Energy regulation and control method and system of virtual power plant based on prediction model

Abstract

The invention discloses an energy regulation and control method and system of a virtual power plant based on a prediction model, and relates to the field of virtual power plant operation control, and the method comprises the following steps: collecting energy data in the virtual power plant; integrating the wind power data and the load data, and performing feature extraction on the integrated data; based on the predicted power generation data of the distributed energy and the predicted demand data of the load side resources, constructing a preliminary energy dynamic balance prediction model; performing day-ahead load demand and distributed energy prediction by using the preliminary energy dynamic balance prediction model; according to the preliminary scheduling strategy, using a niche grey wolf optimization algorithm to adjust the operation state of each distributed energy and load side resource in real time; dynamically adjusting and optimizing the energy dynamic balance prediction model; and carrying out load demand and distributed energy prediction again by using the optimized energy dynamic balance prediction model. According to the method, the change of the power system can be responded in advance, and more efficient energy scheduling is realized.

Description

Technical Field

[0001] The present invention relates to the field of virtual power plant operation control, and in particular, to an energy control method and system for a virtual power plant based on a prediction model. Background Art

[0002] With the continuous growth of global energy demand, the limitation of fossil fuel reserves and the negative impact of fossil fuel use on the environment, many countries have adopted strategies to develop renewable energy and increase distributed power generation to solve this energy crisis, such as wind turbines, solar photovoltaic panels, etc. This has led to a high penetration rate of renewable energy in the existing power system. Large-scale and high-proportion access to the power grid will bring a series of challenges in energy control and system operation. At the same time, renewable energy usually has the characteristics of nature, intermittent, randomness and volatility, and it is difficult to directly arrange its production plan, which will have an adverse impact on the operation of the power grid. With the rapid development of smart grid technology, the global academic and industrial circles have gradually proposed the concept of virtual power plants.

[0003] Virtual Power Plant (VPP) is a new type of power grid operation mode. Through advanced information technology, it centrally dispatches and operates a variety of distributed energy sources, load-side resources (such as electric vehicles, energy storage facilities, cogeneration facilities, etc.) and renewable energy (such as wind energy, solar energy, etc.) scattered in the power grid, making it behave like a unified and controllable power plant in the power market.

[0004] The main functions of a virtual power plant include two parts: one is the ability to predict load and renewable energy power generation; the other is the ability to optimize scheduling in real time, so that the operating status of distributed energy and load-side resources is always in the optimal state, thereby improving the economy and reliability of power grid operation.

[0005] Existing power plants have the following major energy problems: wind power and load data are generally random, uncertain and volatile, which brings difficulties to the energy supply planning and scheduling of power plants; if wind power and load data cannot be accurately predicted, it is difficult to make reasonable production plans, which may lead to an imbalance in energy supply and demand in the power grid. Power plants usually adopt a centralized system structure and control method, which makes it difficult for power plants to fully utilize distributed energy and energy storage resources to achieve more flexible and economical operations.

[0006] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0007] In order to overcome the above problems, the present invention aims to propose an energy control method and system for a virtual power plant based on a prediction model, with the aim of solving the problem that it is difficult to make a reasonable production plan if the wind power and load data cannot be accurately predicted.

[0008] To this end, the specific technical solution adopted by the present invention is as follows:

[0009] According to one aspect of the present invention, a method for energy regulation of a virtual power plant based on a prediction model is provided, and the method for energy regulation of a virtual power plant comprises the following steps:

[0010] S1. Collect energy data in the virtual power plant and extract wind power data and load data from the energy data;

[0011] S2. Integrate wind power data and load data using a linear regression model, and extract features from the integrated data to obtain predicted power generation data of distributed energy and predicted demand data of load-side resources;

[0012] S3. Based on the predicted power generation data of the distributed energy and the predicted demand data of the load-side resources, a preliminary energy dynamic balance prediction model is constructed using a recurrent neural network;

[0013] S4. Use the preliminary energy dynamic balance forecasting model to forecast the load demand and distributed energy resources in the day ahead, and formulate a preliminary dispatching strategy based on the forecasting results;

[0014] S5. According to the preliminary dispatching strategy, the operating status of each distributed energy source and load-side resource is adjusted in real time using the niche grey wolf optimization algorithm;

[0015] S6. Dynamically adjust and optimize the energy dynamic balance prediction model based on real-time operation data;

[0016] S7. Use the optimized energy dynamic balance prediction model to re-predict load demand and distributed energy, and optimize the scheduling strategy based on the new prediction results;

[0017] S8. Execute steps S5-S7 in a loop until optimal energy regulation is achieved.

[0018] Optionally, the step of integrating wind power data and load data using a linear regression model and performing feature extraction on the integrated data to obtain predicted power generation data of distributed energy and predicted demand data of load-side resources includes the following steps:

[0019] S21. Acquire wind power data of the virtual power plant, wherein the wind power data at least includes wind speed, wind direction, temperature, humidity and atmospheric pressure;

[0020] S22, obtaining load data of the virtual power plant, wherein the load data at least includes load demand, equipment operation status and environmental parameters;

[0021] S23, performing linear regression integration on the wind speed, wind direction, temperature, humidity and atmospheric pressure of the wind power data and the load demand, equipment operation status and environmental parameters of the load data to obtain a comprehensive data set including both wind power data and load data;

[0022] S24, extracting features of the data in the comprehensive data set, wherein the features at least include statistical features, frequency domain features, and time domain features;

[0023] S25. Based on the extracted features, a linear regression model is used to obtain the predicted power generation data of distributed energy and the predicted demand data of load-side resources.

[0024] Optionally, the method of obtaining the predicted power generation data of distributed energy and the predicted demand data of load-side resources using a linear regression model based on the extracted features includes the following steps:

[0025] S251. According to the characteristics of the data in the comprehensive data set, combined with the historical power generation data and load demand data, a linear regression model is constructed;

[0026] S252, dividing the comprehensive data set into a training set and a test set, wherein the training set is used to train the linear regression model, and the test set is used to evaluate the prediction ability of the linear regression model;

[0027] S253, training a linear regression model using a training set to learn the relationship between data features, power generation data, and load demand data;

[0028] S254, use the trained linear regression model to predict the test set and evaluate the prediction performance of the model;

[0029] S255. If the prediction performance of the linear regression model meets the requirements, it is used to predict future distributed energy generation data and load-side resource demand data;

[0030] If the prediction performance of the linear regression model does not meet the requirements, reselect the characteristics of the data in the comprehensive dataset and adjust the linear regression model parameters.

