An industrial park power consumption load prediction method and system

Abstract

The application relates to the technical field of cloud edge collaboration, in particular to an industrial area power load prediction method and system applied to a cloud server, which comprises the following steps: receiving a non-uniform continuous time hidden semi-Markov model sent by N edge computing terminals and performing decryption; according to a preliminary encrypted power load prediction vector of a power plant and a randomly generated password, a predicted future power load of the power plant and a weight and a threshold value of a power load prediction model of the power plant satisfying a preset accuracy requirement are obtained; according to the future power load and the weight and the threshold value satisfying the preset accuracy requirement, an industrial area power load prediction model is trained; the above operation is repeated until the industrial area power load prediction model reaches a preset accuracy requirement, and the power load of the industrial area is predicted through the industrial area power load model, so that the cloud edge collaboration technology is used to accurately predict the power load of the industrial area in a future period of time.

Description

Technical Field

[0001] This invention relates to the field of cloud-edge collaboration technology, and in particular to a method and system for predicting the electricity load of industrial areas. Background Technology

[0002] Excessive or insufficient electricity load in industrial areas can negatively impact the balance of power supply and demand and ensure electricity safety. Therefore, accurately predicting the electricity load of industrial areas over a future period, and adjusting it accordingly to avoid overloading or underloading, has become a pressing technical challenge. Currently, traditional methods, primarily manual or combined manual and machine-based forecasting, suffer from poor efficiency and accuracy, and typically only allow for overall regional forecasting, failing to predict individual power-consuming plants.

[0003] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main objective of this invention is to provide a method, apparatus, equipment, and storage medium for predicting the electricity load of an industrial area, aiming to solve the problem that it is difficult to accurately predict the electricity load of an industrial area in the future in the prior art, so as to adjust the electricity load of the industrial area in a timely manner based on the prediction results and avoid the electricity load being too high or too low.

[0005] To achieve the above objectives, the present invention provides a method for predicting the electricity load of an industrial area, wherein the method is applied to N edge computing terminals, where N is an integer greater than 1;

[0006] The method for predicting the electricity load of the industrial zone includes the following steps:

[0007] Obtain information on the holiday schedules of power-consuming factories, their historical electricity load, and weather forecast data;

[0008] Establish an electricity load forecasting model for power-consuming factories, and update the parameters of the electricity load forecasting model based on the holiday schedule information of the power-consuming factories, the historical electricity load of the power-consuming factories, and weather forecast data;

[0009] Repeat the above operation until the power load prediction model of the power-consuming factory reaches the preset power load prediction accuracy requirement of the power-consuming factory, and predict the future power load of the power-consuming factory using the power load prediction model of the power-consuming factory that meets the preset power load prediction accuracy requirement.

[0010] Optionally, updating the parameters of the electricity load prediction model based on the holiday schedule information of the power-consuming plant, the historical electricity load of the power-consuming plant, and weather forecast data includes:

[0011] The holiday schedule information of the power-consuming factory, the historical power load of the power-consuming factory, and the weather forecast data are input into the power load prediction model of the power-consuming factory to obtain the expected power load prediction result of the power-consuming factory. Based on the expected power load prediction result and the actual power load prediction result of the power-consuming factory, the estimated values ​​of weights and thresholds are obtained.

[0012] Perform a Taylor expansion on the estimated weights and retain the linear terms of the Taylor expansion;

[0013] Based on the linear terms of the Taylor expansion, the error function of the system is obtained, and nonlinear recursive least squares tracking is performed on the error function of the system to obtain the weights and thresholds after least squares tracking.

[0014] Optionally, after repeating the above operations until the power consumption load prediction model of the power-consuming plant reaches the preset power consumption load prediction accuracy requirement, and predicting the future power consumption load of the power-consuming plant using the power consumption load prediction model that meets the preset power consumption load prediction accuracy requirement, the method further includes:

[0015] Based on the non-uniform continuous-time hidden semi-Markov model, a password is randomly generated to encrypt the predicted future electricity load of the power-consuming factory and the weights and thresholds of the power load prediction model of the power-consuming factory that meet the preset accuracy requirements, so as to obtain a preliminary encrypted power load prediction vector of the power-consuming factory.

[0016] Based on the power load prediction vector of the pre-encrypted power plant, the randomly generated password, and the state transition matrix and initial state transition probability vector of the non-uniform continuous-time hidden semi-Markov model corresponding to the password, the corresponding non-uniform continuous-time hidden semi-Markov model is constructed in reverse.

