Multi-virtual power plant fast response control method containing wind, light storage and demand side response

Abstract

The present application relates to the technical field of power regulation, in particular to a multi-virtual power plant fast response control method containing wind, light, energy storage and demand side response. In the spatial prediction process, the demand side is divided into nodes and sub-nodes, the distribution characteristics of the energy output proportion of the sub-nodes in the nodes are determined by constructing a two-dimensional space method through statistical historical data, and the energy proportion of each sub-node is predicted based on the spatial distribution of the demand side to obtain a spatial prediction value. In the time prediction process, the corresponding prediction fluctuation adjustment amount at each time is analyzed based on historical data, and the time prediction value of each sub-node can be determined on the time scale by constructing an equation and solving it. The combination of the two prediction processes obtains the final prediction value and controls the power distribution. The final prediction value obtained by the present application can effectively represent the power consumption characteristics of each specific position of the demand side at the future time, thereby enabling the multi-virtual power plant to make scientific and effective power allocation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power regulation, in particular to a multi-virtual power plant fast response control method containing wind-solar energy storage and demand side response. BACKGROUND

[0002] The wind-solar energy storage system aggregates distributed energy sources (such as wind power, photovoltaic), energy storage systems (such as lithium batteries, pumped storage), and controllable loads through physical aggregation to form a unified energy management unit. At the same time, combined with the demand side response mechanism, the demand side is guided to adjust the load by using the electricity demand response incentive to reduce the peak-valley difference, and the fast response and peak load shifting are realized through collaborative control. Distributed multi-virtual power plants (VPP) are used to realize fast response control, and the renewable energy consumption capacity and grid system stability of the entire grid system are improved.

[0003] In dynamic load management, the use of power system demand side load can be predicted, and grid scheduling instructions can be given according to the predicted load to flexibly adjust the control strategy and issue it to energy storage, charging piles, inverters and other systems and equipment, so as to realize fast response to the demand side and ensure the safe, efficient and low-carbon operation of the grid.

[0004] However, after the distributed generator set in the wind-solar energy storage system is connected to the grid, the dynamic power imbalance of the power system is caused, and the predicted value of each position monitored by each regional monitoring data needs to be obtained directly when the obtained load prediction and wind-solar power prediction are performed. The overall energy use is obtained by accumulating the obtained predicted values, which may cause the prediction results to contain noise due to the diversity of the power consumption in each small area, and the noise may affect the final result when the overall accumulation is performed. Directly predicting the overall may make it difficult to reflect the granularity of the demand side characteristics of each region. SUMMARY

[0005] In order to solve the technical problem that the power consumption data prediction of the demand side is not accurate in the prior art, resulting in that the multi-virtual power plant cannot effectively respond to the demand side, the purpose of the present application is to provide a multi-virtual power plant fast response control method containing wind-solar energy storage and demand side response, and the technical solution adopted is as follows:

[0006] The present application provides a multi-virtual power plant fast response control method containing wind-solar energy storage and demand side response, which comprises the following steps:

[0007] The demand side is divided into multiple nodes according to regions, and each node contains multiple sub-nodes; in a node, for each sub-node, the total energy output in the historical data is taken as the horizontal axis, and the proportion of the energy output of the sub-node in the node under the total energy output is taken as the vertical axis to obtain a two-dimensional space of each sub-node; a first prediction value of the total energy output of each node at the current time is obtained by using an ARIMA algorithm, a data point corresponding to the first prediction value in the two-dimensional space is a reference data point of each sub-node, a predicted energy proportion of each sub-node is obtained according to the reference data point, and a space prediction value is obtained;

[0008] A real-time power consumption data sequence of the sub-node in the current time period is obtained, and a historical data segment sequence linearly related to the real-time power consumption data sequence is screened out in a historical database; a prediction fluctuation adjustment amount of each time in the current time period is obtained according to the element fluctuation of each historical data segment sequence; a second prediction value of each time in the current time period is obtained by using an ARIMA algorithm; an equation is constructed and solved according to the cumulative value of the second prediction value and the prediction fluctuation adjustment amount, and an initial value of the power consumption of the sub-node is obtained, and a time prediction value of the sub-node at the current time is further obtained in combination with the prediction fluctuation adjustment amount;

[0009] A confidence weight is obtained according to the data fluctuation of the reference data point and the historical data segment sequence; the space prediction value and the time prediction value of each sub-node are weighted and fused based on the confidence weight to obtain a final prediction value of each sub-node; and the power distribution of the multiple virtual power plants is controlled according to the final prediction value.