[0031] Optionally, the method of using a preliminary energy dynamic balance prediction model to predict the load demand and distributed energy resources in the previous day, and formulating a preliminary scheduling strategy according to the prediction results includes the following steps:

[0032] S41. Use the preliminary energy dynamic balance prediction model to input the statistical characteristics, frequency domain characteristics, and time domain characteristics of the load data to perform a load demand forecast for the day ahead;

[0033] S42, using a preliminary energy dynamic balance prediction model, inputting statistical characteristics, frequency domain characteristics, and time domain characteristics of wind power data to perform a day-ahead distributed energy generation prediction;

[0034] S43. Formulate a preliminary dispatching strategy based on the forecast results of the load demand forecast and the distributed energy generation forecast.

[0035] Optionally, the method of adjusting the operating status of each distributed energy source and load-side resource in real time using the niche grey wolf optimization algorithm according to the preliminary scheduling strategy includes the following steps:

[0036] S51, according to the preliminary dispatching strategy, determine the initial operating state of each distributed energy source and load-side resource, and regard the initial operating state as a gray wolf individual in the microhabitat gray wolf optimization algorithm;

[0037] S52, setting a fitness function, evaluating the quality of each individual gray wolf, and calculating the fitness value of each individual gray wolf;

[0038] S53, sorting all gray wolf individuals according to the fitness value, selecting the best three gray wolf individuals, and recording them as α gray wolf, β gray wolf and δ gray wolf, wherein the running state corresponding to the best fitness value is recorded as the current optimal solution;

[0039] S54, calculating an update strategy according to the operating states and corresponding fitness values ​​of the α gray wolf, the β gray wolf, and the δ gray wolf, and updating the operating states of the remaining gray wolf individuals;

[0040] S55, according to the update strategy, update the operating status of the remaining gray wolf individuals, obtain a new operating status, and re-evaluate the fitness values ​​of the remaining gray wolf individuals to find a new optimal solution;

[0041] S56, determining whether the maximum number of iterations has been reached;

[0042] If so, output the optimal solution;

[0043] If not, execute step S52;

[0044] S57. According to the optimal solution output by the microhabitat grey wolf optimization algorithm, the operating status of each distributed energy source and load-side resource is adjusted in real time.

[0045] Optionally, setting a fitness function, evaluating the quality of each individual gray wolf, and calculating the fitness value of each individual gray wolf comprises the following steps:

[0046] S521. Define the fitness function of the niche grey wolf optimization algorithm according to the scheduling problem of distributed energy and load-side resources, and ensure the computability and interpretability of the fitness function;

[0047] S522, determining each individual gray wolf, obtaining the running status data of each individual gray wolf, and ensuring the accuracy and timeliness of the data;

[0048] S523, calculating and recording the fitness value of each individual gray wolf through the fitness function, and setting a mechanism for automatically recalculating the fitness value to update the running status data of each individual gray wolf in real time;

[0049] S524. Regularly check the validity of the fitness function and adjust the fitness function as needed.

[0050] Optionally, the step of calculating an update strategy based on the operating states and corresponding fitness values ​​of the α gray wolf, the β gray wolf, and the δ gray wolf, and updating the operating states of the remaining gray wolf individuals comprises the following steps:

[0051] S541, setting the moving step length parameter A and the distance adjustment parameter C of the niche gray wolf optimization algorithm;

[0052] S542, calculating the new position of each non-optimal gray wolf according to the positions of the α gray wolf, the β gray wolf and the δ gray wolf and the moving step length parameter A and the distance adjustment parameter C;

[0053] S543, performing range detection on the newly calculated non-optimal gray wolf position to determine whether it exceeds the preset solution space range;

[0054] If it exceeds the solution space, the parameters need to be adjusted to return it to the solution space;

[0055] S544, substituting the adjusted new position into the expression of the running state to obtain a new running state, and updating the running state of the non-optimal gray wolf.

[0056] Optionally, calculating the new position of each non-optimal gray wolf includes:

[0057] Calculate the distance D between the non-optimal gray wolf and the optimal gray wolf α respectively α , the distance Dβ between the non-optimal gray wolf and the optimal gray wolf β and the distance D between the non-optimal gray wolf and the optimal gray wolf δ δ ;

[0058] According to the distance D α , distance D β and distance D δ , calculate the new position of the non-optimal gray wolf after it moves according to the behavior of the optimal gray wolf α, the optimal gray wolf δ, and the optimal gray wolf δ;

[0059] Wherein, the distance D α The calculation formula is as follows:

[0060] D α =|C1·α-A1·X|

[0061] The distance D β The calculation formula is as follows:

[0062] D β =|C2·β-A2·X|

[0063] The distance D δ The calculation formula is as follows:

[0064] D δ =|C3·δ-A3·X|

[0065] Where D α Represents the distance between the non-optimal gray wolf and the optimal gray wolf α;

[0066] D β represents the distance between the optimal gray wolf and the optimal gray wolf β;

[0067] D δ represents the distance between the optimal gray wolf and the optimal gray wolf δ;

[0068] C1, C2, and C3 represent randomly generated distance adjustment parameters;

[0069] A1, A2 and A3 represent the calculated moving step parameters respectively;

[0070] X represents the current position of the non-optimal gray wolf.

[0071] Optionally, the re-forecasting of load demand and distributed energy using the optimized energy dynamic balance forecasting model and optimizing the dispatching strategy according to the new forecasting results includes the following steps:

[0072] S71. Use the optimized energy dynamic balance prediction model, input historical data and current state data of load data, and predict future load demand;

[0073] S72, using the optimized energy dynamic balance prediction model, inputting historical data and current status data of wind power data, and predicting the future distributed energy supply;

[0074] S73. Compare the predicted load demand and distributed energy supply to find out the time periods of energy shortage or surplus;

[0075] S74. If there is a period of insufficient energy, the output of distributed energy is increased, energy storage equipment is started to supplement the supply, and load-side management is performed to reset the operating status of each distributed energy source;

[0076] S75. If there is a period of excess energy, the output of distributed energy is reduced, excess energy is transmitted to the power grid, and the excess energy is stored, and the operating state of each load-side resource is reset;

[0077] S76: Evaluate the effect of the new scheduling strategy and compare the key energy indicators before and after the scheduling;

[0078] If there is a significant difference between the evaluation results and the actual results of the dispatch strategy, the energy dynamic balance prediction model will be adjusted.