[0017] The corresponding non-uniform continuous-time hidden semi-Markov model constructed in reverse is sent to the cloud server.

[0018] To achieve the above objectives, the present invention also provides a method for predicting the electricity load of an industrial area, wherein the method for predicting the electricity load of an industrial area is applied to a cloud server;

[0019] The method for predicting the electricity load of the industrial zone includes the following steps:

[0020] Receive the non-uniform continuous-time hidden semi-Markov model sent by the N edge computing terminals, and decrypt it to obtain a preliminary encrypted power load prediction vector for the power-consuming factory.

[0021] Based on the pre-encrypted power load prediction vector of the power-consuming factory and the randomly generated password, the predicted future power load of the power-consuming factory and the weights and thresholds of the power load prediction model of the power-consuming factory that meet the preset accuracy requirements are obtained.

[0022] Based on the predicted future electricity load of the power-consuming factory and the weights and thresholds of the power load prediction model of the power-consuming factory that meet the preset accuracy requirements, train the power load prediction model for the industrial area.

[0023] Repeat the above operations until the industrial zone electricity load prediction model reaches the preset accuracy requirement, and predict the industrial zone electricity load using the industrial zone electricity load model.

[0024] Optionally, training the industrial zone electricity load prediction model based on the predicted future electricity load of the power-consuming factory and the weights and thresholds of the electricity load prediction model of the power-consuming factory that meet the preset accuracy requirements includes:

[0025] The predicted future electricity load of the power-consuming factory and the weights and thresholds of the electricity load prediction model of the power-consuming factory that meet the preset accuracy requirements are extracted by a convolutional neural network to obtain the industrial area electricity load feature vector, and an industrial area electricity load feature vector library is established based on the industrial area electricity load feature vector.

[0026] The industrial area electricity load feature vectors in the industrial area electricity load feature vector library are classified by a classifier, and the classified industrial area electricity load feature vectors are clustered iteratively to update the classifier and cluster centers;

[0027] The industrial zone electricity load prediction model is updated based on the updated classifier and cluster centers.

[0028] Optionally, the step of classifying the industrial area electricity load feature vectors in the industrial area electricity load feature vector library using a classifier and updating the classifier includes:

[0029] The characteristic vector of the electricity load in the industrial area is classified to obtain the classification results and loss function;

[0030] The loss function is processed using cross-entropy loss, classification labels, and classification results to obtain the classification loss;

[0031] The classification loss is backpropagated through the backpropagation relation to update the classifier parameters.

[0032] Optionally, the step of performing clustering iteration on the classified industrial area electricity load feature vector and updating the cluster centers includes:

[0033] K industrial area electricity load feature vectors are randomly selected from the industrial area electricity load feature vector database as initial cluster centers, where K is an integer greater than 1;

[0034] Calculate the distance between the remaining industrial zone electricity load feature vectors in the industrial zone electricity load feature vector library and the initial cluster center;

[0035] Based on the distance between the electricity load feature vector of each industrial zone and the initial cluster center, the electricity load feature vector of the industrial zone is divided into K clusters;

[0036] Calculate the mean of all feature vectors of the K clusters, and use the mean as the updated cluster center.

[0037] Furthermore, to achieve the above objectives, the present invention also proposes an edge computing terminal, which includes: a memory, a processor, and an industrial area power load prediction program stored in the memory and running on the processor, wherein the industrial area power load prediction program is configured to implement the industrial area power load prediction method as described above.

[0038] Furthermore, to achieve the above objectives, the present invention also proposes a cloud server, the cloud server comprising: a memory, a processor, and an industrial area power load forecasting program stored on the memory and running on the processor, the industrial area power load forecasting program being configured to implement the industrial area power load forecasting method as described above.

[0039] In addition, to achieve the above objectives, the present invention also proposes an industrial area power load forecasting system, which includes the aforementioned N edge computing terminals and cloud server.

[0040] Furthermore, to achieve the above objectives, the present invention also proposes a storage medium storing an industrial area power load prediction program, which, when executed by a processor, implements the industrial area power load prediction method as described above.