[0010] Further, the obtaining of the prediction energy proportion of each sub-node according to the reference data point and the obtaining of the space prediction value of each sub-node include:

[0011] For each reference data point, the stability of each reference data point is obtained according to the vertical coordinate difference between the reference data point and other reference data points; the vertical coordinates of each reference data point are weighted and averaged based on the stability to obtain the prediction energy proportion of each sub-node, and the space prediction value of the sub-node is obtained by multiplying the prediction energy proportion and the space prediction value of the node.

[0012] Further, the stability obtaining method includes:

[0013] For any one reference data point, the vertical coordinate difference between the reference data point and each other reference data point is obtained, the cumulative value of the vertical coordinate difference is negatively correlated and normalized to obtain the stability of the reference data point.

[0014] Further, the screening process of the historical data segment sequence includes:

[0015] In the historical database, a candidate historical data segment sequence in the same time period as the current time period is obtained, a Pearson correlation coefficient between the real-time power consumption data sequence and each candidate historical data segment sequence is obtained, and a candidate historical data segment sequence with a Pearson correlation coefficient greater than a preset threshold is taken as a historical data segment sequence linearly correlated with the real-time power consumption data sequence.

[0016] Further, the method for obtaining the predicted fluctuation adjustment amount comprises:

[0017] For any one historical data segment sequence, a data difference value between an element at each time point and an element at a previous time point is taken as a first-order difference value at each time point, and the first-order difference value at an initial time point is 0;

[0018] All historical data segment sequences are counted, and an average first-order difference value at each time point is taken as the predicted fluctuation adjustment amount.

[0019] Further, an equation expression constructed according to the accumulated value of the second prediction value and the predicted fluctuation adjustment amount is:

[0020] ; wherein is an initial power consumption value, is the accumulated value of the second prediction value, is the predicted fluctuation adjustment amount at a first time point except the initial time point, is the predicted fluctuation adjustment amount at a second time point except the initial time point, is the predicted fluctuation adjustment amount at an i-th time point except the initial time point, is the number of time points in the current time period.

[0021] Further, the time prediction value is in the equation expression .

[0022] Further, the method for obtaining the confidence weight comprises:

[0023] A variance of the ordinate values of all reference data points is taken as a first variance, an average variance of the elements in each historical data segment sequence is obtained, and an average variance of all historical data segment sequences is taken as a second variance; a ratio of the first variance to the second variance is normalized to obtain the confidence weight.

[0024] Further, the method for obtaining the final prediction value comprises:

[0025] The confidence weight is taken as a weight of the time prediction value, a result of subtracting the confidence weight from a positive integer 1 is taken as a weight of a space prediction value, and the time prediction value and the space prediction value are weighted and summed to obtain the final prediction value.

[0026] Further, the controlling power distribution of the multi-virtual power plant according to the final prediction value comprises:

[0027] Obtaining energy output proportions of different virtual power plants, and performing power distribution on a node where each sub-node is located according to the energy output proportions and the final prediction value.

[0028] The present application has the following beneficial effects:

[0029] In order to effectively predict the power consumption data of each position of the demand side, the prediction process is divided into spatial prediction and time prediction two parts. In the spatial prediction process, the demand side is divided into nodes and sub-nodes, the distribution characteristics of the energy output proportion of the sub-nodes in the nodes are determined by constructing a two-dimensional space method based on statistical historical data, and then the energy output proportion of each sub-node is predicted based on the spatial distribution of the demand side to obtain a spatial prediction value. In the time prediction process, the present application divides the time sequence range into time points, and analyzes the corresponding prediction fluctuation adjustment amount at each time point based on historical data. By constructing an equation and solving, the time prediction value of each sub-node can be determined in the time scale. In the spatial prediction process, the diversity of the power consumption process in the small area can be considered. In the time prediction process, the characteristic granularity of each small area can be reflected through detailed time point analysis. The final prediction value obtained by combining the two prediction processes can effectively represent the power consumption characteristics of each specific position of the demand side at the future time, so that the multi-virtual power plant can make scientific and effective power allocation. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor based on these drawings.

