dimensional space can be mapped to the high-dimensional space through kernel function skills, so that a linearly separable hyperplane is searched in the high-dimensional space, and the problem of data classification is effectively solved. Compared with other classification models, the SVM algorithm selected by the machine learning classification model has the advantages that the structural risk minimization principle enables the model to still obtain better generalization capability when processing small sample data, and in a coal mine transportation scene, data sample distribution is unbalanced possibly due to the influence of factors such as equipment faults, weather changes and the like, and the characteristic of the SVM algorithm can reduce the risk of overfitting. In addition, the SVM algorithm can flexibly adapt to different data distribution by adjusting the kernel function and parameters thereof, and has stronger adaptability to complex and changeable data in a coal mine transportation chain. Therefore, the SVM algorithm is selected as a training algorithm of the machine learning classification model, so that the operation data of the coal mine transportation chain can be better processed, a precise scheduling instruction is generated, and the efficiency and the safety of coal mine transportation are improved.
Illustratively, the ARMA model is constructed in the following manner:
The ARMA model (autoregressive moving average model) is used for modeling and prediction of time series data. In the system, the ARMA model is mainly used for predicting the running state of a coal mine transportation chain, especially the key indexes such as the transportation quantity or the equipment state change in a future period. The construction process comprises normalization processing, smooth decomposition, parameter calculation and error adjustment of the data.
First, the data processing module performs normalization processing according to the collected sample data set. This is to ensure that the data is in the same dimension in order to more accurately build the time series model. The normalized data is subjected to smooth decomposition to obtain three parts: deterministic data whose order does not change over time,White noise with 1 unit root,White noise with-1 root of unit. Specifically, the ARMA model attempts to decompose the time series into a trend component (deterministic data that does not change over time) and a random disturbance component (white noise).
The core formula of the ARMA model is as follows:
,
wherein,Representing the observed value at the current moment in time,Is the coefficient of the auto-regressive,Is thatWhite noise with a root of 1 per unit,Is the coefficient of the moving average and,Is thatWhite noise with-1 root of unit. By this formula, the ARMA model describes how the observations at each moment in the time series are commonly determined by past observations (autoregressive part) and past error terms (moving average part).
In the construction process, it is first necessary to determine whether the predicted value of the ARMA model coincides with the actual value. If the predicted result meets the expectations, the model is built. If the model prediction value does not coincide with the actual value, the system calculates the error of the model prediction value and the actual value, adjusts the autoregressive coefficient and the moving average coefficient in the ARMA model through iterative optimization, and optimizes the model parameters by adopting a machine learning algorithm (such as particle swarm optimization or genetic algorithm) so as to obtain the optimal prediction result.
Illustratively, the machine learning classification model is applied in combination with the ARMA model as follows:
In a dispatching module of the system, an ARMA model and a machine learning classification model provide basis for dispatching instructions of a coal mine transportation chain. Specifically, the ARMA model is used for carrying out time series prediction on the future running state of the coal mine transportation chain, and provides necessary future data prediction for a scheduling module. And the machine learning classification model is used for adjusting the scheduling strategy in real time according to the running data monitored in real time and combining the prediction result of the ARMA model to generate the scheduling instruction. The combination of the two makes the dispatching system not only have the capability of predicting the future, but also respond to the changed running state in real time, thereby realizing efficient intelligent dispatching.
It is assumed that during operation of the coal mine conveyor chain, fluctuations in the load of the conveyor vehicle (e.g., the amount of conveyor coal) are monitored. The real-time monitoring module acquires the data and then transmits the data to the machine learning classification model. After data preprocessing and normalization processing, the support vector machine model is trained to finally form a machine learning classification model, and the model can give scheduling instructions in real time according to the current state (such as vehicle load, vehicle speed and the like) of the coal mine transportation chain. Meanwhile, the ARMA model can predict traffic fluctuation in a future period of time through analysis of historical load data, and provides trend prediction of future load for a dispatching system. By combining the two models, the scheduling module can accurately judge whether the transportation vehicles need to be increased or decreased or adjust the transportation route, so that more efficient transportation scheduling is realized.
Through the combination, the operation of the coal mine transportation chain can be more intelligent and refined, the resource waste is reduced, and the operation efficiency is improved.
The BP neural network model can also be used as a prediction model in the intelligent monitoring and scheduling system of the coal mine transportation chain.