[0041] This invention discloses a method for predicting the electricity load of an industrial zone, applied to N edge computing terminals, where N is an integer greater than 1. The method includes the following steps: acquiring the holiday schedule information of the power-consuming factories, their historical electricity load, and weather forecast data; establishing an electricity load prediction model for the power-consuming factories, and updating the parameters of the prediction model based on the holiday schedule information, historical electricity load, and weather forecast data; repeating the above operations until the electricity load prediction model reaches a preset accuracy requirement for predicting the electricity load of the power-consuming factories, and predicting the future electricity load of the power-consuming factories using the electricity load prediction model that meets the preset accuracy requirement. An industrial zone electricity load forecasting method is applied to a cloud server; it includes the following steps: receiving the non-uniform continuous-time hidden semi-Markov model sent by N edge computing terminals and decrypting it to obtain a preliminary encrypted electricity load forecast vector for the power-consuming factories; restoring the predicted future electricity load of the power-consuming factories and the weights and thresholds of the electricity load forecasting model of the power-consuming factories to meet the preset accuracy requirements based on the preliminary encrypted electricity load forecast vector of the power-consuming factories and a randomly generated password; training the industrial zone electricity load forecasting model based on the predicted future electricity load of the power-consuming factories and the weights and thresholds of the electricity load forecasting model of the power-consuming factories to meet the preset accuracy requirements; repeating the above operations until the industrial zone electricity load forecasting model reaches the preset accuracy requirements, and predicting the electricity load of the industrial zone through the industrial zone electricity load model, thereby accurately predicting the electricity load of the industrial zone for a future period of time by establishing the electricity load forecasting models of the power-consuming factories and the industrial zone electricity load forecasting model, so as to adjust the electricity load of the industrial zone in a timely manner according to the forecast results and avoid the situation of excessive or insufficient electricity load. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the electrical load prediction device structure in an industrial area, which is part of the hardware operating environment of the embodiment of the present invention.

[0043] Figure 2 This is a flowchart illustrating the first embodiment of the electricity load prediction method for industrial zones according to the present invention.

[0044] Figure 3 This is a flowchart illustrating the second embodiment of the electricity load forecasting method for industrial zones of the present invention.

[0045] Figure 4 This is a flowchart illustrating the third embodiment of the electricity load prediction method for industrial zones according to the present invention.

[0046] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0048] Reference Figure 1 , Figure 1 This is a schematic diagram of the electrical load prediction device in an industrial area, which is part of the hardware operating environment of the embodiment of the present invention.

[0049] like Figure 1 As shown, the power load forecasting equipment for this industrial area may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen, and optionally, it may also include a standard wired interface or a wireless interface. In this invention, the wired interface of the user interface 1003 may be a USB interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (RAM) or a non-volatile memory (NVM), such as a disk storage device. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0050] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the electrical load forecasting equipment for industrial areas, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0051] like Figure 1 As shown, the memory 1005, which is identified as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an industrial area power load prediction program.

[0052] exist Figure 1In the power load forecasting device for the industrial area shown, the network interface 1004 is mainly used to connect to the backend server and communicate with the backend server; the user interface 1003 is mainly used to connect to user equipment; the power load forecasting device for the industrial area calls the power load forecasting program for the industrial area stored in the memory 1005 through the processor 1001 and executes the power load forecasting method for the industrial area provided in this embodiment of the invention.

[0053] Based on the above hardware structure, an embodiment of the power load prediction method for industrial zones of the present invention is proposed.

[0054] Reference Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the power load forecasting method for industrial areas according to the present invention. The first embodiment of the power load forecasting method for industrial areas according to the present invention is presented.

[0055] In the first embodiment, the method for predicting the electricity load of the industrial area includes the following steps:

[0056] Step S10: Receive the non-uniform continuous-time hidden semi-Markov model sent by the N edge computing terminals, and decrypt it to obtain the preliminary encrypted power load prediction vector of the power-consuming factory.

[0057] It should be understood that the execution entity in this embodiment is a cloud server, which has functions such as data processing, data communication and program execution.

[0058] It should be noted that the state sequence is obtained from the kernel function of the multi-state degenerate model of a non-uniform continuous-time hidden semi-Markov process. The state sequence is then transformed using a transformation rate function to obtain the kernel expression of the kernel function of the multi-state degenerate model of the non-uniform continuous-time hidden semi-Markov process. Thus, the kernel expression can be obtained. The non-uniform continuous-time hidden semi-Markov model is a type of Markov chain. Its states cannot be directly observed, but can be observed through a sequence of observation vectors. Each observation vector represents various states through certain probability density distributions, and each observation vector is generated by a state sequence with a corresponding probability density distribution. This means that the non-uniform continuous-time hidden semi-Markov model is a doubly stochastic process. The non-uniform continuous-time hidden semi-Markov model can be applied to various machine learning scenarios.

