Disclosure of Invention
Based on the above, it is necessary to provide a prediction model correction method, device and equipment medium based on active migration learning, so as to effectively avoid the problem of inaccurate prediction results caused by abrupt change of data and change of surrounding factors.
To achieve the above object, a first aspect of the present application provides a prediction model correction method based on active migration learning, the method comprising:
Acquiring a historical observation data set to obtain a predictive digital twin model constructed based on the historical observation data set;
Collecting current observation data, inputting the current observation data into the predictive digital twin model for prediction, comparison and analysis to obtain a first prediction error of the predictive digital twin model;
and performing active migration learning on the predicted digital twin model according to the first prediction error and the current observation data to obtain a corrected predicted digital twin model.
Further, the step of collecting current observation data, and inputting the current observation data into the predictive digital twin model for prediction and comparison analysis to obtain a first prediction error of the predictive digital twin model, specifically includes:
Collecting current observation data in real time, wherein the current observation data comprises input data and real output data corresponding to the input data;
inputting input data in current observation data into the predictive digital twin model to obtain predictive data output by the predictive digital twin model;
and calculating the error between the predicted data and the real output data to obtain a first predicted error size of the predicted digital twin model.
Further, the performing active migration learning on the predicted digital twin model according to the first prediction error and the current observation data to obtain a corrected predicted digital twin model specifically includes:
Comparing the first prediction error with a preset first error threshold, and when the first prediction error is larger than the first error threshold, performing active migration learning on the predicted digital twin model according to current observation data to obtain a corrected predicted digital twin model.
Further, the performing active migration learning on the predicted digital twin model according to the current observation data to obtain a corrected predicted digital twin model specifically includes:
according to the active learning technology, carrying out t-th screening on the current observed data to obtain screened correction data, wherein the initial value of t is 1;
performing migration learning on the predicted digital twin model according to the correction data to obtain a predicted digital twin model to be detected;
acquiring a second prediction error of the digital twin model to be predicted;
when the second prediction error is not greater than a preset second error threshold, the predicted digital twin model to be detected is used as a corrected predicted digital twin model;
And when the second prediction error is greater than the second error threshold, acquiring next observed data, taking the next observed data as current observed data, enabling t=t+1, and continuously executing the step of t-th screening of the current observed data according to the active learning technology to obtain screened correction data until the second prediction error is not greater than the second error threshold.
Further, the t-th screening of the current observation data according to the active learning technology is performed to obtain screened correction data, which specifically includes:
Performing ith coding on the current observation data to obtain a coding vector, wherein the initial value of i is 1;
learning the potential space of the current observation data by using an countermeasure network to obtain potential space vectors;
according to the coding vector of the current observation data, the potential space vector, the coding vector of the historical observation data and the potential space vector, carrying out distance analysis to obtain the similarity between the current observation data and the historical observation data;
when the similarity between the current observation data and the historical observation data is not greater than a preset similarity threshold value, the current observation data is used as correction data;
When the similarity between the current observation data and the historical observation data is greater than a preset similarity threshold value, collecting next observation data, taking the next observation data as the current observation data, enabling i=i+1, and continuing to execute the ith coding of the current observation data to obtain a coding vector until the similarity between the current observation data and the historical observation data is not greater than the preset similarity threshold value.
Further, the similarity between the current observation data and the historical observation data is calculated by the following formula:
LD＝-E[log(D(q(zL|xL)))]-E[log(1-D(q(zU|xU)))]
Wherein q is an encoder; e is a mathematical expectation; xL is the code vector of the historical observation dataset; zL is the potential space vector of the historical observation dataset; d is an observation dataset; xU is the coding vector of the current observation data; zU is the potential spatial vector of the current observation.
Further, the method further comprises:
when the first prediction error is not larger than the first error threshold, acquiring a time span between the current moment and the moment when the predictive digital twin model is actively transferred and learned last time;
And when the time span is greater than a preset time threshold, acquiring a correction database, and correcting the predictive digital twin model according to the correction database to obtain a corrected predictive digital twin model.