[0079] According to another aspect of the present invention, there is also provided an energy control system for a virtual power plant based on a prediction model, the system comprising: a virtual power plant data collection and extraction module, a data integration and feature extraction module, a preliminary energy dynamic balance prediction model construction module, a prediction and scheduling strategy formulation module, a real-time scheduling optimization module, an energy dynamic balance prediction model optimization module, an optimization prediction and scheduling strategy formulation module and a feedback module;

[0080] The virtual power plant data collection and extraction module is used to collect energy data in the virtual power plant and extract wind power data and load data from the energy data;

[0081] The data integration and feature extraction module is used to integrate wind power data and load data using a linear regression model, and perform feature extraction on the integrated data to obtain predicted power generation data of distributed energy and predicted demand data of load-side resources;

[0082] The preliminary energy dynamic balance prediction model construction module is used to construct a preliminary energy dynamic balance prediction model using a recurrent neural network based on the predicted power generation data of the distributed energy and the predicted demand data of the load-side resources;

[0083] The forecasting and scheduling strategy formulation module is used to use the preliminary energy dynamic balance forecasting model to forecast the load demand and distributed energy in the day ahead, and formulate a preliminary scheduling strategy based on the forecast results;

[0084] The real-time scheduling optimization module is used to adjust the operating status of each distributed energy source and load-side resource in real time according to the preliminary scheduling strategy using the niche grey wolf optimization algorithm;

[0085] The energy dynamic balance prediction model optimization module is used to dynamically adjust and optimize the energy dynamic balance prediction model according to real-time operation data;

[0086] The optimization prediction and dispatch strategy formulation module is used to re-predict load demand and distributed energy using the optimized energy dynamic balance prediction model, and optimize the dispatch strategy according to the new prediction results;

[0087] The feedback module is used to cyclically execute the real-time scheduling optimization module, the energy dynamic balance prediction model optimization module and the optimization prediction and scheduling strategy formulation module until the optimal energy regulation is achieved.

[0088] Compared with the prior art, this application has the following beneficial effects:

[0089] 1. The present invention adopts an energy dynamic balance prediction model to predict future load demand and distributed energy supply, and then formulates a scheduling strategy based on the prediction results, which can respond to changes in the power system in advance and achieve more efficient energy scheduling.

[0090] 2. The present invention obtains wind power data and load data, and then integrates them using a linear regression model to obtain a comprehensive data set containing richer information, which can help us understand the operating status and environmental changes of the power system more comprehensively; by extracting statistical features, frequency domain features and time domain features, data can be analyzed from different angles to better understand the inherent laws of the data; the linear regression model is a simple and effective prediction method that can perform predictions quickly, and the model has strong interpretability, which makes it easy to understand the prediction results of the model.

[0091] 3. The present invention uses the microhabitat gray wolf optimization algorithm to adjust the operating status of distributed energy and load-side resources in real time to achieve the optimal operating effect, and can quickly respond to system changes and improve the operating efficiency of the system. The setting of the fitness function enables the gray wolf individual to perform adaptive adjustments according to the actual state of the system. This method can ensure that the system always operates in the optimal state and improve the stability of the system; the microhabitat gray wolf optimization algorithm is a global search algorithm that can avoid falling into a local optimal solution. This method can ensure that the solution found is the global optimal solution, thereby achieving the optimal scheduling strategy.

[0092] 4. The present invention can avoid infinite loops of the algorithm and improve the operating efficiency of the algorithm by setting the maximum number of iterations. This method can ensure the efficient operation of the algorithm while satisfying the optimization effect. In the process of setting the fitness function and the update strategy, parameters can be flexibly adjusted according to actual needs, such as moving step parameters and distance adjustment parameters, so as to better adapt to various operating environments and conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] The above characteristics, features and advantages of the present invention and their implementation methods and methods will become more clearly understood in conjunction with the following description of the embodiments, which are described in detail in conjunction with the accompanying drawings. Herein, a schematic diagram is shown:

[0094] Figure 1 It is a flow chart of an energy control method of a virtual power plant based on a prediction model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0095] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0096] According to an embodiment of the present invention, a method and system for energy regulation of a virtual power plant based on a prediction model are provided.

[0097] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to an embodiment of the present invention, a method for energy regulation of a virtual power plant based on a prediction model is provided, and the method for energy regulation of a virtual power plant comprises the following steps:

[0098] S1. Collect energy data in the virtual power plant and extract wind power data and load data from the energy data.

[0099] It should be explained that when collecting energy data in virtual power plants, the main focus is on the power generation of various distributed energy sources, including but not limited to wind power, photovoltaic power, hydropower, etc., and the load data of the power plant should also be collected. These data can be obtained in real time through various sensors and monitoring equipment, and transmitted to the data center for processing and analysis through advanced communication technology.

[0100] Wind power data mainly includes parameters such as wind speed, wind direction, speed and power of wind turbines. Wind speed and wind direction are important factors affecting the power generation of wind turbines. Through real-time monitoring and analysis of these data, wind power generation can be predicted and the operation strategy of the power plant can be adjusted accordingly.

[0101] Load data mainly includes parameters such as the real-time load of the power plant and the demand load of the power grid. The real-time load of the power plant indicates the current power supply capacity of the power plant, while the demand load of the power grid indicates the current power demand of the power grid. Through real-time monitoring and analysis of these data, the operating status of the power plant can be evaluated, the power demand of the power grid can be predicted, and the power supply strategy of the power plant can be adjusted accordingly. By extracting and analyzing the energy data in the virtual power plant, the operating status of the power plant can be understood, the power demand of the power grid can be predicted, the operating strategy of the power plant can be optimized, and a more efficient and economical energy supply can be achieved.