[0059] In practice, the method of generating a non-uniform continuous-time hidden semi-Markov model based on the password assumes that the state transition matrix A and the initial state transition probability vector ∏ have been randomly given. Then, the N accumulated values ​​of the conditional probability, i.e., αT(i), are obtained by the forward algorithm. Since the password P and the N accumulated values ​​are known, it means that a system of N equations can be obtained. The unknown in the equations is the observation probability matrix B. Therefore, before solving the equations, N*(N-1) values ​​of the observation probability matrix B need to be randomly generated. The remaining N values ​​can be obtained by solving the equations, thus obtaining the complete non-uniform continuous-time hidden semi-Markov model.

[0060] Step S20: Based on the pre-encrypted power load prediction vector of the power-consuming factory and the randomly generated password, the predicted future power load of the power-consuming factory and the weights and thresholds of the power load prediction model of the power-consuming factory that meet the preset accuracy requirements are obtained.

[0061] In specific implementation, the predicted future electricity load of the power-consuming factory and the weights and thresholds of the electricity load prediction model that meet the preset accuracy requirements are extracted using a convolutional neural network to obtain the industrial area electricity load feature vector. An industrial area electricity load feature vector library is then established based on this feature vector. The industrial area electricity load feature vectors in the library are classified using a classifier, and the classified industrial area electricity load feature vectors are subjected to iterative clustering to update the classifier and cluster centers. Based on the updated classifier and cluster centers, Updating the industrial zone electricity load prediction model includes: randomly selecting K industrial zone electricity load feature vectors from the industrial zone electricity load feature vector library as initial cluster centers, where K is an integer greater than 1; calculating the distance between the remaining industrial zone electricity load feature vectors in the industrial zone electricity load feature vector library and the initial cluster centers; dividing the industrial zone electricity load feature vectors into K clusters based on the distance between each industrial zone electricity load feature vector and the initial cluster centers; calculating the mean of all feature vectors in the K clusters, and using the mean as the updated cluster center.

[0062] Step S30: Train the industrial zone electricity load prediction model based on the predicted future electricity load of the power-consuming factory and the weights and thresholds of the power load prediction model of the power-consuming factory that meet the preset accuracy requirements.

[0063] It should be noted that the industrial area electricity load feature vector here adopts the k-cluster mean model. The obtained industrial area electricity load feature vector is added to the sample data to obtain new sample data; the industrial area electricity load feature vector is then used for detection to obtain the detection result.

[0064] Step S40: Repeat the above operation until the industrial zone power load prediction model reaches the preset accuracy requirement, and predict the power load of the industrial zone through the industrial zone power load model.

[0065] In practical implementation, the backpropagation algorithm is currently the most commonly used and effective algorithm for training Artificial Neural Networks (ANNs). Its main idea is as follows: the training set data is input into the input layer of the ANN, passes through the hidden layer, and finally reaches the output layer to output the result. This is the forward propagation process of the ANN. Since there is an error between the output result of the ANN and the actual result, the error between the estimated value and the actual value is calculated first, and this error is propagated backward from the output layer to the hidden layer until it reaches the input layer. During the backpropagation process, the values ​​of various parameters are adjusted according to the error. The above process is iterated continuously until convergence.

[0066] In this embodiment, the power load prediction method for the industrial zone is applied to a cloud server; it includes the following steps: receiving the non-uniform continuous-time hidden semi-Markov model sent by the N edge computing terminals, and decrypting it to obtain a preliminary encrypted power load prediction vector for the power-consuming factory; restoring the predicted future power load of the power-consuming factory and the weights and thresholds of the power load prediction model of the power-consuming factory to meet the preset accuracy requirements based on the preliminary encrypted power load prediction vector and a randomly generated password; training the power load prediction model for the industrial zone based on the predicted future power load of the power-consuming factory and the weights and thresholds of the power load prediction model of the power-consuming factory to meet the preset accuracy requirements; repeating the above operations until the power load prediction model for the industrial zone reaches the preset accuracy requirements, and predicting the power load of the industrial zone through the power load prediction model, thereby receiving data transmitted from the edge computing terminals, establishing and training the power load prediction model for the industrial zone based on the data, and accurately predicting the power load of the industrial zone through the power load prediction model for the industrial zone.

[0067] Reference Figure 3 , Figure 3 This is a flowchart illustrating the second embodiment of the electricity load forecasting method for industrial zones of the present invention, based on the above. Figure 2 The first embodiment shown presents a second embodiment of the method for predicting the electricity load of industrial areas according to the present invention.

