Virtual power plant power load switching control method and system

Abstract

The invention relates to a virtual power plant power load switching control method and system, and the method comprises the steps: obtaining the power load data of a virtual power plant and the corresponding power supply capability data, and generating a control strategy training sample set based on the power load data and the corresponding power supply capability data; analyzing each power supply capability data in the training sample set based on the stability of the power system to obtain a plurality of sub-supply capability data of each power supply capability; constructing a power load prediction and control model, and training the power load prediction and control model based on the power load data and the power supply capability data in the training sample set and multiple sub-supply capability data of each power supply capability to obtain a power load switching control model; and inputting power load data of a to-be-controlled virtual power plant into the power load switching control model to obtain a power load switching strategy of the to-be-controlled virtual power plant. Therefore, the reliability and flexibility of power supply are improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of power system management, and in particular to a virtual power plant power load switching control method and system. Background Art

[0002] In the existing power system, virtual power plants, as a new type of power resource management method, achieve optimal configuration and scheduling of power resources by integrating distributed power generation resources, energy storage equipment and controllable loads. Virtual power plants can effectively improve the operating efficiency and reliability of power systems, reduce operating costs, and promote the use of renewable energy. However, with the continuous development of the power market and the increasing complexity of power demand, traditional power load control methods have been unable to meet the requirements of modern power systems for flexibility and response speed.

[0003] Existing power load control methods usually rely on static load forecasting models and simple control strategies. These methods are often unable to accurately predict the dynamic changes in power demand and are difficult to adapt to fluctuations in power supply capacity. In addition, traditional control strategies lack in-depth analysis of power system stability, making it difficult to optimize economic performance and environmental impact while ensuring power supply reliability. In terms of power supply capacity analysis, existing technologies often ignore the subdivision and dynamic adjustment of power supply capacity, resulting in the inability to fully utilize the potential of power resources and difficulty in coping with rapid changes in the power market.

[0004] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the prior art: the existing methods lack effective data preprocessing and feature extraction means when processing power load and supply capacity data, resulting in low model training efficiency and limited prediction accuracy; at the same time, the prior art fails to fully consider the stability of the power system and multi-objective optimization requirements in the power load switching control, resulting in insufficient adaptability and comprehensive performance of the control strategy. Summary of the invention

[0005] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0006] The purpose of the embodiments of the present invention is to provide a virtual power plant power load switching control method and system, which, through a series of innovative training methods and technical strategies, achieves efficient, stable and economical operation of the power system, while reducing environmental impact and improving the reliability and flexibility of power supply.

[0007] In a first aspect, an embodiment of the present invention relates to a method for controlling power load switching of a virtual power plant, the method comprising:

[0008] Acquire power load data and corresponding power supply capacity data of the virtual power plant, and generate a control strategy training sample set based on the power load data and the corresponding power supply capacity data;

[0009] Analyze each power supply capacity data in the training sample set based on the stability of the power system to obtain multiple sub-supply capacity data of each power supply capacity;

[0010] Constructing a power load prediction and control model, and training the power load prediction and control model based on the power load data, power supply capacity data, and multiple sub-supply capacity data of each power supply capacity in the training sample set to obtain a power load switching control model;

[0011] The power load data of the virtual power plant to be controlled is input into the power load switching control model to obtain the power load switching strategy of the virtual power plant to be controlled.

[0012] Furthermore, each power supply capacity data in the training sample set is analyzed based on the stability of the power system to obtain multiple sub-supply capacity data of each power supply capacity, including:

[0013] The initial state of each power supply capacity is taken as the first type, and the stability of the state is calculated; each state of the power supply capacity is traversed in turn, and the stability of the current state is calculated. If the stability of the current state is the same as the stability of the previous state, the current state and the previous state are of the same type, otherwise the current state and the previous state are of different types;

[0014] A plurality of sub-supply capability data of the power supply capability is obtained according to the type of each state in the power supply capability.

[0015] Furthermore, according to the sub-supply capacity data of the power supply capacity, a composite loss function with load forecast error penalty is used to calculate the training loss of the power load forecasting and control model.