[0102] S2. Use a linear regression model to integrate wind power data and load data, and perform feature extraction on the integrated data to obtain predicted power generation data of distributed energy and predicted demand data of load-side resources.

[0103] Preferably, the step of integrating wind power data and load data using a linear regression model and extracting features from the integrated data to obtain predicted power generation data of distributed energy and predicted demand data of load-side resources includes the following steps:

[0104] S21. Acquire wind power data of the virtual power plant, wherein the wind power data at least includes wind speed, wind direction, temperature, humidity and atmospheric pressure;

[0105] S22, obtaining load data of the virtual power plant, wherein the load data at least includes load demand, equipment operation status and environmental parameters;

[0106] S23, performing linear regression integration on the wind speed, wind direction, temperature, humidity and atmospheric pressure of the wind power data and the load demand, equipment operation status and environmental parameters of the load data to obtain a comprehensive data set including both wind power data and load data;

[0107] S24, extracting features of the data in the comprehensive data set, wherein the features at least include statistical features, frequency domain features, and time domain features;

[0108] S25. Based on the extracted features, a linear regression model is used to obtain the predicted power generation data of distributed energy and the predicted demand data of load-side resources.

[0109] Preferably, the method of obtaining the predicted power generation data of distributed energy and the predicted demand data of load-side resources based on the extracted features using a linear regression model comprises the following steps:

[0110] S251. According to the characteristics of the data in the comprehensive data set, combined with the historical power generation data and load demand data, a linear regression model is constructed;

[0111] S252, dividing the comprehensive data set into a training set and a test set, wherein the training set is used to train the linear regression model, and the test set is used to evaluate the prediction ability of the linear regression model;

[0112] S253, training a linear regression model using a training set to learn the relationship between data features, power generation data, and load demand data;

[0113] S254, use the trained linear regression model to predict the test set and evaluate the prediction performance of the model;

[0114] S255. If the prediction performance of the linear regression model meets the requirements, it is used to predict future distributed energy generation data and load-side resource demand data;

[0115] If the prediction performance of the linear regression model does not meet the requirements, reselect the characteristics of the data in the comprehensive dataset and adjust the linear regression model parameters.

[0116] It should be explained that in step S25, the extracted features are used to train the linear regression model, including statistical features, frequency domain features and time domain features. Statistical features can help understand the basic distribution characteristics of the data, such as mean, variance, etc.; frequency domain features can reveal the periodicity and frequency of the data; and time domain features can help understand the trend of data changes over time. In addition, in order to improve the predictive performance of the model, the data set is usually divided into a training set and a test set. The training set is used to train the model, and the test set is used to evaluate the predictive performance of the model. During the model training process, care should be taken to prevent overfitting, that is, the model performs well on the training set but performs poorly on the test set. If the predictive performance of the linear regression model does not meet the requirements, it may be necessary to reselect the features or adjust the parameters of the linear regression model. This needs to be judged based on the specific situation, such as the characteristics of the data, the complexity of the problem, etc.

[0117] S3. Based on the predicted power generation data of the distributed energy and the predicted demand data of the load-side resources, a preliminary energy dynamic balance prediction model is constructed using a recurrent neural network.

[0118] It should be explained that in this process, RNN is used to build a preliminary energy dynamic balance prediction model. The model needs to predict the power generation of distributed energy and the demand for load-side resources. These data are all time series data, so RNN is a suitable choice. Building an RNN model usually requires the following steps: Define the structure of the RNN, including the input layer, hidden layer, and output layer. The hidden layer contains recurrent connections to store historical information. Select activation functions, such as tanh, ReLU, etc., for the hidden layer and output layer. Select optimization algorithms, such as SGD, Adam, etc., for training the model. Define loss functions, such as MSE, cross entropy, etc., to measure the prediction performance of the model. Compile the model and set relevant parameters, such as learning rate, number of iterations, etc. Train the model, train the model with training data, and continuously update the parameters of the model to minimize the loss function. Evaluate the model, and evaluate the prediction performance of the model with test data. If the performance is not satisfactory, you need to redefine the model structure or adjust the parameters. Use the model for prediction, input new time series data, and get the prediction results of the model.

[0119] S4. Use the preliminary energy dynamic balance forecasting model to forecast the load demand and distributed energy in the day ahead, and formulate a preliminary dispatching strategy based on the forecast results.

[0120] Preferably, the method of using a preliminary energy dynamic balance prediction model to predict the load demand and distributed energy resources in the day ahead, and formulating a preliminary scheduling strategy according to the prediction results includes the following steps:

[0121] S41. Use the preliminary energy dynamic balance prediction model to input the statistical characteristics, frequency domain characteristics, and time domain characteristics of the load data to perform a load demand forecast for the day ahead;

[0122] S42, using a preliminary energy dynamic balance prediction model, inputting statistical characteristics, frequency domain characteristics, and time domain characteristics of wind power data to perform a day-ahead distributed energy generation prediction;

[0123] S43. Formulate a preliminary dispatching strategy based on the forecast results of the load demand forecast and the distributed energy generation forecast.

[0124] It needs to be explained that this process involves the day-ahead dispatch of the energy system, that is, to formulate a reasonable power generation dispatch strategy based on the predicted load demand and distributed energy generation data. This is an important link in the operation of the power system. Its goal is to reasonably allocate power resources to meet the power needs of users under the premise of meeting the safety and economy of the power system operation. First, the energy dynamic balance prediction model is used to predict the day-ahead load demand and distributed energy generation. This process requires the input of statistical characteristics, frequency domain characteristics and time domain characteristics of load data and wind power data. A preliminary dispatch strategy is formulated based on the prediction results. The power system dispatch strategy generally needs to consider various factors, including the operating status of the power grid, the output of the power plant, the power market price, the power generation forecast results of renewable energy, etc. The goal of formulating the dispatch strategy is to ensure the stable operation of the power system while maximizing economic benefits.

[0125] S5. According to the preliminary dispatch strategy, the microhabitat grey wolf optimization algorithm is used to adjust the operating status of each distributed energy source and load-side resource in real time.