[0068] In the second embodiment, before step S10, the method further includes:

[0069] Step S01: Obtain information on the holiday schedules of power-consuming factories, their historical electricity load, and weather forecast data.

[0070] It should be understood that the execution subject of this embodiment is an edge computing terminal, which has functions such as data processing, data communication and program execution.

[0071] It should be noted that obtaining information on the holiday schedules of power-consuming factories, their historical electricity loads, and weather forecasts can affect their production, and further impact their overall production.

[0072] Step S02: Establish an electricity load prediction model for the power-consuming factory, and update the parameters of the electricity load prediction model based on the factory's holiday schedule information, historical electricity load, and weather forecast data.

[0073] In specific implementation, the holiday schedule information of the power-consuming factories, the historical power load of the power-consuming factories, and the weather forecast data are input into the power load prediction model of the power-consuming factories to obtain the expected power load prediction results of the power-consuming factories. Based on the expected power load prediction results and the actual power load prediction results of the power-consuming factories, the estimated values ​​of weights and thresholds are obtained. The estimated values ​​of weights are subjected to Taylor expansion, and the linear terms of Taylor expansion are retained. Based on the linear terms of Taylor expansion, the error function of the system is obtained, and nonlinear recursive least squares tracking is performed on the error function of the system to obtain the weights and thresholds after least squares tracking.

[0074] It should be noted that the weights of the neural network are used as system parameters to reflect how the system's output changes with the input. The Recursive Least Squares (RLS) method is used to calculate these weights. These weights generally conform to the random walk law, and the relationship between the weights at time k and time k-1 can be recursively calculated given an initial value. This process has been proven to be convergent, but it still suffers from slow convergence speed and low identification accuracy, thus requiring improvement. As can be seen from the preceding algorithm recursive formula, a key process of the algorithm is the change of P(k). The method of this change has been modified to simplify it, and then it is applied as a learning algorithm to neural network learning.

[0075] It should be understood that the power plant's holiday schedule information, historical power load, and weather forecast data are input into the power load prediction model to obtain the actual output. Based on the expected output and the actual output, the estimated weights are obtained. The estimated weights are then subjected to a Taylor expansion, retaining the linear terms of the Taylor expansion. Based on the linear terms of the Taylor expansion, the system's error function is obtained, and nonlinear recursive least squares tracking is performed on the system's error function to obtain the least squares-tracked weights. Based on the linear terms of the Taylor expansion, the error function is calculated, and it is determined whether the error function meets the preset accuracy requirements. If the error function is less than the expected convergence accuracy, the gradient direction is determined based on the error function. When the gradient direction approaches zero along the x-axis, the first and second coordinate directions are selected as the search directions, which are the coordinate directions corresponding to the weights. The weights are then adjusted according to the coordinate directions corresponding to the weights to obtain new weights.

[0076] Step S03: Repeat the above operation until the power load prediction model of the power plant reaches the preset power load prediction accuracy requirement of the power plant, and predict the future power load of the power plant by using the power load prediction model of the power plant that meets the preset power load prediction accuracy requirement.

[0077] In specific implementation, after predicting the future electricity load of the power-consuming factory, the method further includes: encrypting the predicted future electricity load of the power-consuming factory and the weights and thresholds of the power load prediction model of the power-consuming factory to meet preset accuracy requirements using a randomly generated password based on a non-uniform continuous-time hidden semi-Markov model, thereby obtaining a preliminary encrypted electricity load prediction vector of the power-consuming factory; constructing a corresponding non-uniform continuous-time hidden semi-Markov model in reverse based on the preliminary encrypted electricity load prediction vector, the randomly generated password, and the state transition matrix and initial state transition probability vector of the non-uniform continuous-time hidden semi-Markov model randomly corresponding to the password; and sending the corresponding non-uniform continuous-time hidden semi-Markov model constructed in reverse to a cloud server for decryption, and updating the prediction model based on the decrypted data.