In order to achieve the above object, a second aspect of the present application provides a prediction model correction device based on active migration learning, wherein the device includes: the system comprises a data acquisition module, a data analysis module and a model correction module;
The data acquisition module is used for acquiring historical observation data and obtaining a predictive digital twin model constructed based on the historical observation data set;
The data analysis module is used for collecting current observation data, inputting the current observation data into the predictive digital twin model for prediction and comparison analysis, and obtaining a first prediction error of the predictive digital twin model;
The model correction module is used for carrying out active migration learning on the predicted digital twin model according to the first prediction error and the current observation data to obtain a corrected predicted digital twin model.
To achieve the above object, a third aspect of the present application provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, causes the processor to perform the steps of the method according to the first aspect.
To achieve the above object, a fourth aspect of the present application provides a computer device comprising a memory and a processor, characterized in that the memory stores a computer program which, when executed by the processor, causes the processor to perform the steps of the method according to the first aspect.
The embodiment of the invention has the following beneficial effects:
The invention provides a prediction model correction method based on active transfer learning, which comprises the following steps: acquiring a historical observation data set, and constructing a predictive digital twin model based on the historical observation data set; collecting current observation data, inputting the current observation data into a predictive digital twin model for prediction, comparison and analysis, and obtaining a first prediction error of the predictive digital twin model; and performing active migration learning on the predicted digital twin model according to the first prediction error and the current observation data to obtain a corrected predicted digital twin model, and performing active migration learning on the predicted digital twin model through the observation data acquired in real time to obtain a corrected predicted digital twin model, thereby effectively preventing the problems of inaccurate prediction results caused by abrupt change of the data and inaccurate prediction results caused by changes of factors influencing the prediction results over time.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. 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 existing load prediction model has the problems of limited processing capacity for nonlinear relation, high complexity for calculation of large data volume, insufficient sensitivity to change of external factors, difficulty in processing abrupt change of data, long-time dependency relation and the like, the embodiment of the application provides a prediction model correction method based on active transfer learning, which can update training data and a prediction model in real time by combining with the active transfer learning, so as to quickly and accurately obtain a prediction result, and solve the problems of long training time and small training data volume of the model of the original deep learning method. Referring to fig. 1, fig. 1 is a flow chart of a prediction model correction method based on active migration learning according to an embodiment of the present application, which specifically includes:
Step 110, acquiring a historical observation data set, and obtaining a predicted digital twin model constructed based on the predicted digital twin model historical observation data set based on a digital twin technology.
In the embodiment of the invention, a digital twin framework is firstly constructed, a physical entity of the framework is historical observation data, a predictive digital twin model is constructed according to the historical observation data, and the model is a virtual entity. The predictive digital twin model may be a transducer model, a neural network architecture for natural language processing tasks, with efficiency and performance improved through self-attention mechanisms and parallel processing.
The predictive digital twin model may be a load predictive digital twin model in one embodiment of the invention for predicting the magnitude of a load based on input data. Since the load change is affected by weather factors such as time, temperature, humidity, precipitation, etc., the input data may be weather data.
Specifically, historical meteorological data and power load data can be collected, a historical observation data set is generated according to the historical meteorological data and the power load data, and a load prediction digital twin model is constructed according to the historical observation data set as a training set. The historical observation data set contains a plurality of historical observation data, each of which contains input data (meteorological data), and output data (electrical load data). Each historical observation data is a training sample, and the predicted digital twin model is trained through the training sample, so that a trained load training model can be obtained.
And 120, collecting current observation data, and inputting the current observation data into a predictive digital twin model for prediction, comparison and analysis to obtain a first prediction error of the predictive digital twin model.
After the predictive digital twin model is obtained via step 110, the predictive digital twin model may be put into service and monitored and corrected during use.
Specifically, the observation data are collected in real time, the collected observation data are utilized to detect the predictive digital twin model, and the error size existing in the current prediction process of the predictive digital twin model, namely the error size between the predicted value output by the predictive digital twin model and the collected true value is calculated.