[0016] Furthermore, the load forecast error penalty matrix corresponding to the power supply matrix is ​​calculated by the following steps:

[0017] Calculate the intra-mask Euclidean distance matrix EDM of the mth sub-mask matrix based on the mth sub-mask matrix of the power demand matrix m ; Based on the Euclidean distance matrix EDM within the mask m Calculate the intra-mask penalty term penal of the mth sub-mask matrix int,m ;

[0018] Calculate the masked Euclidean distance matrix EDM of the mth sub-mask matrix based on the mth sub-mask matrix of the power supply matrix ext,m ; Based on the masked Euclidean distance matrix EDM ext,m Calculate the out-of-mask penalty term penal of the mth sub-mask matrix ext,m ;

[0019] The penalty term penal in the mask int,m And the mask out-of-band penalty penal ext,m Normalize and add to get the distance penalty matrix penal corresponding to the mth sub-mask matrix m ;

[0020] The distance penalty matrix penal corresponding to each sub-mask matrix m The load forecast error penalty matrix Penal corresponding to the power supply matrix is ​​obtained by splicing.

[0021] Furthermore, the intra-mask Euclidean distance matrix EDM of the m-th sub-mask matrix m The same dimension as the mth sub-mask matrix of the power demand matrix;

[0022] Compute the intra-mask penalty term for the mth sub-mask matrix.

[0023] Calculate the out-of-mask penalty term for the m-th sub-mask matrix.

[0024] Furthermore, the outer mask Euclidean distance matrix EDM of the m-th sub-mask matrix ext,m The same dimension as the mth sub-mask matrix of the power supply matrix;

[0025] In a second aspect, an embodiment of the present invention relates to a virtual power plant power load switching control system, comprising:

[0026] A feature extraction module, for extracting features of power load data and power supply capacity data using a deep learning network;

[0027] A prediction module for predicting future power loads using a time series prediction algorithm based on the extracted features;

[0028] An optimization module, for generating a power load switching control strategy using an optimization algorithm based on the prediction results and power supply capacity data;

[0029] Evaluation module for evaluating the performance of power load switching control strategies, including economics, reliability, and environmental impact.

[0030] Further, generating a control strategy training sample set based on the power load data and the corresponding power supply capacity data includes:

[0031] Gaussian process regression is used to interpolate power load data and corresponding power supply capacity data to handle missing data;

[0032] Perform wavelet transform on power load data and power supply capacity data to extract time-frequency features;

[0033] For each sample data, principal component analysis is performed on the power load data and power supply capacity data to reduce the data dimension and improve the model training efficiency;

[0034] Based on the data after dimensionality reduction, a clustering algorithm is used to classify the power load types and power supply capacity types to generate training samples.

[0035] The above-mentioned embodiments of the present disclosure have the following beneficial effects: the use of the virtual power plant power load switching control method of the present invention can improve the operating efficiency and stability of the power system. Through the feature extraction of the deep learning network, combined with the time series prediction algorithm, the future power load can be predicted more accurately, thereby making the power load switching control strategy more accurate and timely. In addition, the power load switching control strategy generated by the optimization algorithm can optimize the economy, reliability and environmental impact while ensuring the reliability of power supply, and realize the efficient use of power resources.

[0036] Furthermore, the present invention can effectively reduce the prediction error of the model during the training process and improve the accuracy of power load prediction by introducing a composite loss function with load prediction error penalty. At the same time, by calculating the load prediction error penalty matrix corresponding to the power supply matrix, the uncertainty and variability in the power supply capacity data can be better handled, making the power load switching control strategy more flexible and adaptable. The application of these technical means can not only improve the response speed and regulation capability of the power system, but also reduce operating costs and improve the economic benefits and environmental friendliness of the power system.

[0037] In addition, the invention summary does not list all the features required for the embodiments of the present invention, and other combinations of these feature groups can also become embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, in which:

[0039] Figure 1 A schematic flow chart of a method for controlling power load switching in a virtual power plant provided in one embodiment of the present invention;

[0040] Figure 2A schematic diagram of the structure of a virtual power plant power load switching control system provided by one embodiment of the present invention;

[0041] Figure 3 A schematic diagram schematically shows the structure of a computing device according to an embodiment of the present invention;

[0042] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0043] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0044] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, device, apparatus, method or computer program product. Therefore, the present disclosure may be implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0045] According to an embodiment of the present invention, a method and system for controlling power load switching of a virtual power plant are proposed.