[0031] Figure 1 A flow chart of a multi-virtual power plant fast response control method with wind, light, energy storage and demand side response provided by an embodiment of the present application;

[0032] Figure 2 An output power waveform diagram provided by an embodiment of the present application. DETAILED DESCRIPTION

[0033] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined inventive purpose, the following describes in detail the specific implementation, structure, features and effects of a multi-virtual power plant fast response control method with wind and light energy storage and demand side response according to the present application, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0035] The specific scheme of the multi-virtual power plant fast response control method with wind and light energy storage and demand side response provided by the present application is described in detail below in combination with the accompanying drawings.

[0036] Please refer to Figure 1 which shows a flow chart of a multi-virtual power plant fast response control method with wind and light energy storage and demand side response provided by one embodiment of the present application. The method comprises:

[0037] Step S1: dividing the demand side into multiple nodes, each node containing multiple sub-nodes; in one node, for each sub-node, taking the total energy output in the historical data as the horizontal axis and the energy output proportion of the sub-node in the node under the total energy output as the vertical axis to obtain the two-dimensional space of each sub-node; using the ARIMA algorithm to obtain the spatial prediction value of the total energy output of each node at the current time, the data point corresponding to the first prediction value in the two-dimensional space being the reference data point of each sub-node, and obtaining the predicted energy proportion of each sub-node and the spatial prediction value according to the reference data point.

[0038] In the embodiment of the present application, the demand side is divided into multiple nodes, each node containing multiple sub-nodes. The node is the entire park industrial park or large area such as street, block, etc. in the demand side, and the sub-node is the small area such as each factory, industrial building, enterprise, residential user, commercial user, etc. in the large area. Through the division of nodes and sub-nodes, different scales can be set and analyzed in space, which is convenient for subsequent analysis of the multiple power consumption situations of each sub-node.

[0039] In the embodiment of the present application, the power consumption data of the target can be collected in the node and sub-node through the smart meter. There will be a total meter in each large area, and each sub-area meter is associated with the total meter, i.e. the node and sub-node in the embodiment of the present application can also be regarded as the node with the smart meter as the object. The power consumption data and energy output data in the embodiment of the present application are embodied by power. Please refer to Figure 2Fig. 1 shows an output power waveform diagram provided by one embodiment of the present application, i.e. energy output data of nodes and sub-nodes can be obtained at each sampling time.

[0040] In the embodiment of the present application, in order to facilitate prediction analysis, it is necessary to divide time periods in a time sequence range, and then in the subsequent prediction process, the first prediction value and the second prediction value obtained by the conventional prediction algorithm can be obtained by analyzing the electricity consumption data of the time period in which the current time is located. In the embodiment of the present application, 24 hours of a day are divided into five time periods, which are: 00:00-7:20 representing the night time period; 07:20-11:30 representing the morning time period; 11:30-15:00 representing the noon time period; 15:00-17:20 representing the afternoon time period; and 17:20-00:00 representing the evening time period. For example, if the current time is 12:00 noon, the corresponding time period is the noon time period, and the current time period to be analyzed is 11:30-12:00.

[0041] Step S1 of the embodiment of the present application is a prediction process under a spatial scale, and the energy output proportion of a sub-node in a node can be used as the electricity consumption feature of the sub-node. In order to analyze various electricity consumption features of the sub-node, it is necessary to statistically evaluate the historical data, and the historical data of a node is counted in the embodiment of the present application. The total energy output of the node is taken as the horizontal axis, and the energy output proportion of each sub-node in the node is taken as the vertical axis under each total energy output, i.e. a two-dimensional space of each sub-node can be obtained. That is, one two-dimensional space represents the energy proportion of a sub-node in the node of the corresponding large area, and the two-dimensional space contains various electricity consumption features of the sub-node. It should be noted that the energy output proportion is the ratio of the power output power of a sub-node to the total power output power of the corresponding node.

