Search

CN-122000986-A - Distributed energy aggregation and prediction method for virtual power plant

CN122000986ACN 122000986 ACN122000986 ACN 122000986ACN-122000986-A

Abstract

The invention discloses a distributed energy aggregation and prediction method of a virtual power plant, which comprises the steps of establishing a dynamic aggregation weight distribution mechanism aiming at high-priority resource combination, dynamically adjusting weight coefficients of resources in an aggregation process according to real-time weather conditions and load demand prediction results, calculating to obtain an overall output prediction value of the virtual power plant through a weighted fusion algorithm, generating a corresponding uncertainty interval range, and recalculating an overall operation state evaluation index of the virtual power plant through corrected aggregation mechanism parameters, wherein the parameters comprise key performance parameters such as prediction precision, adjustment capability, economic benefit and the like, and establishing an operation state database record history optimization result to form a continuously improved intelligent distributed energy aggregation and prediction system.

Inventors

  • MA JIE
  • YANG YU
  • LIU YAGUANG

Assignees

  • 河北亮能售电有限公司

Dates

Publication Date
20260508
Application Date
20251218

Claims (10)

  1. 1. The distributed energy aggregation and prediction method for the virtual power plant is characterized by comprising the following steps of: acquiring historical operation data of distributed energy resources, and processing the historical operation data through a multidimensional feature extraction algorithm and a time sequence analysis method to obtain a characteristic difference matrix; Calculating a time sequence complementary coefficient according to the correlation degree among the resources of the characteristic difference matrix, obtaining an intra-group complementary intensity value through a Pearson correlation coefficient method, and determining a high-priority resource combination according to the intra-group complementary intensity value; Establishing a dynamic aggregation weight distribution mechanism aiming at high-priority resource combinations, dynamically adjusting weight coefficients of all resources in an aggregation process according to real-time weather conditions and load demand prediction results, calculating to obtain a virtual power plant overall output prediction value through a weighted fusion algorithm, and generating a corresponding uncertainty interval range; After the integral output predicted value is obtained, a multi-objective optimization model is established, an optimal resource allocation scheme is obtained by solving by adopting a particle swarm optimization algorithm with the maximum economy and the optimal running stability as objective functions; Formulating a real-time scheduling instruction according to the optimal resource allocation scheme, and enabling the running state of the virtual power plant to always meet the power grid scheduling requirement through a closed-loop feedback control mechanism; continuously monitoring deviation conditions of actual operation data and predicted values of each resource by adopting a sliding time window method, and automatically correcting aggregation parameters when the accumulated deviation exceeds a preset error threshold value to obtain corrected aggregation mechanism parameters for the next round of prediction calculation; And recalculating the overall running state evaluation index of the virtual power plant through the corrected aggregation mechanism parameters, and establishing a running state database record history optimization result to obtain the continuously improved intelligent distributed energy aggregation and prediction system.
  2. 2. The method of claim 1, wherein the obtaining a characteristic difference matrix comprises: Acquiring historical operation data of distributed energy resources, extracting output characteristics from the historical operation data by adopting a multidimensional characteristic extraction algorithm to obtain an initial characteristic data set, wherein the historical operation data comprises wind power, photovoltaic, energy storage and controllable load; processing the initial characteristic data set by a time sequence analysis method, and extracting a power fluctuation amplitude, a response time constant and an adjustment capacity range to obtain a standardized output characteristic parameter; carrying out normalization processing on the output characteristic parameters by adopting a normalization processing method, and eliminating dimension differences to obtain a normalization characteristic data set; And constructing a characteristic difference matrix according to the standardized characteristic data set, and calculating the output characteristic difference among the distributed energy resources to obtain the characteristic difference matrix.
  3. 3. The method of claim 1, wherein the determining a high priority resource combination comprises: extracting key feature vectors of each distributed energy resource by adopting a principal component analysis method according to the characteristic difference matrix to obtain a feature data set after dimension reduction; grouping wind power, photovoltaic, energy storage and controllable load through a k-means clustering algorithm according to the feature data set after dimension reduction to obtain a resource collaborative clustering result; Aiming at the resource collaborative clustering result, calculating the time sequence complementary coefficient of the resources in each group, and obtaining the complementary intensity value in the group by adopting a Pearson correlation coefficient method; if the complementary strength value in the group exceeds a preset threshold value, calculating the matching strength of the energy storage in the group and the controllable load through a gray correlation analysis method to obtain an intra-group matching coefficient; according to the intra-group matching coefficients, fusing complementary intensity values and the matching coefficients in the group by adopting a weighted average method to obtain comprehensive synergy scores of the groups; and sequencing the cooperative priorities of the groups by comprehensive cooperative grading, and determining the high-priority resource combination.
  4. 4. The method of claim 1, wherein the calculating a virtual power plant overall output predicted value and generating a corresponding uncertainty interval range comprises: acquiring output data of wind power and photovoltaic according to real-time meteorological conditions, and acquiring output variation trend of each resource by adopting a time sequence analysis method; If the output change trend exceeds a preset fluctuation threshold, short-term characteristic data are extracted through a sliding window method, and a resource output characteristic vector is obtained; According to the resource output characteristic vector, predicting future output values of all resources by adopting a support vector regression algorithm to obtain a single resource output prediction result; Acquiring a real-time load demand curve of the virtual power plant through a load demand prediction result, and determining a load demand fluctuation range; If the deviation between the load demand fluctuation range and the single resource output prediction result exceeds a preset threshold value, a linear interpolation method is adopted to adjust the dynamic weight coefficient of each resource, and an optimized weight distribution scheme is obtained; and calculating the whole output predicted value of the virtual power plant by adopting a weighted fusion algorithm according to the optimized weight distribution scheme to obtain a comprehensive output predicted result.