[0126] Preferably, the method of adjusting the operating status of each distributed energy source and load-side resource in real time using the niche grey wolf optimization algorithm according to the preliminary scheduling strategy comprises the following steps:

[0127] S51, according to the preliminary dispatching strategy, determine the initial operating state of each distributed energy source and load-side resource, and regard the initial operating state as a gray wolf individual in the microhabitat gray wolf optimization algorithm;

[0128] S52, setting a fitness function, evaluating the pros and cons of each gray wolf individual (i.e., the operating status of distributed energy and load-side resources), and calculating the fitness value of each gray wolf individual;

[0129] S53, sorting all gray wolf individuals according to the fitness value, selecting the best three gray wolf individuals, and recording them as α gray wolf, β gray wolf and δ gray wolf, wherein the running state corresponding to the best fitness value is recorded as the current optimal solution;

[0130] S54, according to the running states of the α gray wolf, the β gray wolf and the δ gray wolf (i.e., their positions in the solution space) and the corresponding fitness values, calculate an update strategy, and update the running states of the remaining gray wolf individuals;

[0131] S55, according to the update strategy, update the operating status of the remaining gray wolf individuals, obtain a new operating status, and re-evaluate the fitness values ​​of the remaining gray wolf individuals to find a new optimal solution;

[0132] S56, determining whether the maximum number of iterations has been reached;

[0133] If so, output the optimal solution;

[0134] If not, execute step S52;

[0135] S57. According to the optimal solution output by the microhabitat grey wolf optimization algorithm, the operating status of each distributed energy source and load-side resource is adjusted in real time.

[0136] Preferably, the setting of the fitness function, evaluating the quality of each individual gray wolf, and calculating the fitness value of each individual gray wolf comprises the following steps:

[0137] S521. Define the fitness function of the niche grey wolf optimization algorithm according to the scheduling problem of distributed energy and load-side resources, and ensure the computability and interpretability of the fitness function;

[0138] S522, determining each individual gray wolf, obtaining the running status data of each individual gray wolf, and ensuring the accuracy and timeliness of the data;

[0139] S523, calculating and recording the fitness value of each individual gray wolf through the fitness function, and setting a mechanism for automatically recalculating the fitness value to update the running status data of each individual gray wolf in real time;

[0140] S524. Regularly check the validity of the fitness function and adjust the fitness function as needed.

[0141] Preferably, the step of calculating the update strategy according to the running status and corresponding fitness values ​​of the α gray wolf, the β gray wolf and the δ gray wolf, and updating the running status of the remaining gray wolf individuals comprises the following steps:

[0142] S541, setting the moving step length parameter A and the distance adjustment parameter C of the niche gray wolf optimization algorithm;

[0143] S542, calculating the new position of each non-optimal gray wolf according to the positions of the α gray wolf, the β gray wolf and the δ gray wolf and the moving step length parameter A and the distance adjustment parameter C;

[0144] S543, performing range detection on the newly calculated non-optimal gray wolf position to determine whether it exceeds the preset solution space range;

[0145] If it exceeds the solution space, the parameters need to be adjusted to return it to the solution space;

[0146] S544, substituting the adjusted new position into the expression of the running state to obtain a new running state, and updating the running state of the non-optimal gray wolf.

[0147] Preferably, calculating the new position of each non-optimal gray wolf comprises:

[0148] Calculate the distance D between the non-optimal gray wolf and the optimal gray wolf α respectively α , the distance Dβ between the non-optimal gray wolf and the optimal gray wolf β and the distance D between the non-optimal gray wolf and the optimal gray wolf δ δ ;

[0149] According to the distance D α , distance D β and distance D δ , calculate the new position of the non-optimal gray wolf after it moves according to the behavior of the optimal gray wolf α, the optimal gray wolf δ, and the optimal gray wolf δ;

[0150] Wherein, the distance D α The calculation formula is as follows:

[0151] D α =|C1·α-A1·X|

[0152] The distance D β The calculation formula is as follows:

[0153] D β =|C2·β-A2·X|

[0154] The distance D δ The calculation formula is as follows:

[0155] D δ =|C3·δ-A3·X|

[0156] Where D α Represents the distance between the non-optimal gray wolf and the optimal gray wolf α;

[0157] D β represents the distance between the optimal gray wolf and the optimal gray wolf β;

[0158] D δrepresents the distance between the optimal gray wolf and the optimal gray wolf δ;

[0159] C1, C2, and C3 represent randomly generated distance adjustment parameters;

[0160] A1, A2 and A3 represent the calculated moving step parameters respectively;

[0161] X represents the current position of the non-optimal gray wolf.

[0162] The calculation formula for calculating the new position of the non-optimal gray wolf after moving according to the behavior of the optimal gray wolf α, the optimal gray wolf δ and the optimal gray wolf δ is as follows:

[0163] X new =α-A1*D α =β-A2*D β =δ-A3*D δ

[0164] Where, X new Indicates the new position after moving;

[0165] A1, A2 and A3 represent the calculated moving step parameters respectively;

[0166] D α Represents the distance between the non-optimal gray wolf and the optimal gray wolf α;

[0167] D β represents the distance between the optimal gray wolf and the optimal gray wolf β;

[0168] D δ Represents the distance between the optimal gray wolf and the optimal gray wolf δ.

[0169] It needs to be explained that, first of all, all gray wolf individuals need to be evaluated, which is usually done through a function called fitness function. This function gives a value based on the current state of each gray wolf individual, which is its fitness value. The higher the fitness value, the better the gray wolf individual. Next, sort the fitness values ​​of all gray wolf individuals. This shows which gray wolf individuals are better and which are relatively weaker. Then, select the three with the highest fitness values ​​from all gray wolf individuals, which are α, β and δ. These three gray wolf individuals can be regarded as leaders in the current optimization process, and other gray wolf individuals will try their best to learn or imitate their behavior. Finally, the running state corresponding to the optimal fitness value is the current optimal solution. In other words, the current state is considered to be the best of all possible states. This optimal solution will be recorded as the current best result.

[0170] S6. Dynamically adjust and optimize the energy dynamic balance prediction model based on real-time operating data.