And 130, performing active migration learning on the predicted digital twin model according to the first prediction error and the current observation data to obtain a corrected predicted digital twin model.
Active transfer learning is a method combining active learning and transfer learning, and improves the learning effect and efficiency of new tasks by selectively marking a small amount of data and utilizing the existing knowledge.
In the embodiment of the invention, whether the predictive digital twin model is required to be actively transferred and learned according to the observation data acquired in real time is determined by the current predictive error of the predictive digital twin model, so that parameters in the predictive digital twin model are updated, the corrected predictive digital twin model is obtained, and the prediction result of the corrected predictive digital twin model is more accurate.
The method for correcting the predictive digital twin model is based on digital twin frame operation. Referring to fig. 2, fig. 2 is a load prediction digital twin model in an embodiment of the present invention, where the load prediction digital twin model mainly includes three parts, namely a physical entity, a virtual entity, and interactions between the two entities, and the states and behaviors of the physical entity are reflected in real time by the virtual entity so as to monitor, simulate, and optimize.
In an embodiment of the present invention, the physical entity in the load prediction digital twin frame includes historical load data and historical meteorological data, and the digital twin model is a load prediction digital twin model constructed according to the historical load data and the historical meteorological data. And calculating an error between a predicted value and a true value of the load in the real-time load prediction process, and carrying out active migration learning on the load prediction digital twin model according to the size of the error and the load sample to correct the load prediction digital twin model, so that the real-time updating of the training sample set and the power load prediction digital twin model is realized, and the influence of time on the power load is adapted.
The power grid load can be detected, simulated, estimated and predicted by constructing a load prediction digital twin model, and if abnormal load conditions occur in the process, the model analyzes reasons and carries out model correction, and a correct decision is given to a load detection device. In addition, the method and the device perform active migration learning on the predicted digital twin model through the first prediction error and the observation data acquired in real time to obtain the corrected predicted digital twin model, and effectively prevent the problems that the predicted result is inaccurate due to mutation of the data and the predicted result is inaccurate due to change of factors influencing the predicted result along with time.
In step 120 of the embodiment of the present invention, current observation data is collected, and the current observation data is input into a predicted digital twin model for prediction and comparison analysis, so as to obtain a first prediction error size of the predicted digital twin model, which specifically includes:
step121, collecting current observation data in real time, wherein the current observation data comprises input data and real output data corresponding to the input data.
Step122, inputting the input data in the current observation data into the predictive digital twin model to obtain the predictive data output by the predictive digital twin model.
Step123, calculating an error between the predicted data and the real output data to obtain a first predicted error size of the predicted digital twin model.
In the embodiment of the invention, the data acquisition device is used for acquiring the observation data, wherein the observation data is a group of data, and the group of data comprises input data and real output data corresponding to the input data.
In an embodiment of the present invention, the predicted digital twin model is a load predicted digital twin model, and the current observation data includes weather data of a current time and electrical load data under the weather data. It will be appreciated that the data obtained by the measurement is real data.
Inputting the meteorological data into a load prediction digital twin model to perform load prediction, obtaining prediction data output by the load prediction digital twin model, and performing error calculation on the prediction data and electric load data under the meteorological data to obtain the current first prediction error of the load prediction digital twin model.
After the current first prediction error of the load prediction digital twin model is obtained, the first prediction error can be analyzed to determine how to modify the load prediction digital twin model according to current observation data. In step 130 of one embodiment of the present invention, active migration learning is performed on the predicted digital twin model according to the first prediction error and the current observation data to obtain a modified predicted digital twin model, which specifically includes:
step131, comparing the first prediction error with a preset first error threshold, and when the first prediction error is greater than the first error threshold, performing active migration learning on the predicted digital twin model according to the current observation data to obtain a corrected predicted digital twin model.
In the embodiment of the invention, the prediction error of the prediction digital twin model is monitored, and when the prediction error is overlarge, the prediction digital twin model is corrected.
Specifically, a first error threshold is preset, wherein the first error threshold is an error maximum value which allows prediction of the existence of the digital twin model. And comparing the first prediction error with a first error threshold value, and judging whether the current prediction error of the predictive digital twin model is overlarge.