[0042] In order to realize the prediction of each sub-node in space, the embodiment of the present application first obtains the first prediction value of the total energy output of each node at the current time by using the ARIMA algorithm, i.e. the first prediction value is a large-scale prediction value under a large area, and then the prediction energy proportion of each sub-node can be predicted to determine the spatial prediction value of the sub-node under the spatial scale. In the two-dimensional space, the horizontal axis is the total energy output, and the first prediction value can be corresponded in the two-dimensional space. The data points corresponding to the first prediction value represent the various electricity consumption situations of the sub-node under the total energy output of the first prediction value, and therefore the data points corresponding to the first prediction value in the two-dimensional space are recorded as the reference data points of each sub-node. The prediction energy proportion of each sub-node can be obtained by counting the reference data points, and the spatial prediction value of the corresponding sub-node can be obtained by multiplying the prediction energy proportion by the first prediction value.

[0043] It should be noted that the time period of the embodiment of the present application is divided in units of days, and the time sequence range in the historical database is set to one week, that is, the historical data in the last one week is counted to construct the two-dimensional space. In order to make the two-dimensional space more accurate, the same historical time as the current time in the last one week can be further selected for construction.

[0044] Preferably, in an embodiment of the present application, although the reference data points represent various power consumption situations of the sub-nodes, some power consumption situations are not common. If the ordinate information of all reference data points is directly counted, the final prediction result will be affected by such uncommon power consumption situations. The more uncommon the power consumption situation is, the more discrete the reference data points will be in the reference data point set. Therefore, the weight can be determined based on the dispersion between the reference data points and other reference data points, and then the accurate prediction energy ratio result is obtained, which specifically includes:

[0045] For each reference data point, the stability of each reference data point is obtained according to the difference in ordinate between the reference data point and other reference data points, that is, the smaller the difference in ordinate, the more stable the reference data point is in the reference data point set, and the greater the reference is. The ordinate of each reference data point is weighted and averaged based on the stability to obtain the prediction energy ratio of each sub-node. The prediction energy ratio is multiplied by the spatial prediction value of the node to obtain the spatial prediction value of the sub-node.

[0046] Further, the method for obtaining the stability includes:

[0047] For any one reference data point, the difference in ordinate between the reference data point and each other reference data point is obtained, the cumulative value of the difference in ordinate is negatively correlated and normalized to obtain the stability of the reference data point. In the embodiment of the present application, the method for negatively correlating and normalizing is: obtaining the reciprocal of the cumulative value of the difference in ordinate, taking the reciprocal as the power of an exponential function with a natural constant as the base, and the mapping result of the exponential function is the stability. In other embodiments of the present application, other methods for negatively correlating and normalizing can also be selected, which are not described and limited here.

[0048] Step S2: obtaining the real-time power consumption data sequence of the sub-node in the current time period, screening out the historical data segment sequence linearly correlated with the real-time power consumption data sequence in the historical database; obtaining the prediction fluctuation adjustment amount of each time in the current time period according to the element volatility in each historical data segment sequence; obtaining the second prediction value of each time in the current time period by using the ARIMA algorithm; constructing an equation according to the cumulative value of the second prediction value and the prediction fluctuation adjustment amount and solving it to obtain the initial value of the power consumption of the sub-node, and further combining the prediction fluctuation adjustment amount to obtain the time prediction value of the sub-node at the current time.

[0049] The embodiment of the present application further obtains the prediction result in the time scale. For the power demand side, the fluctuation of the data demand side obtained in different time scales is different. When the data prediction is performed, the energy use at each time point is disturbed by the change of the demand side. For example, there is a time difference of strong demand peak in each factory or each residential area. Generally, the factory can adjust the peak load by the three-shift form to reduce the peak demand in a certain time period, so as to achieve the effect of peak load shaving and smoothing of the power grid. The demand peak of the power grid of each residential area has great similarity in time segmentation. According to the demand characteristics of the demand side in each sub-region in a large area, the demand side response adjustment is comprehensively adjusted. Therefore, the overall prediction can cause the fluctuation of the data in different time to be ignored. When the data to be analyzed is obtained, the prediction value in a period of time can be split into different time fluctuations.