[0046] It should be noted that any number of elements in the drawings is for illustration rather than limitation, and any naming is only for distinction and does not have any limiting meaning.

[0047] The principle and spirit of the present invention are explained in detail below with reference to several representative embodiments of the present invention.

[0048] Reference below Figure 1 , Figure 1 A flowchart of a method for controlling power load switching in a virtual power plant provided by an embodiment of the present invention is shown in FIG. It should be noted that the embodiment of the present invention can be applied to any applicable scenario.

[0049] Figure 1 The virtual power plant power load switching control method provided by one embodiment of the present invention shown in the figure includes:

[0050] Step 101, obtaining power load data and corresponding power supply capacity data of a virtual power plant, and generating a control strategy training sample set based on the power load data and the corresponding power supply capacity data;

[0051] Step 102, analyzing each power supply capacity data in the training sample set based on the stability of the power system to obtain multiple sub-supply capacity data of each power supply capacity;

[0052] Step 103, constructing a power load prediction and control model, training the power load prediction and control model based on the power load data, power supply capacity data and multiple sub-supply capacity data of each power supply capacity in the training sample set to obtain a power load switching control model;

[0053] Step 104 , inputting the power load data of the virtual power plant to be controlled into the power load switching control model to obtain the power load switching strategy of the virtual power plant to be controlled.

[0054] It should be noted that this embodiment relates to a virtual power plant power load switching control method, which obtains the power load data of the virtual power plant and the corresponding power supply capacity data, and generates a control strategy training sample set based on these data. This allows the system to accurately match power supply and demand and optimize the allocation and use of power resources.

[0055] Specifically, obtaining power load data and power supply capacity data involves collecting real-time data from various nodes of the virtual power plant, including but not limited to power generation, power consumption, energy storage status, etc. These data are collected through smart meters, sensors, and data acquisition systems. The generation of control strategy training sample sets depends on data preprocessing, feature extraction, and label definition to ensure that the model can learn effective control strategies from historical data.

[0056] Preferably, in order to improve the quality and availability of data, data cleaning and missing value processing techniques, such as interpolation, time series prediction, etc., can be used to ensure the integrity and accuracy of the training sample set. In addition, machine learning algorithms, such as random forests, support vector machines or neural networks, can be used to build power load forecasting and control models, and cross-validation and other methods can be used to optimize model parameters and improve the generalization ability of the model.

[0057] In some embodiments, each power supply capability data in the training sample set is analyzed based on the stability of the power system to obtain multiple sub-supply capability data of each power supply capability, including:

[0058] The initial state of each power supply capacity is taken as the first type, and the stability of the state is calculated; each state of the power supply capacity is traversed in turn, and the stability of the current state is calculated. If the stability of the current state is the same as the stability of the previous state, the current state and the previous state are of the same type, otherwise the current state and the previous state are of different types;

[0059] A plurality of sub-supply capability data of the power supply capability is obtained according to the type of each state in the power supply capability.

[0060] Specifically, the stability analysis of power supply capacity includes evaluating the initial state of each power supply capacity and calculating its stability index. Stability can be measured in many ways, for example, it can be defined based on the volatility of power supply, frequency response characteristics, or supply and demand balance. In practical applications, a series of parameters can be set to quantify stability, such as supply fluctuation threshold, frequency deviation tolerance, etc. Through these parameters, the stability of each state can be determined, and the power supply capacity can be divided into different sub-supply capacities accordingly.

[0061] Preferably, in order to more accurately perform stability analysis of power supply capacity, advanced data analysis techniques, such as machine learning algorithms, can be used to automatically identify and classify the stability status of power supply capacity. In addition, adaptive algorithms can be introduced to dynamically adjust the parameters of stability analysis based on real-time data to adapt to changes in power system operating conditions. When classifying power supply capacity, in addition to stability, other factors such as cost-effectiveness, environmental impact, etc. can also be considered to achieve more comprehensive power supply capacity management.

[0062] In some embodiments, according to the sub-supply capacity data of the power supply capacity, a composite loss function with load forecast error penalty is used to calculate the training loss of the power load forecasting and control model.

