CN-122001011-A - Virtual power plant dynamic scheduling method based on multi-source data fusion

CN122001011ACN 122001011 ACN122001011 ACN 122001011ACN-122001011-A

Abstract

The invention discloses a virtual power plant dynamic scheduling method based on multi-source data fusion, which comprises the steps of obtaining multi-source power data in a virtual power plant, unifying the multi-source power data to obtain an initial data set, constructing a feature matrix according to the initial data set and historical power generation output data, extracting key factors according to the feature matrix to obtain a feature data matrix, carrying out time sequence analysis on the feature data matrix through a deep learning model to obtain a power generation output prediction result, obtaining real-time load change data, obtaining load fluctuation characteristics according to the real-time load change data, carrying out resource optimization distribution according to the load fluctuation characteristics and a power generation processing prediction result, obtaining a preliminary scheduling plan, and optimizing the preliminary scheduling plan through real-time feedback data of loads to obtain an optimized scheduling scheme.

Inventors

LUO YANJING
ZHAO ZHENG
DUAN JINGXUAN

Assignees

河北亮能售电有限公司

Dates

Publication Date: 20260508
Application Date: 20260109

Claims (10)

1. The virtual power plant dynamic scheduling method based on multi-source data fusion is characterized by comprising the following steps of: acquiring multi-source power data in the virtual power plant, and unifying the multi-source power data to obtain an initial data set; Constructing a feature matrix according to the initial data set and the historical power generation output data, and extracting key factors according to the feature matrix to obtain a characterization data matrix; Carrying out time sequence analysis on the characterization data matrix through a deep learning model to obtain a power generation output prediction result; Acquiring real-time load change data, and acquiring load fluctuation characteristics according to the real-time load change data; According to the load fluctuation characteristics and the power generation processing prediction results, carrying out resource optimization distribution through an optimization algorithm to obtain a preliminary scheduling plan; and optimizing the preliminary scheduling plan through real-time feedback data of the load to obtain an optimized scheduling scheme.
2. The method of claim 1, wherein the step of determining the position of the substrate comprises, The acquisition process of the initial data set comprises the following steps: acquiring multi-source power data, wherein the multi-source power data comprises power, power supply voltage and environment data, output data and transformer substation logs of different green power equipment; And carrying out standardized conversion and data unified processing on the multi-source power data to obtain an initial data set.
3. The method of claim 1, wherein the step of determining the position of the substrate comprises, The initial data set acquisition further comprises: the initial data set is preprocessed, wherein the preprocessing includes data cleaning, noise removal, and missing value replenishment.
4. The method of claim 1, wherein the step of determining the position of the substrate comprises, The acquisition process of the characterization data matrix comprises the following steps: The method comprises the steps of obtaining historical power generation output data, constructing an initial feature matrix according to the initial data set and the historical power generation output data, setting core factors, taking data of other dimensions as influencing factors, obtaining the association degree between the core factors and the influencing factors, judging the association degree, obtaining a simplified feature matrix, correcting the data in the simplified feature matrix according to historical trend, sorting the feature importance of the corrected simplified feature matrix through a machine learning model to obtain feature combinations with high priority, carrying out association judgment on different features in the feature combinations, obtaining the optimized feature matrix based on the association judgment result, and verifying the optimized feature matrix to obtain a feature data matrix.
5. The method of claim 1, wherein the step of determining the position of the substrate comprises, The process for obtaining the predicted result of the power generation output comprises the following steps: And processing the data under a plurality of time sequences in the characteristic data matrix through a long short-time memory network model to obtain a power generation output prediction result, wherein the characteristic data matrix comprises power generation output data and key influence factors on power generation processing data.
6. The method of claim 1, wherein the step of determining the position of the substrate comprises, The acquisition process of the load fluctuation characteristic comprises the following steps: Acquiring real-time load change data, carrying out data fusion on the real-time load change data and a power generation output prediction result to obtain a comprehensive load data set, obtaining dynamic load demands according to the comprehensive load data set, constructing a dynamic load demand curve according to the dynamic load demands, and obtaining load fluctuation characteristics according to the dynamic load demand curve, wherein the load fluctuation characteristics comprise mutation points and periodic change characteristics of the dynamic load demand curve.
7. The method of claim 1, wherein the step of determining the position of the substrate comprises, The obtaining process of the preliminary dispatch plan comprises the following steps: And carrying out resource optimization distribution through a genetic algorithm according to the load fluctuation characteristic and the power generation processing prediction result, wherein in the genetic algorithm, the target is that the weighted sum of the load fluctuation characteristic, the economic index and the reliability index is minimized, the load fluctuation characteristic is obtained according to the power generation output prediction result and the real-time load change data, the reliability index is obtained according to the relation between the power generation output prediction result and the power grid and the fault score, and the primary scheduling plan comprises the power supply proportion of the green power and the power grid.
8. The method of claim 1, wherein the step of determining the position of the substrate comprises, The acquisition process of the optimized scheduling scheme comprises the following steps: And acquiring real-time feedback data of the load, judging a threshold value of the real-time feedback data to obtain a load change trend, and adjusting the preliminary scheduling plan according to the load change trend.
9. The method of claim 1, wherein the step of determining the position of the substrate comprises, And optimizing in the virtual power plant according to the processing parameters in the generation process of the optimized scheduling scheme.
10. A virtual power plant dynamic scheduling system based on multi-source data fusion, for performing the method of any one of claims 1-9.