[0171] It should be explained that in actual operation, the prediction performance of the prediction model may change due to changes in the operating status of the power system, environmental conditions, equipment performance and other factors. Therefore, it is necessary to dynamically adjust and optimize the energy dynamic balance prediction model based on real-time operation data to ensure its prediction performance. This is a common method for dynamically adjusting and optimizing the model. When new data is received, the parameters of the model can be updated with these data. For example, an online learning method can be used to train the model once each time new data is received. Alternatively, a certain amount of new data can be collected regularly and then used to retrain the model. Over time, some features may lose their predictive power, and new features may appear. Therefore, feature selection methods can be used to regularly update the features of the model. This can be achieved by evaluating the impact of each feature on the prediction performance. If there are multiple prediction models, model fusion methods can be used to improve the prediction performance.

[0172] S7. Use the optimized energy dynamic balance prediction model to re-predict load demand and distributed energy, and optimize the scheduling strategy based on the new prediction results.

[0173] Preferably, the method of re-forecasting load demand and distributed energy using the optimized energy dynamic balance forecasting model and optimizing the dispatching strategy according to the new forecasting results includes the following steps:

[0174] S71. Use the optimized energy dynamic balance prediction model, input historical data and current state data of load data, and predict future load demand;

[0175] S72, using the optimized energy dynamic balance prediction model, inputting historical data and current status data of wind power data, and predicting the future distributed energy supply;

[0176] S73. Compare the predicted load demand and distributed energy supply to find out the time periods of energy shortage or surplus;

[0177] S74. If there is a period of insufficient energy, the output of distributed energy is increased, energy storage equipment is started to supplement the supply, and load-side management is performed to reset the operating status of each distributed energy source;

[0178] S75. If there is a period of excess energy, the output of distributed energy is reduced, excess energy is transmitted to the power grid, and the excess energy is stored, and the operating state of each load-side resource is reset;

[0179] S76: Evaluate the effect of the new scheduling strategy and compare the key energy indicators before and after the scheduling;

[0180] If there is a significant difference between the evaluation results and the actual results of the dispatch strategy, the energy dynamic balance prediction model will be adjusted.

[0181] It should be explained that, first, the optimized energy dynamic balance prediction model is used to input the historical data and current status data of load data and wind power data to predict future load demand and distributed energy supply. Then, the predicted load demand and distributed energy supply are compared to find the time periods with insufficient or excessive energy. This is a key step because it can help identify the time periods and locations that need to be adjusted. Then, according to the comparison results, the corresponding scheduling strategy is formulated. Specifically, if there is a time period of insufficient energy, it is necessary to increase the output of distributed energy, start energy storage equipment to supplement the supply, and perform load-side management to reset the operating status of each distributed energy. Conversely, if there is a time period of excess energy, it is necessary to reduce the output of distributed energy, transmit excess energy to the power grid, store excess energy, and reset the operating status of each load-side resource. Finally, the effect of the new scheduling strategy is evaluated, and the key energy indicators before and after the scheduling are compared. If there is a large difference between the evaluation results and the actual results of the scheduling strategy, it means that there may be problems with the prediction model, and it needs to be further adjusted and optimized.

[0182] S8. Execute steps S5-S7 in a loop until optimal energy regulation is achieved.

[0183] According to another embodiment of the present invention, there is also provided an energy control system for a virtual power plant based on a prediction model, the system comprising: a virtual power plant data collection and extraction module, a data integration and feature extraction module, a preliminary energy dynamic balance prediction model construction module, a prediction and scheduling strategy formulation module, a real-time scheduling optimization module, an energy dynamic balance prediction model optimization module, an optimization prediction and scheduling strategy formulation module and a feedback module;

[0184] The virtual power plant data collection and extraction module is used to collect energy data in the virtual power plant and extract wind power data and load data from the energy data;

[0185] The data integration and feature extraction module is used to integrate wind power data and load data using a linear regression model, and perform feature extraction on the integrated data to obtain predicted power generation data of distributed energy and predicted demand data of load-side resources;

[0186] The preliminary energy dynamic balance prediction model construction module is used to construct a preliminary energy dynamic balance prediction model using a recurrent neural network based on the predicted power generation data of the distributed energy and the predicted demand data of the load-side resources;

[0187] The forecasting and scheduling strategy formulation module is used to use the preliminary energy dynamic balance forecasting model to forecast the load demand and distributed energy in the day ahead, and formulate a preliminary scheduling strategy based on the forecast results;

[0188] The real-time scheduling optimization module is used to adjust the operating status of each distributed energy source and load-side resource in real time according to the preliminary scheduling strategy using the niche grey wolf optimization algorithm;

[0189] The energy dynamic balance prediction model optimization module is used to dynamically adjust and optimize the energy dynamic balance prediction model according to real-time operation data;

[0190] The optimization prediction and dispatch strategy formulation module is used to re-predict load demand and distributed energy using the optimized energy dynamic balance prediction model, and optimize the dispatch strategy according to the new prediction results;

[0191] The feedback module is used to cyclically execute the real-time scheduling optimization module, the energy dynamic balance prediction model optimization module and the optimization prediction and scheduling strategy formulation module until the optimal energy regulation is achieved.

[0192] Specifically, in order to facilitate a better understanding by those skilled in the art, the technical terms or some nouns that may be involved in the present application are explained in the following embodiments:

[0193] Linear regression model: A linear regression model is a prediction model based on the linear relationship between input variables and output variables. The parameters of the model are determined by minimizing the square error between the actual output value and the predicted output value.

[0194] A recurrent neural network (RNN) is a neural network that can process sequence data. It has memory and can remember previous calculation results and use them for current calculations. This makes RNN very suitable for processing time series data such as speech, text, etc.

[0195] In summary, with the help of the above-mentioned technical scheme of the present invention, the present invention adopts an energy dynamic balance prediction model to predict future load demand and distributed energy supply, and then formulates a scheduling strategy based on the prediction results, which can respond to changes in the power system in advance and achieve more efficient energy scheduling; the present invention obtains wind power data and load data, and then integrates them using a linear regression model to obtain a comprehensive data set containing richer information, which can help us understand the operating status and environmental changes of the power system more comprehensively; by extracting statistical features, frequency domain features and time domain features, data can be analyzed from different angles, so as to better understand the inherent laws of the data; the linear regression model is a simple and effective prediction method that can quickly make predictions, and the model has strong interpretability, which is easy to understand the prediction results of the model; the present invention can adjust distributed energy in real time by using the microhabitat gray wolf optimization algorithm The operating status of the resources on the source and load sides can achieve the optimal operating effect, respond quickly to system changes, and improve the operating efficiency of the system. The setting of the fitness function enables the gray wolf individuals to make adaptive adjustments according to the actual state of the system. This method can ensure that the system always operates in the optimal state and improve the stability of the system; the niche gray wolf optimization algorithm is a global search algorithm that can avoid falling into the local optimal solution. This method can ensure that the solution found is the global optimal solution, thereby realizing the optimal scheduling strategy; by setting the maximum number of iterations, the infinite loop of the algorithm can be avoided and the operating efficiency of the algorithm can be improved. This method can ensure the efficient operation of the algorithm while satisfying the optimization effect; in the process of setting the fitness function and updating the strategy, parameters can be flexibly adjusted according to actual needs, such as the moving step size parameter and the distance adjustment parameter, so as to better adapt to various operating environments and conditions.