[0063] The calculation formula of the composite loss function is:

[0064]

[0065] Where L represents the total loss value, N represents the total number of power supply capacity data, and w i represents the weight of the i-th power supply capacity, PD i represents the ith power demand matrix, PS i represents the i-th power supply matrix, penal i represents the load forecasting error penalty matrix corresponding to the i-th power supply matrix, λ represents the regularization parameter, and ||·|| represents the Frobenius norm of the matrix.

[0066] Specifically, the calculation of the composite loss function involves multiple parameters and concepts. Among them, the total loss value is the objective function that needs to be minimized during the model training process. It consists of multiple parts, including the weight of the power supply capacity data, the power demand matrix, the power supply matrix, and the load forecast error penalty matrix. These parameters can be set based on historical data and expert knowledge. For example, weights can be assigned according to the importance of power supply capacity, and the forecast error penalty matrix can be adjusted according to the size of the forecast error.

[0067] Preferably, in order to further improve the training effect of the model, a variety of methods can be used to refine the calculation of the composite loss function. For example, a regularization parameter can be introduced to prevent the model from overfitting, or different distance metrics can be used to calculate the prediction error. In addition, different optimization algorithms can be considered to solve the minimum value of the loss function, such as gradient descent, stochastic gradient descent, or Adam optimizer. These methods can be selected according to the complexity of the model and the characteristics of the training data to achieve the best training effect.

[0068] In some embodiments, the load forecast error penalty matrix corresponding to the power supply matrix is ​​calculated by the following steps:

[0069] Calculate the intra-mask Euclidean distance matrix EDM of the mth sub-mask matrix based on the mth sub-mask matrix of the power demand matrix m ; Based on the Euclidean distance matrix EDM within the mask m Calculate the intra-mask penalty term penal of the mth sub-mask matrix int,m ;

[0070] Calculate the masked Euclidean distance matrix EDM of the mth sub-mask matrix based on the mth sub-mask matrix of the power supply matrix ext,m ; Based on the masked Euclidean distance matrix EDM ext,m Calculate the out-of-mask penalty term penal of the mth sub-mask matrix ext,m ;

[0071] The penalty term penal in the mask int,m And the mask out-of-band penalty penal ext,m Normalize and add to get the distance penalty matrix penal corresponding to the mth sub-mask matrix m ;

[0072] The distance penalty matrix penal corresponding to each sub-mask matrix m The load forecast error penalty matrix Penal corresponding to the power supply matrix is ​​obtained by splicing.

[0073] It should be noted that this embodiment describes in detail how to calculate the load forecast error penalty matrix corresponding to the power supply matrix. This step is to analyze the sub-mask matrix of the power demand matrix, calculate the Euclidean distance matrix inside and outside the mask, and generate penalty items based on this, and finally integrate these penalty items into a matrix for use in the model training process.

[0074] Specifically, the sub-mask matrix of the power supply matrix refers to dividing the power supply matrix into multiple regions or subsets, each subset corresponding to a sub-mask matrix. The Euclidean distance matrix inside the mask refers to calculating the shortest distance from each point inside the sub-mask matrix to the mask boundary. These distances can be determined by computational geometry methods, such as using the Dijkstra algorithm or the Floyd-Warshall algorithm. The penalty term inside the mask is calculated based on these distances and can be defined using different mathematical functions, such as exponential functions or linear functions. The calculation of the Euclidean distance matrix outside the mask and the penalty term outside the mask is similar to that inside the mask, but the focus is on the area outside the sub-mask matrix.

[0075] Preferably, in order to more accurately calculate the penalty terms inside and outside the mask, more advanced numerical analysis methods, such as Gaussian blur or machine learning algorithms, can be used to optimize the calculation of distances and the generation of penalty terms. In addition, an adaptive mechanism can be introduced to dynamically adjust the weight of the penalty term according to the performance of the model during training to improve the generalization ability and prediction accuracy of the model. In practical applications, the real-time data and historical trends of the power system can also be considered to adjust the penalty terms inside and outside the mask in real time to adapt to changes in the operating status of the power system. These methods can further improve the accuracy and robustness of the model for power load forecasting.