Illustratively, the BP neural network model is constructed in the following specific manner:
The BP neural network model is used as one of prediction models and is responsible for scheduling prediction according to the input of real-time monitoring data and historical data. The construction process comprises normalization processing, network training, error optimization and iterative adjustment. The core of the BP neural network model is that the BP neural network model is calculated step by step from an input layer to an output layer through a hidden layer in a multi-layer perceptron mode, and parameters are optimized so as to accurately predict the running state of a transportation chain, and finally a scheduling instruction is generated. See a schematic flow chart of a BP neural network model building method shown in fig. 7. The detailed process of BP neural network model construction is as follows:
Step S702 data preprocessing and normalization
Specifically, firstly, the data processing module processes the collected real-time running state data of the coal mine transportation chain to obtain a preprocessed sample data set. In order to ensure the training effect of the neural network model, all input data need to be subjected to normalization processing. The purpose of normalization is to compress the range of values of different feature data to a uniform range, typically between 0 and 1, so as to avoid the adverse effect of excessive or insufficient values of certain features on network training.
The normalized data set is used as an input data set of the BP neural network, and the normalized data set contains various characteristic information for training a model. The parameters of the BP neural network model are used as parameters of the BP neural network model, so that the network is helped to better establish the relation between different characteristic data and the prediction result.
Step S704 construction and training of BP neural network model
Specifically, the BP neural network consists of three layers, namely an input layer, a hidden layer and an output layer. In the construction process, it is first necessary to determine the structure of the neural network, including the number of neurons, the dimensions of inputs and outputs.
Assuming that the input layer has m neurons and the output layer has n neurons, the input data set contains m eigenvalues and the output data set contains n predicted values. These input and output data will be calculated by the weight parameters and the activation function to obtain the predicted output of the BP neural network.
The core formula of the BP neural network is as follows:
,
wherein,Is the output of the neural network and,Is a connection inputAnd the weights of the neurons of the population,Is a bias term. The learning goal of the BP neural network is to adjust weight parameters through a back propagation algorithm and minimize the predicted valueAnd the actual value.
Step S706 error calculation and optimization
Specifically, in the network training process, the error between the predicted value and the actual value needs to be calculated, and the weight parameter of the neural network is adjusted through a back propagation algorithm. The key formulas for error calculation are as follows:
,
wherein,As a result of the overall error, the total error,Is a predicted value of the neural network,Is the actual value. To minimize the error, the BP neural network steps the weights and biases through a gradient descent method until the error falls within an acceptable range.
The function of the error function is to measure the difference between the predicted value and the actual value and to guide the training process of the neural network. Specifically, the network will calculate the gradient based on the error back propagation and update the weights and offsets of each layer by iterative optimization algorithms (e.g., gradient descent or Adam optimizer) to reduce the error.
Step S708, network training and parameter updating
Specifically, the training of the BP neural network is a iterative process. In each iteration, the network first calculates the error between the predicted value and the actual value and adjusts the network parameters by error back propagation. And if the error meets the preset precision requirement, ending the training process, and completing the model construction. If the error is too large, iteration is continued, and network performance is optimized by adjusting super-parameters such as the number of neurons, the learning rate, the activation function and the like until a preset error range is reached.
Step S710 application of BP neural network model
Specifically, the trained BP neural network model will be used for real-time scheduling of coal mine transportation chains. In the real-time scheduling process, the running state data collected by the real-time monitoring module is used as input to be transmitted to a trained BP neural network model, the model calculates a prediction result according to the input data, and a scheduling instruction is generated according to the prediction result. By the mode, the BP neural network can flexibly cope with different running state changes in the coal mine transportation chain, a scheduling scheme is timely optimized, and scheduling efficiency is improved.
It is assumed that the load of the transport vehicle is a key indicator to be predicted in the coal mine transport chain. The system firstly acquires current load data of the vehicle through a real-time monitoring module, wherein the current load data comprises information such as traffic volume, vehicle speed, transportation route and the like. The data processing module cleans and normalizes the data to obtain a sample data set.
These data will be passed as input to the BP neural network model for training. During training, the network continuously adjusts the weight parameters until the error between the output value (predicted load) and the actual load is minimized. Once the network training is completed, the model can be used to predict the load over a period of time in the future and generate scheduling instructions based on the prediction. For example, if the predicted load is too high, the dispatch system may instruct the number of vehicles to increase or the route to be transported to be adjusted, and if the predicted load is too low, the system may reduce the vehicles or optimize the route.
In this way, the BP neural network can not only make accurate scheduling prediction according to real-time data, but also automatically adjust parameters along with the change of the data so as to cope with various changes possibly occurring in a transportation chain.