[0050] In order to realize the effective prediction of each time point in the current time period, the embodiment of the present application first obtains the real-time power consumption data sequence of the sub-node in the current time period, and then screens the historical data segment sequence linearly related to the real-time power consumption data sequence in the historical database. The historical data segment sequence is used as the data reference for prediction in the time scale. It should be noted that in order to ensure the reference of the historical data, the historical data segment sequence should belong to the same time period range as the real-time power consumption data sequence. For example, the current time period is 11:30 to 12:00, and the historical data segment sequence is also 11:30 to 12:00. That is, the historical data segment sequence is the data segment between 11:30 and 12:00 on different days in a week, and is linearly related to the real-time power consumption data sequence.

[0051] Preferably, in the embodiment of the present application, the screening process of the historical data segment sequence includes:

[0052] In the historical database, the selected historical data segment sequence in the same time period as the current time period is obtained, the Pearson correlation coefficient between the real-time power consumption data sequence and each selected historical data segment sequence is obtained, and the selected historical data segment sequence with the Pearson correlation coefficient greater than the preset threshold is used as the historical data segment sequence linearly related to the real-time power consumption data sequence. The Pearson correlation coefficient is set to 0.8 in the embodiment of the present application.

[0053] In order to analyze the change in the time scale, the embodiment of the present application quantifies the fluctuation between each time point by using the historical data segment sequence. Because the historical data segment sequence is a strong correlation sequence of the real-time power consumption data sequence, it can be used as a reference to analyze the power consumption data change characteristics at each time point, that is, the predicted fluctuation adjustment amount of each time point in the current time period is obtained according to the element fluctuation in each historical data segment sequence. The predicted fluctuation adjustment amount represents the change amount of the next time point relative to the previous time point when electricity consumption occurs at the next time point for the current time point. Based on this, the second predicted value of each time point in the current time period can be further obtained by using the ARIMA algorithm, the cumulative value of the second predicted value is used as a known value, the initial value of the initial time point is set as an unknown value, and the equation can be constructed by combining the predicted fluctuation adjustment amount of each time point. The electricity consumption initial value of the sub-node at the initial time point can be obtained by solving the equation, and the time prediction value of the sub-node at the current time point can be further obtained by combining the predicted fluctuation adjustment amount. That is, the significance of constructing the equation is that the second predicted value is the predicted value of each time point based on the overall historical time point. By accumulating and using it for equation solving, the predicted value of each small scale can be obtained from the large scale in time. The introduction of the predicted fluctuation adjustment amount in the equation can avoid ignoring the fluctuation of the data at different times, so that the equation can better reflect the effective power consumption accumulation process of each time point in the current time period.

[0054] Preferably, in the embodiment of the present application, the method for obtaining the predicted fluctuation adjustment amount comprises:

[0055] For any one historical data segment sequence, the data difference value between each time point and the previous time point element is used as the first-order difference value of each time point. Because there is no previous time point at the initial time point, the first-order difference value of the initial time point is set to 0.

[0056] All historical data segment sequences are counted, and the average first-order difference value of each time point is used as the predicted fluctuation adjustment amount. That is, the predicted fluctuation adjustment amount of the initial time point is 0.

[0057] Preferably, the equation expression constructed according to the cumulative value of the second predicted value and the predicted fluctuation adjustment amount is:

[0058] ; wherein is the electricity consumption initial value, is the cumulative value of the second predicted value, is the predicted fluctuation adjustment amount of the first time point except the initial time point, is the predicted fluctuation adjustment amount of the second time point except the initial time point, is the predicted fluctuation adjustment amount of the i-th time point except the initial time point, is the number of time points in the current time period.

[0059] The time prediction value at the current time is .

[0060] Step S3: obtaining a confidence weight according to the data volatility of the reference data points and the historical data segment sequence; weighting and fusing the spatial prediction value and the time prediction value of each sub-node based on the confidence weight to obtain a final prediction value of each sub-node; and controlling the power distribution of the multi-virtual power plant according to the final prediction value.

[0061] In the above steps S1 and S2, the prediction of the spatial scale and the prediction of the time scale are respectively performed, and two kinds of prediction values are obtained for each sub-node. It is further needed to fuse the two kinds of prediction values to obtain the final prediction value. Therefore, the embodiment of the present application adopts the idea of weighted fusion, and obtains a confidence weight according to the data volatility of the reference data points and the historical data segment sequence, that is, the greater the data volatility, the lower the information reference degree at this scale, and the higher the reference degree at the other scale. The final prediction value of each sub-node can be obtained by weighting and fusing the spatial prediction value and the time prediction value of each sub-node through the confidence weight.