[0078] In this embodiment, the power load forecasting method for industrial areas is applied to N edge computing terminals, where N is an integer greater than 1; it includes the following steps: acquiring the holiday schedule information of power-consuming factories, the historical power load of power-consuming factories, and weather forecast data; establishing a power load forecasting model for power-consuming factories, and updating the parameters of the power load forecasting model based on the holiday schedule information, the historical power load of power-consuming factories, and weather forecast data; repeating the above operations until the power load forecasting model of the power-consuming factories reaches the preset power load forecasting accuracy requirement of the power-consuming factories, and predicting the future power load of the power-consuming factories using the power load forecasting model of the power-consuming factories that meets the preset power load forecasting accuracy requirement; the power load forecasting method for industrial areas is applied to cloud servers; the power load forecasting method for industrial areas includes the following steps: connecting The process involves receiving the non-uniform continuous-time hidden semi-Markov model sent by the N edge computing terminals, decrypting it to obtain a preliminary encrypted power load prediction vector for the power-consuming factories; restoring the predicted future power load of the power-consuming factories and the weights and thresholds of the power load prediction model to meet the preset accuracy requirements based on the preliminary encrypted power load prediction vector and a randomly generated password; training the industrial zone power load prediction model based on the predicted future power load of the power-consuming factories and the weights and thresholds of the power load prediction model to meet the preset accuracy requirements; repeating the above operations until the industrial zone power load prediction model reaches the preset accuracy requirements, and predicting the power load of the industrial zone through the industrial zone power load model, thereby realizing the prediction of the future power load of each power-consuming factory and the future power load of the entire industrial zone.

[0079] Reference Figure 4 , Figure 4 This is a flowchart illustrating the third embodiment of the electricity load forecasting method for industrial zones of the present invention, based on the above. Figure 2 The first embodiment shown presents a third embodiment of the present invention for predicting the electricity load of industrial areas.

[0080] In the third embodiment, step S30 includes:

[0081] Step S301: The predicted future electricity load of the power-consuming factory and the weights and thresholds of the electricity load prediction model of the power-consuming factory that meet the preset accuracy requirements are extracted by a convolutional neural network to obtain the industrial area electricity load feature vector, and an industrial area electricity load feature vector library is established based on the industrial area electricity load feature vector.

[0082] In specific implementation, a preprocessing index set is generated based on the predicted future electricity load of the power-consuming plant and the weights and thresholds of the power load prediction model of the power-consuming plant that meet the preset accuracy requirements. The header data index of the preprocessing index set is extracted and placed at the end of the refined index set. A second data index is selected from the index set after extracting the header data index, and then placed at the end of the refined index set. A third data index is selected from the index set after extracting the second data index, based on the header data index, the second data index, and the current iteration number, and then placed at the end of the refined index set. The index set after extracting the third data index is determined as the preprocessing index set, and the iterative processing continues until the current preprocessing index is determined. The target dataset is an empty set. The refined index set and the preprocessed index set before iterative processing are concatenated to form the target dataset. The target dataset is input into a pre-obtained convolutional neural network to obtain the power load feature vector of the industrial area. An original index set is generated based on the predicted future power load of the power-consuming factory and the weights and thresholds of the power load prediction model of the power-consuming factory that meet the preset accuracy requirements. One data index is randomly selected from the original index set and determined as the benchmark index. The benchmark index is placed at the end of the preprocessed index set. The data index with the lowest correlation with the benchmark index in the original index set after the benchmark index is removed is determined as the benchmark index, and the iterative processing continues. The above operations are repeated until the original index set generates the preprocessed index set.

[0083] It should be understood that in each iteration, the benchmark indicator is placed at the end of the preprocessed indicator set, which is equivalent to arranging the data indicators in the order they were placed in the preprocessed indicator set. In specific implementation, a pointer is set to point to the benchmark indicator, making it the head of the preprocessed indicator set; the initial preprocessed indicator set is empty, which is equivalent to placing the benchmark indicator at the end of the preprocessed indicator set. Subsequent indicators are sorted according to their correlation with each other. The data indicator with the lowest correlation to the benchmark indicator is selected as the next data indicator in the preprocessed indicator set. Then, the pointer moves one step to the tail to select the next data indicator as the benchmark indicator. Similarly, the indicator with the lowest correlation to the benchmark indicator is selected as the next data indicator. Data indicators that have already been sorted do not participate in the correlation calculation. The above steps are repeated until the pointer points to a data indicator. At this point, the original indicator set is empty, and the final arrangement is the preprocessed indicator set. A convolutional neural network is then used to extract the industrial area's electricity load feature vector.

[0084] Step S302: Classify the industrial area electricity load feature vectors in the industrial area electricity load feature vector library using a classifier, and perform clustering iteration on the classified industrial area electricity load feature vectors to update the classifier and cluster centers.