[0196] Although the present invention has been disclosed as above with preferred embodiments, the embodiments are merely examples for the convenience of description and are not intended to limit the present invention. Those skilled in the art may make several changes and modifications without departing from the spirit and scope of the present invention. The scope of protection claimed by the present invention shall be based on the claims.

Claims

1. A method for energy control of a virtual power plant based on a prediction model, characterized in that: The energy control method of the virtual power plant includes the following steps: S1. Collect energy data in the virtual power plant and extract wind power data and load data from the energy data; S2. Integrate wind power data and load data using a linear regression model, and extract features from the integrated data to obtain predicted power generation data of distributed energy and predicted demand data of load-side resources; S3. Based on the predicted power generation data of the distributed energy and the predicted demand data of the load-side resources, a preliminary energy dynamic balance prediction model is constructed using a recurrent neural network; S4. Use the preliminary energy dynamic balance forecasting model to forecast the load demand and distributed energy resources in the day ahead, and formulate a preliminary dispatching strategy based on the forecasting results; S5. According to the preliminary dispatching strategy, the operating status of each distributed energy source and load-side resource is adjusted in real time using the niche grey wolf optimization algorithm; S6. Dynamically adjust and optimize the energy dynamic balance prediction model based on real-time operation data; S7. Use the optimized energy dynamic balance prediction model to re-predict load demand and distributed energy, and optimize the scheduling strategy based on the new prediction results; S8. Execute steps S5-S7 in a loop until optimal energy regulation is achieved.

2. The energy control method of a virtual power plant based on a prediction model according to claim 1 is characterized in that: The method of integrating wind power data and load data using a linear regression model and extracting features from the integrated data to obtain predicted power generation data of distributed energy and predicted demand data of load-side resources includes the following steps: S21. Acquire wind power data of the virtual power plant, wherein the wind power data at least includes wind speed, wind direction, temperature, humidity and atmospheric pressure; S22, obtaining load data of the virtual power plant, wherein the load data at least includes load demand, equipment operation status and environmental parameters; S23, performing linear regression integration on the wind speed, wind direction, temperature, humidity and atmospheric pressure of the wind power data and the load demand, equipment operation status and environmental parameters of the load data to obtain a comprehensive data set including both wind power data and load data; S24, extracting features of the data in the comprehensive data set, wherein the features at least include statistical features, frequency domain features, and time domain features; S25. Based on the extracted features, a linear regression model is used to obtain the predicted power generation data of distributed energy and the predicted demand data of load-side resources.

3. The energy control method of a virtual power plant based on a prediction model according to claim 2 is characterized in that: The method of obtaining the predicted power generation data of distributed energy and the predicted demand data of load-side resources based on the extracted features using a linear regression model includes the following steps: S251. According to the characteristics of the data in the comprehensive data set, combined with the historical power generation data and load demand data, a linear regression model is constructed; S252, dividing the comprehensive data set into a training set and a test set, wherein the training set is used to train the linear regression model, and the test set is used to evaluate the prediction ability of the linear regression model; S253, training a linear regression model using a training set to learn the relationship between data features, power generation data, and load demand data; S254, use the trained linear regression model to predict the test set and evaluate the prediction performance of the model; S255. If the prediction performance of the linear regression model meets the requirements, it is used to predict future distributed energy generation data and load-side resource demand data; If the prediction performance of the linear regression model does not meet the requirements, reselect the characteristics of the data in the comprehensive dataset and adjust the linear regression model parameters.

4. The energy control method of a virtual power plant based on a prediction model according to claim 1 is characterized in that: The method of using the preliminary energy dynamic balance prediction model to predict the load demand and distributed energy resources in the day ahead and formulating a preliminary dispatching strategy based on the prediction results includes the following steps: S41. Use the preliminary energy dynamic balance prediction model to input the statistical characteristics, frequency domain characteristics, and time domain characteristics of the load data to perform a load demand forecast for the day ahead; S42, using a preliminary energy dynamic balance prediction model, inputting statistical characteristics, frequency domain characteristics, and time domain characteristics of wind power data to perform a day-ahead distributed energy generation prediction; S43. Formulate a preliminary dispatching strategy based on the forecast results of the load demand forecast and the distributed energy generation forecast.

5. The energy control method of a virtual power plant based on a prediction model according to claim 1 is characterized in that: According to the preliminary scheduling strategy, the operation status of each distributed energy source and load-side resource is adjusted in real time using the niche grey wolf optimization algorithm, including the following steps: S51, according to the preliminary dispatching strategy, determine the initial operating state of each distributed energy source and load-side resource, and regard the initial operating state as a gray wolf individual in the microhabitat gray wolf optimization algorithm; S52, setting a fitness function, evaluating the quality of each individual gray wolf, and calculating the fitness value of each individual gray wolf; S53, sorting all gray wolf individuals according to the fitness value, selecting the best three gray wolf individuals, and recording them as α gray wolf, β gray wolf and δ gray wolf, wherein the running state corresponding to the best fitness value is recorded as the current optimal solution; S54, calculating an update strategy according to the operating states and corresponding fitness values ​​of the α gray wolf, the β gray wolf, and the δ gray wolf, and updating the operating states of the remaining gray wolf individuals; S55, according to the update strategy, update the operating status of the remaining gray wolf individuals, obtain a new operating status, and re-evaluate the fitness values ​​of the remaining gray wolf individuals to find a new optimal solution; S56, determining whether the maximum number of iterations has been reached; If so, output the optimal solution; If not, execute step S52; S57. According to the optimal solution output by the microhabitat grey wolf optimization algorithm, the operating status of each distributed energy source and load-side resource is adjusted in real time.