[0076] In some embodiments, the intra-mask Euclidean distance matrix EDM of the m-th sub-mask matrix m The dimension of the m-th sub-mask matrix of the power demand matrix is ​​the same as that of the m-th sub-mask matrix; the intra-mask Euclidean distance matrix EDM of the m-th sub-mask matrix m The element value at position (i,j,k) is:

[0077]

[0078] Among them, dist(i,j,k) represents the shortest Euclidean distance from point (i,j,k) to the mask boundary of the sub-mask matrix, Mask m Represents the mask area of ​​the mth sub-mask matrix.

[0079] Specifically, the intra-mask Euclidean distance matrix of the sub-mask matrix is ​​a matrix with the same dimension as the sub-mask matrix of the power demand matrix, and the value of each element represents the shortest Euclidean distance from the corresponding point to the boundary of the sub-mask region. This distance can be determined by computational geometry methods, for example, the Bresenham algorithm or the Dijkstra algorithm can be used to calculate the distance from the point to the nearest boundary. In practical applications, the calculation of this distance needs to take into account the topological structure of the power network and the characteristics of power flow.

[0080] Preferably, in order to improve the efficiency and accuracy of the calculation of the Euclidean distance matrix within the mask, parallel computing technology can be used to process large-scale power demand matrices. In addition, an adaptive algorithm can be introduced to optimize the distance calculation process, for example, automatically adjusting the calculation parameters according to the dynamic changes in power demand. In some cases, machine learning methods can also be used to predict and estimate the distance within the mask, especially when the data is sparse or the computing resources are limited.

[0081] In some embodiments, according to the formula

[0082] penal int,m =(max(EDM m )-EDM m )⊙GT m

[0083] Calculate the intra-mask penalty term of the mth sub-mask matrix, where EDM m Represents the intra-mask Euclidean distance matrix of the mth sub-mask matrix, GT m represents the mth sub-mask matrix, max(·) means replacing all elements of the matrix with the maximum element value in the matrix, and ⊙ means multiplying the corresponding elements of the matrix.

[0084] It should be noted that this embodiment describes in detail how to calculate the intra-mask penalty term of the sub-mask matrix. This step is to generate penalty terms by combining the intra-mask Euclidean distance matrix with the sub-mask matrix using specific mathematical operations, which will be used to construct the subsequent load forecast error penalty matrix.

[0085] Specifically, the calculation of the intra-mask penalty term involves the intra-mask Euclidean distance matrix and the sub-mask matrix. The intra-mask Euclidean distance matrix is ​​a matrix containing the distance from each point to the nearest mask boundary, while the sub-mask matrix is ​​a binary matrix used to identify points inside and outside the mask area.

[0086] More specifically, when calculating the penalty term within the mask, we first need to determine an appropriate mathematical function to process the two matrices. For example, we can use the maximum replacement function to replace all elements in the masked Euclidean distance matrix with the maximum value, and then multiply the elements with the sub-mask matrix. Such an operation can highlight the points within the mask area and provide more accurate error penalty information for load forecasting.

[0087] Preferably, in order to improve the flexibility and adaptability of the calculation of the penalty term within the mask, a parameterized method can be introduced to adjust the calculation of the penalty term. For example, a penalty coefficient can be set to adjust the penalty intensity of the points within the mask.

[0088] In addition, different mathematical operations, such as weighted average or exponential function, can be used to enhance the model's sensitivity to different types of load forecast errors. In practical applications, these parameters can be optimized through methods such as cross-validation according to the specific needs and historical data of the power system to achieve the best forecasting performance. These improvements can make the model more accurately reflect the dynamic changes of power load, thereby improving the operating efficiency and stability of the power system.

[0089] In some embodiments, according to the formula

[0090] penal ext,m =(max(EDM ext,m )-EDM ext,m )⊙(1-GT m )

[0091] Calculate the mask-outside penalty term of the m-th sub-mask matrix, where EDM ext,m Represents the mask-outside Euclidean distance matrix of the m-th sub-mask matrix, GT m represents the mth sub-mask matrix, max(·) means replacing all elements of the matrix with the maximum element value in the matrix, and ⊙ means multiplying the corresponding elements of the matrix.

[0092] Specifically, the calculation of the off-mask penalty involves the off-mask Euclidean distance matrix and the sub-mask matrix. The off-mask Euclidean distance matrix is ​​a matrix that records the distance from each point to the nearest mask boundary, but only for those points outside the mask area. The sub-mask matrix is ​​an identification matrix used to distinguish points inside and outside the mask area.