[0085] It should be noted that the process involves performing Softmax classification on the industrial zone's electricity load feature vectors, using the cross-entropy loss function to obtain the classification result. This loss function is then processed using the cross-entropy loss, classification labels, and the softmax classification result to obtain the classification loss. The classification loss is then backpropagated through gradient propagation to update the classifier parameters. The process continues until all industrial zone electricity load feature vectors are classified, resulting in the target classifier. K industrial zone electricity load feature vectors are randomly selected from the sample library as initial cluster centers. In this context, K is an integer greater than 1; calculate the distance between each remaining industrial area electricity load feature vector in the industrial area electricity load feature vector library and the initial cluster center; based on the distance between each industrial area electricity load feature vector and the initial cluster center, divide the industrial area electricity load feature vector into K clusters; calculate the mean of all feature vectors in the K clusters, and use the mean as the new cluster center; return to the step of randomly selecting K industrial area electricity load feature vectors from the industrial area electricity load feature vector library as initial cluster centers, until the position of the new cluster center no longer changes, the iteration stops, the target cluster center is obtained, and the model is trained for the testing phase.

[0086] Step S303: Update the industrial zone electricity load prediction model based on the updated classifier and cluster centers.

[0087] In specific implementation, based on the target cluster centers, the distance between the electricity load feature vector of each industrial zone and each cluster center is calculated; based on the distance between the current electricity load feature vector of each industrial zone and each cluster center, the cluster to which the electricity load feature vector of each industrial zone belongs is determined, and the clusters of industrial zone electricity load feature vectors with predicted results are selected; the clusters of feature vectors with predicted results are classified using a target classifier to obtain the classification results. The above operations are repeated until the prediction of future industrial zone electricity load is completed.

[0088] In this embodiment, the non-uniform continuous-time hidden semi-Markov model sent by the N edge computing terminals is received and decrypted to obtain a preliminary encrypted power load prediction vector for the power-consuming factory. Based on the preliminary encrypted power load prediction vector and a randomly generated password, the predicted future power load of the power-consuming factory and the weights and thresholds of the power load prediction model satisfying the preset accuracy requirements are obtained. Features of the predicted future power load and the weights and thresholds of the power load prediction model satisfying the preset accuracy requirements are extracted using a convolutional neural network to obtain an industrial area power load feature vector. An industrial area power load feature vector library is established based on the industrial area power load feature vector. The industrial area power load feature vectors in the library are classified using a classifier, and the classified industrial area power load feature vectors are clustered iteratively to update the classifier and cluster centers. The industrial area power load prediction model is updated based on the updated classifier and cluster centers. Repeat the above operations until the industrial zone electricity load prediction model reaches the preset accuracy requirement, and predict the industrial zone electricity load through the industrial zone electricity load model, thereby training and updating the industrial zone electricity load prediction model to accurately predict the future electricity load of the industrial zone.

[0089] Furthermore, this embodiment of the invention also proposes a storage medium storing an industrial area power load prediction program, wherein when the industrial area power load prediction is executed by a processor, the steps of the industrial area power load prediction method described above are implemented.

[0090] Since this storage medium can adopt the technical solutions of all the above embodiments, it has at least the beneficial effects brought about by the technical solutions of the above embodiments, which will not be described in detail here.

[0091] Furthermore, this embodiment of the invention also proposes an industrial zone power load forecasting system, which includes the aforementioned N edge computing terminals and the aforementioned cloud server.

[0092] Since the power load forecasting system for industrial areas can adopt the technical solutions of all the above embodiments, it has at least the beneficial effects brought about by the technical solutions of the above embodiments, and will not be described in detail here.

[0093] Other embodiments or specific implementations of the power load forecasting system for industrial areas described in this invention can refer to the above-described method embodiments, and therefore have at least all the beneficial effects brought about by the technical solutions of the above embodiments, which will not be repeated here.

[0094] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0095] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. In the unit claims listing several devices, several of these devices may be embodied by the same hardware item. The use of the terms first, second, and third, etc., does not indicate any order and can be interpreted as names.