  5. 5. The method of claim 1, wherein the obtaining the optimal resource allocation scheme comprises: After the integral output predicted value is obtained, a multi-objective optimization model is constructed, an objective function is defined as being optimal in economical maximization and operation stability, constraint conditions are set as technical parameter limit of each resource and safe operation requirement of a power grid, and a particle swarm optimization algorithm is adopted to calculate and obtain optimal output distribution proportion and resource allocation scheme of each resource; Acquiring real-time output data of each resource according to the optimal output distribution proportion of each resource, extracting the output change trend of each resource through time sequence analysis, and judging whether the output of each resource meets the technical parameter limit; If the resource output exceeds the technical parameter limit, the output distribution proportion is adjusted by a linear interpolation method, and a corrected output distribution scheme is obtained; Acquiring real-time running state data of the virtual power plant according to the corrected output distribution scheme, analyzing the uncertainty of each resource output distribution scheme by a Monte Carlo simulation method, and determining the uncertainty interval range of the running state; Acquiring real-time load demand data of a power grid according to an uncertainty interval range of an operation state, and judging whether load demand fluctuation exceeds a preset threshold value; If the load demand fluctuation exceeds a preset threshold, adjusting the output distribution proportion of each resource by a weighted average method to obtain an optimized resource scheduling scheme; Acquiring real-time output monitoring data of each resource according to the optimized resource scheduling scheme, extracting short-term output characteristic vectors through a sliding window method, and judging whether the output of each resource meets the safety operation requirement of a power grid; if the resource output does not meet the power grid safe operation requirement, recalculating the output distribution proportion of each resource by a secondary planning method to obtain an adjusted scheduling scheme; According to the adjusted scheduling scheme, comprehensive operation data of the virtual power plant are obtained, the integral operation state characteristics of the virtual power plant are generated through a data fusion method, and the operation stability index of the virtual power plant is determined; and acquiring economic evaluation data according to the operation stability index of the virtual power plant, and calculating the economic index of the virtual power plant by a cost accounting method to obtain a comprehensive optimized operation scheme.
  6. 6. The method of claim 1, wherein said enabling the virtual power plant operating state to always meet the grid scheduling requirement via a closed loop feedback control mechanism comprises: Acquiring real-time operation data of a virtual power plant, and generating a real-time state feature vector of the virtual power plant through a data fusion technology; according to the real-time state feature vector, a preset load demand prediction model is adopted, and a load demand prediction value in a future period of time is calculated; after the load demand predicted value is obtained, the load demand rising amplitude is calculated by comparing the load demand predicted value with the historical data; if the load demand rising amplitude exceeds a preset threshold value, determining the calling sequence of the energy storage resources through a priority ordering algorithm; generating a scheduling instruction of the energy storage resource according to the calling sequence of the energy storage resource, and rapidly responding to the load demand change; Acquiring real-time wind power output data, and calculating a wind power output prediction deviation rate through comparison with a wind power output prediction model; If the wind power output prediction deviation rate is greater than a preset threshold value, determining participation of controllable loads through a controllable load management platform; generating a scheduling instruction of the controllable load according to the participation degree of the controllable load, and performing balance adjustment; acquiring a real-time running state of the virtual power plant, and adjusting a resource scheduling instruction through a closed-loop feedback control algorithm; And updating the running state of the virtual power plant according to the adjusted resource scheduling instruction, so as to ensure that the power grid scheduling requirement is met.
  7. 7. The method of claim 1, wherein performing the aggregation parameter modification comprises: Continuously acquiring operation data of each resource by adopting a sliding time window method, and cleaning the acquired data by a data preprocessing module to obtain a standardized operation data set; calculating the deviation value of the actual operation data of each resource and the predicted value through the comparison of the standardized operation data set and the predicted model to obtain a deviation data set; If the accumulated deviation value in the deviation data set exceeds a preset error threshold, identifying a deviation source through a deviation analysis module, and determining a resource characteristic parameter to be corrected; and updating the parameter weight by adopting an online learning algorithm according to the determined resource characteristic parameters to obtain a modified aggregation parameter set.
  8. 8. The method of claim 1, wherein the continuously improved intelligent distributed energy aggregation and prediction system comprises: Recalculating the overall running state evaluation index of the virtual power plant through the corrected aggregation mechanism parameters; Analyzing the running state change trend of the virtual power plant through the state change feature set, and extracting key influence factors by adopting a principal component analysis algorithm to obtain a key factor set; If the factor value in the key factor set exceeds a preset threshold, determining an operation parameter to be optimized through a factor screening module, and generating an optimized parameter set; Adjusting a resource allocation strategy of the virtual power plant according to the optimized parameter set, and updating the resource scheduling weight by adopting a linear regression algorithm to obtain an updated scheduling weight set; generating a real-time resource scheduling instruction through the updated scheduling weight set, outputting the real-time resource scheduling instruction to a virtual power plant control system, and acquiring a scheduling execution data set; If the execution deviation in the dispatching execution data set exceeds a preset threshold value, identifying a deviation source through a deviation analysis module, and determining a control parameter set to be corrected; Updating the virtual power plant running state model according to the corrected control parameter set, and optimizing model parameters by adopting a support vector machine algorithm to obtain an optimized running state model; And generating a predicted value sequence of the virtual power plant through the optimized running state model, outputting the predicted value sequence to a running state database, and updating a historical optimizing data set.
  9. 9. A computer device comprising a memory, a processor and a computer program stored on the memory, characterized in that the processor executes the computer program to carry out the steps of the method according to any one of claims 1-8.
  10. 10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any of claims 1-8.