[0093] Furthermore, when calculating the penalty term outside the mask, a specific mathematical function can be used. For example, all elements in the Euclidean distance matrix outside the mask are replaced with the maximum value through the maximum value replacement function, and then multiplied element by element with the complement of the sub-mask matrix, that is, 1 minus the sub-mask matrix. Such an operation can effectively identify and process points outside the mask area, providing the model with the necessary error penalty information.

[0094] Preferably, in order to improve the calculation efficiency and accuracy of the out-of-mask penalty term, multiple strategies can be used to optimize this process. For example, a dynamically adjusted penalty coefficient can be introduced, which can be adjusted according to the real-time operating status and historical data of the power system.

[0095] In addition, advanced numerical analysis methods, such as machine learning algorithms, can also be considered to predict and estimate the distance outside the mask, especially when data is sparse or computing resources are limited. These methods can further improve the model's response speed and prediction accuracy to power load changes, thereby optimizing the operating efficiency and stability of the power system.

[0096] In some embodiments, the outer-mask Euclidean distance matrix EDM of the m-th sub-mask matrix is ext,m The dimension of the mth sub-mask matrix of the power supply matrix is ​​the same as that of the mth sub-mask matrix of the power supply matrix; the masked Euclidean distance matrix EDM ext,m The element value at position (i,j,k) is:

[0097]

[0098] Among them, dist(i,j,k) represents the shortest Euclidean distance from point (i,j,k) to the mask boundary of the sub-mask matrix, Mask m Represents the mask area of ​​the mth sub-mask matrix.

[0099] It should be noted that this embodiment describes in detail how to calculate the mask-outside Euclidean distance matrix of the sub-mask matrix. This matrix is ​​a key component of the power supply matrix, which is used to determine the shortest distance from each point to the boundary of the sub-mask area, especially for those points outside the mask area.

[0100] Specifically, the calculation of the mask-outside Euclidean distance matrix involves analyzing each point in the power supply matrix to determine their distance to the nearest mask boundary. This usually involves complex geometric calculations that can be implemented by various algorithms, such as the Dijkstra algorithm or the A* algorithm.

[0101] Furthermore, in practical applications, the mask area can be defined by the topology of the power network, the path of power flow, or other related factors. The value of each element of the mask-outside Euclidean distance matrix is ​​determined by the shortest distance from the point to the mask boundary. If the point is not in the mask area, the value is 0.

[0102] Preferably, in order to improve the efficiency and accuracy of the calculation of the Euclidean distance matrix outside the mask, parallel computing technology can be used to process large-scale power supply matrices. In addition, adaptive algorithms can be introduced to optimize the distance calculation process, for example, automatically adjusting the calculation parameters according to the dynamic changes in power demand. In some cases, machine learning methods can also be used to predict and estimate the distance outside the mask, especially when the data is sparse or the computing resources are limited.

[0103] The above-mentioned embodiments of the present disclosure have the following beneficial effects: The virtual power plant power load switching control method of the present invention can construct a power load prediction and control model by acquiring power load data and power supply capacity data, thereby realizing accurate control and optimization of power load. This method can effectively improve the operating efficiency and stability of the power system while reducing the risks caused by load fluctuations. By subdividing and dynamically adjusting the power supply capacity, it is possible to respond more flexibly to the rapid changes in the power market and realize efficient use of power resources.

[0104] In addition, the control method can improve the prediction accuracy in the model training process by adopting a composite loss function with load prediction error penalty. This method can reduce the power load control errors caused by prediction errors, thereby improving the reliability of the power system. At the same time, by conducting in-depth analysis of the sub-supply capacity data of the power supply capacity, the power load can be adjusted more finely, and a more refined control of the power system can be achieved, further improving the economy and environmental friendliness of the power system.

[0105] Further references Figure 2 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a virtual power plant power load switching control system. These device embodiments are similar to Figure 2 Corresponding to the method embodiment shown, the model training system for real-time image recognition can be specifically applied to various electronic devices. Figure 2 As shown, a virtual power plant power load switching control system 200 of some embodiments includes:

[0106] A feature extraction module 201 is used to extract features of power load data and power supply capacity data using a deep learning network;

[0107] A prediction module 202, for predicting future power loads using a time series prediction algorithm based on the extracted features;

[0108] An optimization module 203, for generating a power load switching control strategy using an optimization algorithm based on the prediction results and the power supply capacity data;

[0109] The evaluation module 204 is used to evaluate the performance of the power load switching control strategy, including economy, reliability and environmental impact.