6. The energy control method of a virtual power plant based on a prediction model according to claim 5 is characterized in that: The step of setting the fitness function, evaluating the quality of each individual gray wolf, and calculating the fitness value of each individual gray wolf comprises the following steps: S521. Define the fitness function of the niche grey wolf optimization algorithm according to the scheduling problem of distributed energy and load-side resources, and ensure the computability and interpretability of the fitness function; S522, determining each individual gray wolf, obtaining the running status data of each individual gray wolf, and ensuring the accuracy and timeliness of the data; S523, calculating and recording the fitness value of each individual gray wolf through the fitness function, and setting a mechanism for automatically recalculating the fitness value to update the running status data of each individual gray wolf in real time; S524. Regularly check the validity of the fitness function and adjust the fitness function as needed.

7. The energy control method of a virtual power plant based on a prediction model according to claim 5 is characterized in that: The step of calculating the update strategy according to the running status and corresponding fitness values ​​of the α gray wolf, the β gray wolf and the δ gray wolf, and updating the running status of the remaining gray wolf individuals comprises the following steps: S541, setting the moving step length parameter A and the distance adjustment parameter C of the niche gray wolf optimization algorithm; S542, calculating the new position of each non-optimal gray wolf according to the positions of the α gray wolf, the β gray wolf and the δ gray wolf and the moving step length parameter A and the distance adjustment parameter C; S543, performing range detection on the newly calculated non-optimal gray wolf position to determine whether it exceeds the preset solution space range; If it exceeds the solution space, the parameters need to be adjusted to return it to the solution space; S544, substituting the adjusted new position into the expression of the running state to obtain a new running state, and updating the running state of the non-optimal gray wolf.

8. The energy control method of a virtual power plant based on a prediction model according to claim 7 is characterized in that: The calculation of the new position of each non-optimal gray wolf includes: Calculate the distance D between the non-optimal gray wolf and the optimal gray wolf α respectively α , the distance Dβ between the non-optimal gray wolf and the optimal gray wolf β and the distance D between the non-optimal gray wolf and the optimal gray wolf δ δ ; According to the distance D α , distance D β and distance D δ , calculate the new position of the non-optimal gray wolf after it moves according to the behavior of the optimal gray wolf α, the optimal gray wolf δ, and the optimal gray wolf δ; Wherein, the distance D α The calculation formula is as follows: D α =|C1·α-A1·X| The distance D β The calculation formula is as follows: D β =|C2·β-A2·X| The distance D δ The calculation formula is as follows: D δ =|C3·δ-A3·X| Where D α Represents the distance between the non-optimal gray wolf and the optimal gray wolf α; D β represents the distance between the optimal gray wolf and the optimal gray wolf β; D δ represents the distance between the optimal gray wolf and the optimal gray wolf δ; C1, C2, and C3 represent randomly generated distance adjustment parameters; A1, A2 and A3 represent the calculated moving step parameters respectively; X represents the current position of the non-optimal gray wolf.

9. The energy control method of a virtual power plant based on a prediction model according to claim 1 is characterized in that: The method of re-forecasting load demand and distributed energy using the optimized energy dynamic balance forecasting model and optimizing the dispatching strategy according to the new forecasting results includes the following steps: S71. Use the optimized energy dynamic balance prediction model, input historical data and current state data of load data, and predict future load demand; S72, using the optimized energy dynamic balance prediction model, inputting historical data and current status data of wind power data, and predicting the future distributed energy supply; S73. Compare the predicted load demand and distributed energy supply to find out the time periods of energy shortage or surplus; S74. If there is a period of insufficient energy, the output of distributed energy is increased, energy storage equipment is started to supplement the supply, and load-side management is performed to reset the operating status of each distributed energy source; S75. If there is a period of excess energy, the output of distributed energy is reduced, excess energy is transmitted to the power grid, and the excess energy is stored, and the operating state of each load-side resource is reset; S76: Evaluate the effect of the new scheduling strategy and compare the key energy indicators before and after the scheduling; If there is a significant difference between the evaluation results and the actual results of the dispatch strategy, the energy dynamic balance prediction model will be adjusted.

10. An energy control system for a virtual power plant based on a prediction model, used to implement the energy control method for a virtual power plant based on a prediction model as claimed in any one of claims 1 to 9, characterized in that: The system includes: virtual power plant data collection and extraction module, data integration and feature extraction module, preliminary energy dynamic balance prediction model construction module, prediction and scheduling strategy formulation module, real-time scheduling optimization module, energy dynamic balance prediction model optimization module, optimization prediction and scheduling strategy formulation module and feedback module; The virtual power plant data collection and extraction module is used to collect energy data in the virtual power plant and extract wind power data and load data from the energy data; The data integration and feature extraction module is used to integrate wind power data and load data using a linear regression model, and perform feature extraction on the integrated data to obtain predicted power generation data of distributed energy and predicted demand data of load-side resources; The preliminary energy dynamic balance prediction model construction module is used to construct a preliminary energy dynamic balance prediction model using a recurrent neural network based on the predicted power generation data of the distributed energy and the predicted demand data of the load-side resources; The forecasting and scheduling strategy formulation module is used to use the preliminary energy dynamic balance forecasting model to forecast the load demand and distributed energy in the day ahead, and formulate a preliminary scheduling strategy based on the forecast results; The real-time scheduling optimization module is used to adjust the operating status of each distributed energy source and load-side resource in real time according to the preliminary scheduling strategy using the niche grey wolf optimization algorithm; The energy dynamic balance prediction model optimization module is used to dynamically adjust and optimize the energy dynamic balance prediction model according to real-time operation data; The optimization prediction and dispatch strategy formulation module is used to re-predict load demand and distributed energy using the optimized energy dynamic balance prediction model, and optimize the dispatch strategy according to the new prediction results; The feedback module is used to cyclically execute the real-time scheduling optimization module, the energy dynamic balance prediction model optimization module and the optimization prediction and scheduling strategy formulation module until the optimal energy regulation is achieved.

Images