When the first prediction error is larger than a first error threshold, confirming that the current prediction error of the predicted digital twin model is overlarge, and at the moment, performing active migration learning on the predicted digital twin model by utilizing the observation data acquired in real time, and correcting and updating parameters in the predicted digital twin model to obtain the corrected predicted digital twin model.
When the first prediction error is not greater than the first error threshold, the current prediction error of the prediction digital twin model is confirmed to be in an acceptable range, and correction is not needed, so that the digital twin model can be continuously put into use.
In an embodiment of the present invention, step131, performing active migration learning on the predicted digital twin model according to current observation data to obtain a corrected predicted digital twin model, specifically includes:
A. And (3) carrying out the t-th screening on the current observation data according to the active learning technology to obtain screened correction data, wherein the initial value of t is 1.
In an embodiment of the invention, a large number of training samples are needed to be used as data support for predicting the load by using the load prediction digital twin model, but a large amount of time is consumed if all training samples of the time domain simulation are used for model training one by one, so that the training sample set is required to be updated in real time in order to learn the load prediction digital twin model to the most load scenes as possible and to cope with the continuously changing power system environment, and the prediction result of the load prediction digital twin model is more accurate by combining the updated training sample set with the load prediction digital twin model.
The active learning technology can help the predictive digital twin model to acquire sample key features faster, so that the predictive digital twin model can be used for model training according to abundant sample content, accuracy of the predictive digital twin model is improved, time and calculated amount of model training can be reduced as much as possible, efficiency of model training is improved, and in one embodiment of the invention, variable division antagonism active learning technology can be adopted to select data which is not learned by the predictive digital twin model in real-time observation data.
Specifically, analyzing the current observed data to determine whether the current observed data is learned data, and if the current observed data is the learned data, not needing to learn the data again; if the current observation data is the data which is not learned, the current observation data is used as the correction data obtained by screening, and the transfer learning can be carried out on the predictive digital twin model according to the correction data.
In step a of one embodiment of the present invention, a t-th filtering is performed on current observation data according to an active learning technique to obtain filtered correction data, which specifically includes:
A1, carrying out ith coding on the current observation data to obtain a coding vector, wherein the initial value of i is 1.
And A2, learning the potential space of the current observation data by using the countermeasure network to obtain potential space vectors.
A3, carrying out distance analysis according to the coding vector of the current observation data, the coding vector of the potential space vector and the historical observation data and the potential space vector to obtain the similarity between the current observation data and the historical observation data.
A4, when the similarity between the current observation data and the historical observation data is not greater than a preset similarity threshold value, taking the current observation data as correction data.
And A5, when the similarity between the current observation data and the historical observation data is greater than a preset similarity threshold, acquiring next observation data, taking the next observation data as the current observation data, enabling i=i+1, and continuously executing the ith coding on the current observation data to obtain a coding vector until the similarity between the current observation data and the historical observation data is not greater than the preset similarity threshold.
In the embodiment of the present invention, the variable self-encoder (variational autoencoder, VAE) and the countermeasure network are adopted to learn the potential space of the observation data to determine whether the data features of the observation data have been learned, and reference may be made to fig. 3, and fig. 3 is a flow of variable countermeasure active migration learning in the embodiment of the present invention.
The variational antagonism active migration learning is to encode the observed data by using the VAE, wherein the observed data can be current observed data and historical observed data, and the potential space of the observed data is learned by using the antagonism network. If the difference between the current observation data and the historical observation data learned by the predictive digital twin model is large enough, calibrating the observation data is valuable without learning. It will be appreciated that the observation data is added to the historical observation data after each training of the predictive digital twin model with the observation data in order to improve the accuracy of the subsequent analysis results.