[0110] It is understandable that the modules recorded in the virtual power plant power load switching control system 200 are similar to those in the reference Figure 1 Therefore, the operations, features and beneficial effects described above for the virtual power plant power load switching control method are also applicable to the virtual power plant power load switching control system 200 and the modules contained therein, and will not be repeated here.

[0111] In some embodiments, generating a control strategy training sample set based on the power load data and the corresponding power supply capacity data includes:

[0112] Gaussian process regression is used to interpolate power load data and corresponding power supply capacity data to handle missing data;

[0113] Perform wavelet transform on power load data and power supply capacity data to extract time-frequency features;

[0114] For each sample data, principal component analysis is performed on the power load data and power supply capacity data to reduce the data dimension and improve the model training efficiency;

[0115] Based on the data after dimensionality reduction, a clustering algorithm is used to classify the power load types and power supply capacity types to generate training samples.

[0116] Specifically, the virtual power plant power load switching control system includes a feature extraction module, a prediction module, an optimization module and an evaluation module. The feature extraction module uses a deep learning network to analyze and extract key features of power load data and power supply capacity data. The prediction module uses a time series prediction algorithm to predict future power loads based on the extracted features. The optimization module uses an optimization algorithm to generate a power load switching control strategy based on the prediction results and power supply capacity data. The evaluation module is responsible for evaluating the performance of the control strategy, including economy, reliability and environmental impact.

[0117] Preferably, in order to improve the accuracy and efficiency of the system, advanced deep learning technologies such as convolutional neural network CNN or recurrent neural network RNN ​​can be used for the feature extraction module to better capture the spatiotemporal characteristics of the data. In the prediction module, algorithms such as long short-term memory network LSTM or gated recurrent unit GRU can be used to improve the accuracy of time series prediction. The optimization module can use genetic algorithms, particle swarm optimization PSO or simulated annealing algorithms to find the best power load switching strategy. The evaluation module can introduce methods such as multi-objective optimization and life cycle assessment LCA to comprehensively evaluate the comprehensive performance of the control strategy. In addition, the system can integrate real-time data monitoring and adaptive learning mechanisms to dynamically adjust the control strategy to adapt to the continuous changes in the power system.

[0118] Reference below Figure 3 , which shows a schematic diagram of the structure of an electronic device 300 suitable for implementing some embodiments of the present disclosure. The electronic devices in some embodiments of the present disclosure may include but are not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle terminals (such as vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The terminal device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0119] like Figure 3 As shown, the electronic device 300 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 are also stored. The processing device 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0120] Typically, the following devices may be connected to the I / O interface 305: input devices 306 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 308 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 309. The communication devices 309 may allow the electronic device 300 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 3The electronic device 300 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0121] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from the network through the communication device 309, or installed from the storage device 308, or installed from the ROM 302. When the computer program is executed by the processing device 301, the above-mentioned functions defined in the method of some embodiments of the present disclosure are executed.

[0122] The storage medium of the embodiment of the present application stores program instructions that can implement all the above methods, wherein the program instructions can be stored in the above storage medium in the form of a software product, including several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0123] It should be noted that in some embodiments of the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In some embodiments of the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in combination with an instruction execution system, device or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0124] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0125] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0126] The computer-readable medium may be included in the electronic device; or it may exist separately without being assembled into the electronic device. The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device can be written in one or more programming languages ​​or a combination thereof to perform computer program codes for the operations of some embodiments of the present disclosure. The programming languages ​​include object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer, partially on the remote computer, or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect through the Internet).

[0127] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0128] The units described in some embodiments of the present disclosure may be implemented in software or in hardware. The units described may also be provided in a processor, and the functions described above herein may be performed at least in part by one or more hardware logic components. For example, and without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0129] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) and the technical solutions formed.