It should be noted that, in the embodiment of the present invention, the ARMA model and the BP neural network may be used to predict at the same time. Specifically, in an actual coal mine transportation scenario, different specific scheduling methods are adopted for different transportation chain state data. For example, when it is monitored that the average operating time of the mine cars on a particular transportation route exceeds a preset threshold and the traffic of the route is at a high level for a period of time, it is indicative that a congestion condition may occur on the route. At this point, the scheduling system will employ a scheduling method that first predicts traffic volume for the route over a period of time in the future using the ARMA model, and if the predicted traffic volume will remain high, the scheduling module instructs the deployment of mine cars from other relatively free routes to the congested route to increase the transport capacity. Meanwhile, according to analysis of the real-time operation data by the machine learning classification model, the departure time interval of the mine car on the line is adjusted, the interval is properly shortened, and the transportation efficiency is improved. In addition, if the BP neural network model predicts that the coal yield of a certain area will be greatly increased in the next few hours, according to the analysis of the machine learning classification model on real-time operation data, the dispatching system plans the transportation route in advance, preferentially schedules the mine car to go to the area for coal transportation, and reasonably distributes the quantity of the mine cars according to the predicted yield, so that the coal can be transported out of the mine in time and efficiently.
In the intelligent monitoring and scheduling system for the coal mine transportation chain, when parameters of a machine learning classification model need to be updated, a scheduling module carries out iterative training on the model so as to optimize a scheduling result. The aim of each iteration is to adjust the parameters of the model, so that the scheduling instruction is more accurate, and the real-time change of the coal mine transportation chain can be better adapted. The number of iterations is determined based on the update of parameters in the scheduling model, specificallyThe parameters are updated by two partsAndAnd determining, respectively representing the first updating parameter and the second updating parameter corresponding to the scheduling model. The formula is as follows:
,
wherein,Is the total number of iterations that are performed,AndRepresenting the number of iterations required for parameter updating of the predictive model and the machine learning classification model, respectively. Through the formula, the scheduling module can adjust parameters during each training, so that the model is optimized on multiple layers, and the accuracy and adaptability of scheduling decisions are effectively improved.
In the present system, the training step size (i.e., the amplitude adjusted in each iteration) is set to 60 times. This step size means that 60 steps are performed for each update of the model parameters, thereby adjusting the network parameters and gradually improving the prediction accuracy of the scheduling model.
The iteration times are greater than or equal to 60 times, so as to ensure that the model is trained sufficiently and avoid that the data cannot be fitted sufficiently under the condition of too few iteration times. After each iteration, the model is tested on a new training set to check if the error is effectively reduced. If the error is not significantly reduced, the scheduling module continues to iterate until the training reaches the preset accuracy requirement.
Through repeated iterative training, the scheduling model is gradually optimized, and the running state of the coal mine transportation chain can be predicted more accurately. Once trained, the model may be used to generate scheduling instructions. For example, given the number of mine cars and the operating time, the model can predict future transportation demands and generate scheduling instructions based on the prediction. The instructions may direct the arrangement of mine cars in the coal mine conveyor chain, the allocation of the conveyor tasks, etc.
The scheduling module can continuously update the scheduling strategy according to the real-time data and the trained model, so that the efficiency and safety of the coal mine transportation chain are ensured. For example, if the model predicts that the traffic in a certain area will increase, the dispatch system may instruct the number of mine cars to increase or the route of transportation to be adjusted to optimize resource utilization.
In a certain training, the number, the running time and the time consumption of the mine cars recorded by the real-time monitoring module are assumed to be input data, and the input data are transmitted to the neural network model after normalization processing. In the training process, the BP neural network model continuously updates the weight through a gradient descent method, and network parameters are optimized in iteration. After these iterations, the model has been able to accurately predict the mine car load and the transportation mission over a period of time in the future. Based on the prediction, the dispatching module can automatically adjust the quantity of the mine cars and the distribution of the transportation tasks according to the result output by the model, so that the overall transportation efficiency of the coal mine is improved.
By the mode, the model can be predicted according to historical data, can be flexibly adjusted when the data change is monitored in real time, and ensures that the scheduling system has high adaptability and accuracy.