[0062] Preferably, in the embodiment of the present application, the method for obtaining the confidence weight comprises:

[0063] The variance of the ordinate values of all the reference data points is taken as a first variance, and the average variance of all the historical data segment sequences is taken as a second variance. The ratio of the first variance to the second variance is normalized to obtain the confidence weight. That is, the greater the first variance, the greater the volatility of the reference data points, and the lower the weight of the spatial prediction value in the fusion process at the spatial scale, and the higher the weight of the time prediction value in the fusion process at the time scale. Therefore, the confidence weight can be directly taken as the weight of the time prediction value. Since the value domain of the confidence weight is between 0 and 1, the result of subtracting the confidence weight from the integer 1 is taken as the weight of the spatial prediction value. The time prediction value and the spatial prediction value are weighted and summed to obtain the final prediction value.

[0064] In the embodiment of the present application, the normalization operation is realized by using the range standardization, and the normalization of the data is realized by analyzing the maximum and minimum values in the dimension. In other embodiments of the present application, other normalization methods can also be selected, which will not be described here.

[0065] After the final prediction value is determined, the power distribution of the multi-virtual power plant can be controlled according to the final prediction value.

[0066] Preferably, in the embodiments of the present application, considering that the multi-virtual power plant contains different power generation and energy storage parts such as wind and light, the energy output proportions of different virtual power plants need to be obtained before power distribution, and then the power that each virtual power plant needs to distribute at the sub-node is determined according to the multiplication of the energy output proportion and the final prediction value. In the specific distribution process, the power required by all sub-nodes can be integrated, and the virtual power plant can distribute the power to the nodes corresponding to the sub-nodes. It should be noted that if the power generation of the virtual power plant is less than the demand of the demand side in the power distribution process, the transmission path of the optimized power grid can be optimized by EMS. Ensure that power can be transmitted from areas with sufficient supply to areas with high demand, while avoiding overload and loss. And in some embodiments of the present application, adjustment priorities can also be set for different regions to ensure that fast response areas can have priority in power allocation. The specific power allocation process is a technical means known to those skilled in the art, and will not be described here.

[0067] In summary, the embodiments of the present application divide the prediction process into spatial prediction and temporal prediction. In the spatial prediction process, the demand side is divided into nodes and sub-nodes, the distribution characteristics of the energy output proportion of the sub-nodes in the nodes are determined by constructing a two-dimensional space method based on statistical historical data, and then the energy proportion of each sub-node is predicted based on the spatial distribution of the demand side to obtain a spatial prediction value. In the time prediction process, the present application divides the time sequence range into time points, and analyzes the corresponding prediction fluctuation adjustment amount at each time point based on historical data. By constructing equations and solving, the time prediction value of each sub-node can be determined on the time scale. The final prediction value obtained by combining the two prediction processes controls the power distribution of the multi-virtual power plant. The final prediction value obtained by the present application can effectively represent the power consumption characteristics of each specific position of the demand side at future time, thereby enabling the multi-virtual power plant to make scientific and effective power allocation.

[0068] It should be noted that the above-mentioned embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or can be advantageous.