Claims

1. A virtual power plant power load switching control method, characterized in that: The following steps are involved: Acquire power load data and corresponding power supply capacity data of the virtual power plant, and generate a control strategy training sample set based on the power load data and the corresponding power supply capacity data; Analyze each corresponding power supply capacity data in the control strategy training sample set based on the stability of the power system to obtain multiple sub-supply capacity data of each corresponding power supply capacity data; Constructing a power load prediction and control model, and training the power load prediction and control model based on the power load data, power supply capacity data, and multiple sub-supply capacity data of each corresponding power supply capacity data in the training sample set to obtain a power load switching control model; The power load data of the virtual power plant to be controlled is input into the power load switching control model to obtain the power load switching strategy of the virtual power plant to be controlled.

2. The virtual power plant power load switching control method according to claim 1, characterized in that: Based on the stability of the power system, each power supply capacity data in the training sample set is analyzed to obtain multiple sub-supply capacity data of each power supply capacity, including: The initial state of each power supply capacity is taken as the first type, and the stability of the initial state is calculated; each state of the power supply capacity is traversed in turn, and the stability of the current state is calculated. If the stability of the current state is the same as the stability of the previous state, the current state and the previous state are of the same type, otherwise the current state and the previous state are of different types; A plurality of sub-supply capability data of the power supply capability are obtained according to the type of each state in the power supply capability.

3. The virtual power plant power load switching control method according to claim 1, characterized in that: According to multiple sub-supply capacity data of power supply capacity, a load forecast error penalty matrix with load forecast error penalty is adopted to calculate the training loss of the power load forecasting and control model.

4. The virtual power plant power load switching control method according to claim 3 is characterized in that: The load forecast error penalty matrix corresponding to the power supply matrix is ​​calculated by the following steps: Calculate the intra-mask Euclidean distance matrix EDM of the mth sub-mask matrix based on the mth sub-mask matrix of the power demand matrix m ; Based on the Euclidean distance matrix EDM within the mask m Calculate the intra-mask penalty term penal of the mth sub-mask matrix int,m ; Calculate the masked Euclidean distance matrix EDM of the mth sub-mask matrix based on the mth sub-mask matrix of the power supply matrix ext,m ; Based on the masked Euclidean distance matrix EDM ext,m Calculate the out-of-mask penalty term penal of the mth sub-mask matrix ext,m ; The penalty term penal in the mask int,m And the mask out-of-band penalty penal ext,m Normalize and add to get the distance penalty matrix penal corresponding to the mth sub-mask matrix m ; The distance penalty matrix penal corresponding to each sub-mask matrix m The load forecast error penalty matrix Penal corresponding to the power supply matrix is ​​obtained by splicing.

5. The virtual power plant power load switching control method according to claim 4, characterized in that: The intra-mask Euclidean distance matrix EDM of the m-th sub-mask matrix m The dimension is the same as the mth sub-mask matrix of the power demand matrix.

6. The virtual power plant power load switching control method according to claim 4, characterized in that: Compute the intra-mask penalty term for the mth sub-mask matrix.

7. The method for controlling power load switching of a virtual power plant according to claim 4, characterized in that: Calculate the out-of-mask penalty term for the m-th sub-mask matrix.

8. The method for controlling power load switching of a virtual power plant according to claim 4, characterized in that: The masked Euclidean distance matrix EDM of the m-th sub-mask matrix ext,m The same dimension as the mth sub-mask matrix of the power supply matrix.

9. A virtual power plant power load switching control system, characterized in that: The system comprises: A feature extraction module, for extracting features of power load data and power supply capacity data using a deep learning network; A prediction module for predicting future power loads using a time series prediction algorithm based on the extracted features; An optimization module, for generating a power load switching control strategy using an optimization algorithm based on the prediction results and power supply capacity data; Evaluation module for evaluating the performance of power load switching control strategies, including economics, reliability, and environmental impact.

10. The virtual power plant power load switching control system according to claim 9, characterized in that: Generating a control strategy training sample set based on the power load data and the corresponding power supply capacity data includes: Gaussian process regression is used to interpolate power load data and corresponding power supply capacity data to handle missing data; Perform wavelet transform on power load data and power supply capacity data to extract time-frequency features; For each sample data, principal component analysis is performed on the power load data and power supply capacity data to reduce the data dimension and improve the model training efficiency; Based on the data after dimensionality reduction, a clustering algorithm is used to classify the power load types and power supply capacity types to generate training samples.

Images