In order to further improve the performance of the intelligent monitoring and scheduling system of the coal mine transportation chain, a data safety and credibility traceability module combined with the blockchain is constructed, so that the data safety can be effectively ensured, the transparency of the system is improved, and the cross-system cooperation is optimized. The core goal of introducing blockchain technology is to build a de-centralized data storage and management platform so that every data block in the system is tamper-resistant and traceable. By designing the blockchain data storage architecture, key data in the coal mine transportation chain process, such as raw data including mine car quantity, running time, mine car time consumption and the like, and sample data, characteristic data, model parameters and the like generated by a machine learning model can be encrypted and stored in a distributed mode in the blockchain nodes. Each data block contains information such as a time stamp, a data hash value, a hash value of a previous data block and the like, so that the sequency and the non-falsifiability of data in the storage process are ensured, and the data uplink flow can be standardized through an intelligent contract. To achieve this goal, the creation and verification of data blocks relies not only on the consensus mechanism of the blockchain node, but also in combination with monitoring device identification verification and data integrity verification, ensures that only verified data can be successfully written into the blockchain network.
In order to realize a trusted traceability mechanism, each data set is identified by generating a unique traceability code, and the traceability code penetrates through the whole data life cycle and is from acquisition, transmission, processing to various links such as model application and the like. When the transport chain data at a certain moment need to be traced, the system can search the related data blocks and the chains thereof on the block chain according to the tracing code, and clearly display the detailed information of the data generating path, the change record and the processing operation. Therefore, if abnormality occurs, such as data fluctuation exceeds a set reasonable range, the system can automatically trigger a backtracking mechanism to rapidly locate an abnormality source and judge whether the monitoring equipment is in fault, data transmission interference or abnormality of the model. In the process, the blockchain technology helps to quickly find out problems and repair the problems by providing a transparent and non-tamperable record, so that the stable operation of the whole transportation chain is ensured.
Collaborative optimization of the machine learning model and the blockchain data storage system may be enhanced by a data quality feedback mechanism. The trusted data stored by the blockchain may be fed back to the training phase of the machine learning model as a "quality inspector". When the degree of deviation between the training data and the historical quality data stored in the blockchain is found to be too large, the system can pause training and prompt that the data acquisition source or the preprocessing flow needs to be reviewed again. The mechanism effectively avoids the influence of dirty data on the performance of the model, ensures the quality of training data of the model, and improves the accuracy and reliability of prediction. In order to further optimize the evolution process of the model, when the model parameters (such as super parameters of the machine learning model or weights of the neural network) are updated, key indexes (such as accuracy, recall, prediction error and the like) before and after the update and relevant data version information are recorded in the blockchain. The information provides reliable basis for subsequent model updating, and can ensure verifiability and scientificity of each model updating by combining a common-knowledge verification mechanism of a blockchain.
In order to realize cross-system data collaboration, the intelligent monitoring and scheduling system of the coal mine transportation chain can fuse the trusted data on the blockchain with the operation management system of an enterprise through docking with management systems such as enterprise ERP, MES and the like. Based on the trusted data shared by the block chains, coal mine enterprises can more accurately predict the traffic, optimize the production plan, and carry out material purchasing and production scheduling according to the transportation efficiency, so that the scientificity and the accuracy of the whole operation decision are improved. Meanwhile, an industry data sharing alliance is established by combining a plurality of coal mine enterprises, and anonymization processing and sharing are carried out on the common data through a blockchain network, so that iteration and collaborative innovation of technology in the industry can be promoted, and the improvement of the intelligent level of the whole coal mine industry is accelerated.
In the combining process of block chain and machine learning, the algorithm design of data tracing and abnormal early warning is particularly important. The implementation of the anomaly early warning mechanism can be described using the following formula:
,
wherein,Represents an abnormal value at the time t,Is the data value at time t,Is a predicted value based on the history data,Is a preset anomaly threshold. When (when)When the threshold value is exceeded, the system triggers abnormal early warning and starts a tracing flow.
The purpose of the formula is to judge whether the current data deviates from the normal range or not by comparing the predicted values of the real-time data and the historical data, and if the deviation is too large, the judgment is abnormal. Through the mechanism, the system can automatically identify and alarm, reduce the load of manual monitoring and improve the overall operation efficiency. For example, if the variation of the mine car load amount at a certain moment exceeds the historical prediction range, the system can inquire the generation process of the data through the blockchain, analyze whether the fluctuation is caused by equipment failure, data transmission errors or model abnormality, and take measures for repairing in a targeted manner.
In a word, the application of the block chain-based data security and credibility traceability module in the intelligent monitoring and scheduling system of the coal mine transportation chain not only improves the security and transparency of data, but also optimizes the data quality and model training process by combining with a machine learning model, and realizes collaborative innovation and upgrading of the coal mine industry technology in multiple layers and multiple dimensions.