Description

Virtual power plant dynamic scheduling method based on multi-source data fusion Technical Field The invention belongs to the technical field of dynamic scheduling of virtual power plants, and particularly relates to a dynamic scheduling method of a virtual power plant based on multi-source data fusion. Background In the field of energy management and power dispatching, research on how to efficiently integrate resources and realize supply and demand balance has important significance. The field is directly related to the energy utilization efficiency and the stability of system operation, and is a core support for promoting green energy development and smart grid construction. With the wide application of renewable energy sources, the complexity of power systems is increasing, and the demands for accurate prediction, dynamic scheduling and real-time optimization are becoming urgent. However, many current solutions suffer from significant drawbacks in dealing with multi-source data and dynamic environments. The existing method is difficult to adapt to heterogeneous characteristics of data from different sources, so that deviation occurs in the information integration process, and further accuracy of follow-up prediction and scheduling is affected. In addition, when facing the rapid change of load demand, the existing scheme lacks the deep utilization of real-time feedback, and cannot adjust strategies in time to cope with emergency, so that resource waste or insufficient power supply is easily caused. In this context, this field faces significant technical challenges. First, how to effectively integrate data from multiple sources ensures the consistency and reliability of information. Due to the diversity and complexity of data sources, problems such as non-uniform format, non-synchronous time and the like are often encountered in the integration process, and the accurate prediction of the power generation output is directly influenced. The problem is further extended to another key difficulty, namely, how to dynamically formulate a reasonable scheduling scheme based on prediction by combining the change of the real-time load demand. If feedback adjustment cannot be performed in time according to the load state, the scheduling scheme is difficult to reach the optimal value, and even unstable system operation can be caused. Therefore, how to dynamically optimize the power generation scheduling scheme by combining the feedback of the real-time load state on the basis of multi-source data integration becomes a key problem to be solved in the invention. Disclosure of Invention In order to solve the technical problems, the invention provides a virtual power plant dynamic scheduling method based on multi-source data fusion, so as to solve the problems in the prior art. In order to achieve the above purpose, the present invention provides a virtual power plant dynamic scheduling method based on multi-source data fusion, comprising: acquiring multi-source power data in the virtual power plant, and unifying the multi-source power data to obtain an initial data set; Constructing a feature matrix according to the initial data set and the historical power generation output data, and extracting key factors according to the feature matrix to obtain a characterization data matrix; Carrying out time sequence analysis on the characterization data matrix through a deep learning model to obtain a power generation output prediction result; Acquiring real-time load change data, and acquiring load fluctuation characteristics according to the real-time load change data; According to the load fluctuation characteristics and the power generation processing prediction results, carrying out resource optimization distribution through an optimization algorithm to obtain a preliminary scheduling plan; and optimizing the preliminary scheduling plan through real-time feedback data of the load to obtain an optimized scheduling scheme. Optionally, the acquiring process of the initial data set includes: acquiring multi-source power data, wherein the multi-source power data comprises power, power supply voltage and environment data, output data and transformer substation logs of different green power equipment; And carrying out standardized conversion and data unified processing on the multi-source power data to obtain an initial data set. Optionally, after the initial data set is acquired, the method further includes: the initial data set is preprocessed, wherein the preprocessing includes data cleaning, noise removal, and missing value replenishment. Optionally, the acquiring process of the characterization data matrix includes: The method comprises the steps of obtaining historical power generation output data, constructing an initial feature matrix according to the initial data set and the historical power generation output data, setting core factors, taking data of other dimensions as influencing factors, obtaining the association degree b