The encoder of the VAE learns the low dimensional space of the observed data distribution by means of a gaussian mixture model, and the decoder reconstructs the input data. The current observation data and the historical observation data in the historical observation data set are mixed and input into the VAE for training. Wherein, the reconstruction loss expression is:
Wherein q and p are encoder and decoder, respectively; e is a mathematical expectation; xL is the code vector of the historical observation dataset; zL is the potential space vector of the historical observation dataset; beta is the Lagrangian parameter of the optimization problem; dKL is the KL divergence of the observed data; xU is the coding vector of the current observation data; zU is the potential spatial vector of the current observation data; p (z) is a priori value of the unit gaussian selection.
According to the embodiment of the invention, the sample is selected through the countermeasure network, and the discriminator is adopted to distinguish whether the observed data is learned or not, so that the input sample is required to be encoded during the training of the VAE, the judgment of the input sample by the discriminator is deceived, and the unlearned observed data is selected as much as possible.
When the probability distributionsq(zL|xL) andq(zU|xU) for VAE training are similar, the spoof arbiter will determine all of the input data as learned samples, while the arbiter will also calculate the probability that the input data is not learned samples, i.e., a nash equilibrium state is reached.
Specifically, the objective function of the VAE is the binary cross entropy loss:
Wherein D is an observation data set; e is a mathematical expectation; xL is the code vector of the historical observation dataset; zL is the potential space vector of the historical observation dataset; xU is the coding vector of the current observation data; zU is the potential spatial vector of the current observation.
Combining the reconstruction loss and the binary cross entropy loss to obtain an objective function expression of variational resistance to the VAE in the active migration learning:
where λ1 and λ2 are constants.
The basis for finally judging whether the current observed data is the non-learned data is the output value of the discriminator, namely the similarity between the current observed data and the historical observed data, and the similarity is calculated by the following formula:
LD＝-E[log(D(q(zL|xL)))]-E[log(1-D(q(zU|xU)))]
When the LD value is very high, the similarity between the current observed data and the historical observed data is indicated, and the observed data does not need to be learned again; when the value of LD is low, the difference between the current observed data and the historical observed data is large, and the potential space of the current observed data needs to be learned.
In the embodiment of the invention, a similar threshold value is preset, and the similar threshold value is used for distinguishing whether LD is high or low. When LD is not greater than a preset similarity threshold, the difference between the current observation data and the historical observation data is larger, the current observation data can be used as correction data, and the predictive digital twin model is corrected; when LD is larger than a preset similarity threshold, the similarity between the current observation data and the historical observation data is indicated, the next observation data can be continuously collected in real time, the next observation data is taken as the current observation data, and the steps A1-A3 are repeated until the similarity between the current observation data and the historical observation data is not larger than the preset similarity threshold.
B. and performing migration learning on the predicted digital twin model according to the correction data to obtain the predicted digital twin model to be detected.
After the correction data is obtained, the correction data is used as a learning sample to be provided to the predicted digital twin model for transfer learning, and parameters in the predicted digital twin model are updated, so that the updated predicted result of the predicted digital twin model to be measured is more accurate.
C. and obtaining a second prediction error of the digital twin model to be predicted.
In the embodiment of the invention, test data is acquired, and the test data is used for testing the magnitude of the prediction error of the predicted digital twin model to be tested. The test data may be any one of historical observation data, real-time collected observation data or network acquired data.
D. And when the second prediction error is not greater than a preset second error threshold, taking the predicted digital twin model to be detected as a corrected predicted digital twin model.
E. When the second prediction error is greater than a second error threshold, collecting next observation data, taking the next observation data as current observation data, enabling t=t+1, and continuously executing the step of t-th screening on the current observation data according to the active learning technology to obtain screened correction data until the second prediction error is not greater than the second error threshold.
In the embodiment of the invention, whether the updated digital twin model to be predicted meets the prediction requirement is judged, namely whether the prediction result of the digital twin model to be predicted is high-precision is judged. Specifically, a second error threshold value is preset, the second error threshold value is compared with a second prediction error, and when the second prediction error is not larger than the preset second error threshold value, it is indicated that the to-be-detected prediction digital twin model of the to-be-detected prediction digital twin model meets the prediction requirement, and the to-be-detected prediction digital twin model can be used as a finally corrected prediction digital twin model. And when the second prediction error is larger than a preset second error threshold, the fact that the to-be-detected prediction digital twin model of the to-be-detected prediction digital twin model does not meet the prediction requirement is indicated, next observation data are collected, the next observation data are used as current observation data, and the steps A-C are repeatedly executed until the second prediction error is not larger than the second error threshold.