The intelligent scheduling system for the coal mine transportation chain based on the machine learning provided by the embodiment of the invention combines the real-time monitoring, data processing and machine learning algorithms based on the scheme of the intelligent monitoring and scheduling system for the coal mine transportation chain based on the machine learning, and has the following advantages:
improving coal mine transportation efficiency
The system provided by the embodiment of the invention continuously collects the running state data (such as the number of mine cars, running time, time consumption of the mine cars and the like) of the coal mine transportation chain through the real-time monitoring module, predicts and dispatches by utilizing the machine learning construction model, and can adjust transportation tasks and resource allocation in real time. Particularly, when the transportation demand suddenly increases or fails, the system can quickly respond to optimize the dispatching instruction, thereby avoiding resource waste and transportation bottleneck and improving the transportation efficiency of the whole coal mine.
Reducing human intervention and optimizing scheduling decisions
Traditional coal mine transportation scheduling is usually adjusted by relying on manual experience, and in a complex environment, manual decision making is easy to error, and rapid change of transportation requirements are difficult to deal with. The system provided by the embodiment of the invention can carry out autonomous decision according to real-time data by introducing the machine learning classification model, reduce the frequency of human intervention, automatically generate the scheduling instruction and carry out real-time adjustment according to the result of model prediction. The automatic scheduling capability not only improves the intelligent level of the system, but also ensures the accuracy and timeliness of scheduling decisions.
Enhancing system adaptivity and flexibility
When the machine learning model introduced by the system provided by the embodiment of the invention is used for parameter optimization by using a gradient descent method, continuous self-adjustment can be carried out according to different time periods, transportation states and external environment changes. The system has high adaptability, can quickly react when facing complex and dynamic changes in the coal mine transportation chain, and adjusts a prediction model and a scheduling strategy according to the changes, so that the system can operate efficiently.
Improving prediction accuracy and reducing error
The system provided by the embodiment of the invention can more accurately predict the running state of the transportation chain by using the support vector machine, the BP neural network, the ARMA model and other advanced machine learning algorithms. The prediction not only covers the factors of mine car quantity, transportation time, mine car time consumption and the like, but also can provide accurate basis for subsequent scheduling decisions. The high-precision prediction capability can effectively reduce errors of human estimation and avoid excessive or insufficient situations possibly occurring in scheduling.
Optimizing resource allocation and cost control
The system provided by the embodiment of the invention can conduct data processing, prediction and scheduling in real time by introducing the machine learning model, so that the system can more reasonably configure transportation resources. For example, based on the predictive model, the system can be adjusted in time when the mine car is insufficient or overloaded, avoiding wasting of resources or overload in the transportation process. By dynamically adjusting the number of mine cars, the transportation route and the transportation sequence, the system can furthest improve the utilization efficiency of transportation resources and realize cost control.
Real-time monitoring and fault early warning capability
The system provided by the embodiment of the invention not only can monitor the running state of the coal mine transportation chain in real time, but also can perform fault early warning on equipment states, transportation processes and the like according to real-time data. Once an anomaly is found (e.g., excessive transit time, abnormal car operation, etc.), the system can immediately alert and present an improved scheduling scheme based on the machine learning model. This function significantly improves the safety and reliability of the system, avoiding significant losses due to equipment failure or transportation bottlenecks.
Gradual self-learning and model optimization
The self-learning capability of the machine learning model is introduced into the system provided by the embodiment of the invention, so that the scheduling decision of the system can be continuously optimized according to the accumulation and feedback of historical data in the long-term operation process of the system. Through continuous training of machine learning algorithms and iterative optimization of model parameters, the system can gradually improve prediction capability and scheduling accuracy in the face of more diversified operating scenarios. This learning mechanism makes the system more and more accurate in continuous use, and also more and more adaptive and efficient.
Efficient data processing and decision support
The data processing module in the system provided by the embodiment of the invention performs preprocessing operations such as cleaning, feature selection, dimension reduction and the like on the real-time monitoring data, so that the quality of the data is improved, and the training speed and decision efficiency of the machine learning model are also improved. Through the efficient processing of the data, the system can extract useful features from a large amount of real-time data, support the rapid generation of scheduling instructions, reduce decision delay and enhance the real-time response capability of the system.
Of course, it will be appreciated by those skilled in the art that implementing all or part of the above-described methods in the embodiments may be implemented by a computer level to instruct a control device, where the program may be stored in a computer readable storage medium, and the program may include the above-described methods in the embodiments when executed, where the storage medium may be a memory, a magnetic disk, an optical disk, or the like.
Finally, it is further noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises an element.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other.
Although the present invention is disclosed above, the present invention is not limited thereto. Various changes and modifications may be made by one skilled in the art without departing from the spirit and scope of the invention, and the scope of the invention should be assessed accordingly to that of the appended claims.