[0096] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0097] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0098] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0099] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0100] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. An industrial park power consumption load forecasting system characterized by, The power load prediction system of the industrial zone comprises N edge computing terminals and a cloud server; The edge computing terminal comprises a memory, a processor and an industrial zone power load prediction program stored in the memory and executable on the processor; the processor executes the following steps: obtaining holiday arrangement information of a power plant, historical power load of the power plant and weather forecast data; establishing a power load prediction model of the power plant, and updating parameters of the power load prediction model according to the holiday arrangement information of the power plant, the historical power load of the power plant and the weather forecast data; repeating the above operation until the power load prediction model of the power plant reaches a preset power load prediction accuracy requirement of the power plant, and predicting future power load of the power plant through the power load prediction model of the power plant meeting the preset power load prediction accuracy requirement of the power plant; According to the non-uniform continuous time hidden semi-Markov model, the predicted future power load of the power plant and the weight and threshold of the power load prediction model of the power plant meeting the preset accuracy requirement are encrypted by a randomly generated password to obtain a preliminary encrypted power load prediction vector of the power plant; According to the preliminary encrypted power load prediction vector of the power plant, the randomly generated password and the state transition matrix and initial state transition probability vector of the non-uniform continuous time hidden semi-Markov model corresponding to the password, a corresponding non-uniform continuous time hidden semi-Markov model is reversely constructed; the reversely constructed corresponding non-uniform continuous time hidden semi-Markov model is sent to the cloud server; The cloud server comprises a memory, a processor and an industrial zone power load prediction program stored in the memory and executable on the processor; the processor executes the following steps: receiving the non-uniform continuous time hidden semi-Markov model sent by the N edge computing terminals, and decrypting to obtain a preliminary encrypted power load prediction vector of the power plant; restoring the predicted future power load of the power plant and the weight and threshold of the power load prediction model of the power plant meeting the preset accuracy requirement according to the preliminary encrypted power load prediction vector of the power plant and the randomly generated password; According to the predicted future power load of the power plant and the weight and threshold of the power load prediction model of the power plant meeting the preset accuracy requirement, an industrial zone power load prediction model is trained; the above operation is repeated until the industrial zone power load prediction model reaches a preset accuracy requirement, and the power load of the industrial zone is predicted through the industrial zone power load model.

2. The industrial park's electricity consumption load forecasting system of claim 1, wherein, The parameter updating of the power load prediction model according to the holiday arrangement information of the power plant, the historical power load of the power plant and the weather forecast data comprises: The vacation arrangement information of the power consumer plant, the historical power consumption load of the power consumer plant, and meteorological forecast data are input into the power consumption load prediction model of the power consumer plant to obtain a predicted power consumption load of the power consumer plant, and an estimated value of a weight and a threshold is obtained according to the predicted power consumption load of the power consumer plant and an actual power consumption load of the power consumer plant; Taylor expansion is performed on the estimated value of the weight, and a linear term of the Taylor expansion is retained; According to the linear term of the Taylor expansion, an error function of the system is obtained, and nonlinear recursive least squares tracking is performed on the error function of the system to obtain a weight and a threshold after least squares tracking.

3. The industrial park's electricity consumption load forecasting system of claim 1, wherein, The training of the industrial area power consumption load prediction model according to the predicted future power consumption load of the power consumer plant and the weight and the threshold of the power consumption load prediction model of the power consumer plant satisfying the preset accuracy requirement comprises: Feature extraction is performed on the predicted future power consumption load of the power consumer plant and the weight and the threshold of the power consumption load prediction model of the power consumer plant satisfying the preset accuracy requirement through a convolutional neural network to obtain an industrial area power consumption load feature vector, and an industrial area power consumption load feature vector library is established according to the industrial area power consumption load feature vector; Classification is performed on the industrial area power consumption load feature vectors in the industrial area power consumption load feature vector library through a classifier, and the classified industrial area power consumption load feature vectors are iteratively clustered to update the classifier and the cluster centers; The industrial area power consumption load prediction model is updated according to the updated classifier and the cluster centers.

4. The power consumption load forecasting system for an industrial park according to claim 3, wherein The classification of the industrial area power consumption load feature vectors in the industrial area power consumption load feature vector library through the classifier and the updating of the classifier comprise: The industrial area power consumption load feature vectors are classified to obtain a classification result and a loss function; The loss function is processed through a cross-entropy loss, a classification label, and the classification result to obtain a classification loss; The classification loss is gradient backpropagated through a backpropagation relationship, and the classifier parameters are updated.

5. The power consumption load forecasting system for an industrial park according to claim 3 or 4, characterized by, The iterative clustering of the classified industrial area power consumption load feature vectors and the updating of the cluster centers comprise: K industrial area power consumption load feature vectors in the industrial area power consumption load feature vector library are randomly extracted as initial cluster centers, where K is an integer greater than 1; The distances between the remaining industrial area power consumption load feature vectors in the industrial area power consumption load feature vector library and the initial cluster centers are calculated; According to the distances between each industrial area power consumption load feature vector and the initial cluster centers, the industrial area power consumption load feature vectors are divided into K clusters; The means of all feature vectors of the K clusters are calculated, and the means are taken as updated cluster centers.

Images