In one embodiment of the present invention, the training sample set is updated over time by screening real-time data through active transfer learning techniques, and the predictive digital twin model is modified. The specific method also comprises the following steps:
When the first prediction error is not greater than a first error threshold value, acquiring a time span between the current moment and the moment when the predictive digital twin model is actively transferred and learned last time; when the time span is larger than a preset time threshold, a correction database is obtained, and the predicted digital twin model is corrected according to the correction database, so that the corrected predicted digital twin model is obtained.
In the embodiment of the invention, the predictive digital twin model can actively perform the active migration learning after a period of distance except the active migration learning on the predictive digital twin model under the condition of larger predictive error. The correction database used for learning may be newly acquired observation data, or previously acquired observation data without learning, and is not limited thereto.
The invention can realize the selection of the unlearned observation data through the variation antagonism initiative transfer learning, further reduce the model training time, improve the model training efficiency, and show in experimental data that the predictive digital twin model correction process based on the variation antagonism initiative transfer learning, compared with a model training process without a variational countermeasure active transfer learning process, the training time is reduced by 153s, and the prediction accuracy is equivalent, so that remarkable progress is made in improving the model training efficiency.
In an embodiment of the present invention, a prediction model correction device based on active migration learning is provided, and referring to fig. 4, fig. 4 is a structural block diagram of a prediction model correction device based on active migration learning in an embodiment of the present invention, where the device includes: a data acquisition module 401, a data analysis module 402 and a model modification module 403.
The data acquisition module 401 is configured to acquire historical observation data, and obtain a predicted digital twin model constructed based on a historical observation data set based on a digital twin technology.
The data analysis module 402 is configured to collect current observation data, and input the current observation data into the predictive digital twin model for prediction and comparison analysis, so as to obtain a first prediction error of the predictive digital twin model.
The model modification module 403 is configured to perform active migration learning on the predicted digital twin model according to the first prediction error and the current observation data, so as to obtain a modified predicted digital twin model.
According to the prediction model correction device based on active migration learning, the first prediction error and the observation data acquired in real time are used for carrying out active migration learning on the prediction digital twin model, so that the corrected prediction digital twin model is obtained, and the problems that the prediction result is inaccurate due to mutation of the data and the prediction result is inaccurate due to change of factors influencing the prediction result along with time are effectively prevented.
FIG. 5 shows an internal block diagram of a computer device in one embodiment of the application. The computer device may specifically be a terminal or a system. As shown in fig. 5, the computer device includes a processor, a memory, and a network interface connected by a system bus. The memory includes a nonvolatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system, and may also store a computer program which, when executed by a processor, causes the processor to implement the steps of the method embodiments described above. The internal memory may also have stored therein a computer program which, when executed by a processor, causes the processor to perform the steps of the method embodiments described above. It will be appreciated by those skilled in the art that the structure shown in FIG. 5 is merely a block diagram of some of the structures associated with the present inventive arrangements and is not limiting of the computer device to which the present inventive arrangements may be applied, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.
In one embodiment, a computer device is provided that includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method embodiments described above.
In one embodiment, a computer-readable storage medium is provided, in which a computer program is stored which, when executed by a processor, causes the processor to perform the steps of the method embodiments described above.
Those skilled in the art will appreciate that the processes implementing all or part of the methods of the above embodiments may be implemented by a computer program for instructing relevant hardware, and the program may be stored in a non-volatile computer readable storage medium, and the program may include the processes of the embodiments of the methods as above when executed. Any reference to memory, storage, database, or other medium used in embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (SYNCHLINK) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The foregoing examples illustrate only a few embodiments of the application, which are described in detail and are not to be construed as limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of protection of the present application is to be determined by the appended claims.