[0069] Each embodiment in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A fast response control method for multiple virtual power plants including wind, solar, energy storage and demand-side response, characterized in that: The method comprises: The demand side is divided into multiple nodes according to the region, and each node contains multiple sub-nodes. In a node, for each sub-node, the total energy output in the historical data is used as the horizontal axis, and the energy output ratio of the sub-node in the node under the total energy output is used as the vertical axis to obtain the two-dimensional space of each sub-node. The ARIMA algorithm is used to obtain the first predicted value of the total energy output of each node at the current moment. The data point corresponding to the first predicted value in the two-dimensional space is the reference data point of each sub-node. Based on the reference data point, the predicted energy ratio of each sub-node and the spatial prediction value are obtained. Obtain the real-time electricity consumption data sequence of the child node in the current time period, and filter out the historical data segment sequence that is linearly correlated with the real-time electricity consumption data sequence from the historical database; obtain the predicted fluctuation adjustment value for each moment in the current time period based on the element volatility in each historical data segment sequence; use the ARIMA algorithm to obtain the second predicted value for each moment in the current time period; construct and solve an equation based on the accumulated value of the second predicted value and the predicted fluctuation adjustment value to obtain the initial value of the electricity consumption of the child node, and further combine the predicted fluctuation adjustment value to obtain the time prediction value of the child node at the current moment; Confidence weights are obtained based on the data volatility of the reference data point and the historical data segment sequence. The spatial prediction value and the temporal prediction value of each child node are weighted and fused based on the confidence weights to obtain the final prediction value of each child node. The power distribution of multiple virtual power plants is controlled based on the final prediction value. The method for obtaining the predicted fluctuation adjustment amount includes: For any historical data segment sequence, the data difference between each moment and the previous moment is taken as the first-order difference value of each moment; the first-order difference value of the initial moment is 0; Counting all historical data segment sequences, and taking the average first-order difference value at each moment as the predicted fluctuation adjustment amount; The equation constructed based on the accumulated value of the second predicted value and the predicted fluctuation adjustment amount is: ;in is the initial value of power consumption, is the accumulated value of the second predicted value, is the forecast fluctuation adjustment at the first moment other than the initial moment, is the forecast fluctuation adjustment at the second moment other than the initial moment, is the predicted fluctuation adjustment at the i-th moment except the initial moment, is the number of time points in the current time period; The time prediction value is in the equation expression .

2. A fast response control method for multiple virtual power plants with wind, solar, energy storage and demand-side response according to claim 1, characterized in that: Obtaining the predicted energy proportion of each subnode and the spatial prediction value based on the reference data point includes: For each reference data point, the stability of each reference data point is obtained based on the difference in the vertical coordinates between the reference data point and other reference data points; based on the stability, the vertical coordinates of each reference data point are weighted and averaged to obtain the predicted energy share of each child node, and the predicted energy share is multiplied by the spatial prediction value of the node to obtain the spatial prediction value of the child node.

3. The method for rapid response control of multiple virtual power plants with wind, solar, energy storage and demand-side response according to claim 2 is characterized in that: The method for obtaining the stability includes: For any reference data point, the vertical coordinate difference between the reference data point and each other reference data point is obtained, and the accumulated value of the vertical coordinate difference is negatively correlated and normalized to obtain the stability of the reference data point.

4. The method for rapid response control of multiple virtual power plants with wind, solar, energy storage and demand-side response according to claim 1 is characterized in that: The screening process of the historical data segment sequence includes: In the historical database, a candidate historical data segment sequence in the same time period as the current time period is obtained, and the Pearson correlation coefficient between the real-time electricity consumption data sequence and each candidate historical data segment sequence is obtained. The candidate historical data segment sequence with a Pearson correlation coefficient greater than a preset threshold is regarded as a historical data segment sequence linearly correlated with the real-time electricity consumption data sequence.

5. The method for rapid response control of multiple virtual power plants with wind, solar, energy storage and demand-side response according to claim 1 is characterized in that: The method for obtaining the confidence weight includes: The variance of the vertical coordinate values ​​of all reference data points is used as the first variance; the variance of the elements in each historical data segment sequence is obtained, and the average variance of all historical data segment sequences is used as the second variance; the ratio of the first variance to the second variance is normalized to obtain the confidence weight.

6. A fast response control method for multiple virtual power plants with wind, solar, energy storage and demand-side response according to claim 5, characterized in that: The method for obtaining the final prediction value includes: The confidence weight is used as the weight of the time prediction value, the result of subtracting the confidence weight from the positive integer 1 is used as the weight of the spatial prediction value, and the time prediction value and the spatial prediction value are weighted and summed to obtain the final prediction value.

7. The method for rapid response control of multiple virtual power plants with wind, solar, energy storage and demand-side response according to claim 1 is characterized in that: The controlling of power distribution of multiple virtual power plants according to the final prediction value includes: The energy output ratios of different virtual power plants are obtained, and power is distributed to the node where each child node is located according to the energy output ratios and the final predicted value.

Images