Description

Distributed energy aggregation and prediction method for virtual power plant Technical Field The invention belongs to the technical field of power integration, and particularly relates to a distributed energy aggregation and prediction method for a virtual power plant. Background With the deep advancement of global energy transformation, virtual power plants have become a core support for building new electric power systems as a key technical platform for integrating distributed energy resources. The virtual power plant performs unified scheduling and management on various distributed energy resources such as wind power, photovoltaic, energy storage, controllable load and the like, so that the flexibility and stability of a power grid can be obviously improved, and the technical development level of the virtual power plant directly influences the overall operation efficiency of an energy system. The current distributed energy aggregation method mainly adopts a simple superposition or averaging treatment mode, and lacks in-depth consideration of the characteristic difference of different energy resources. The traditional methods often neglect the inherent relevance among various energy resources, so that the aggregation result cannot fully embody the real running state of the system, and the cooperative advantage of the distributed energy is difficult to develop. In actual operation of a virtual power plant, different types of distributed energy resources have distinct output characteristics and response patterns, and such characteristic differences can create complex interactions during the aggregation process. Because of the lack of an effective interaction mechanism identification method, the prior art is difficult to accurately capture key information such as output complementary rules between wind power and photovoltaic, coordination and cooperation modes of energy storage and controllable load and the like. The ambiguity of the interaction mechanism further leads to the lack of an aggregation optimization strategy, so that the system cannot conduct intelligent resource allocation adjustment according to the synergistic effect of different energy resources. When the aggregation process lacks the guidance of a targeted optimization strategy, the whole virtual power plant system cannot realize the optimal combination of various distributed energy resources, and the economical efficiency and the reliability of the system operation are severely restricted. Disclosure of Invention The invention provides a distributed energy aggregation and prediction method for a virtual power plant, which aims to solve the problems in the prior art. In order to achieve the above purpose, the invention provides a distributed energy aggregation and prediction method for a virtual power plant, which comprises the following steps: acquiring historical operation data of distributed energy resources, and processing the historical operation data through a multidimensional feature extraction algorithm and a time sequence analysis method to obtain a characteristic difference matrix; Calculating a time sequence complementary coefficient according to the correlation degree among the resources of the characteristic difference matrix, obtaining an intra-group complementary intensity value through a Pearson correlation coefficient method, and determining a high-priority resource combination according to the intra-group complementary intensity value; Establishing a dynamic aggregation weight distribution mechanism aiming at high-priority resource combinations, dynamically adjusting weight coefficients of all resources in an aggregation process according to real-time weather conditions and load demand prediction results, calculating to obtain a virtual power plant overall output prediction value through a weighted fusion algorithm, and generating a corresponding uncertainty interval range; After the integral output predicted value is obtained, a multi-objective optimization model is established, an optimal resource allocation scheme is obtained by solving by adopting a particle swarm optimization algorithm with the maximum economy and the optimal running stability as objective functions; Formulating a real-time scheduling instruction according to the optimal resource allocation scheme, and enabling the running state of the virtual power plant to always meet the power grid scheduling requirement through a closed-loop feedback control mechanism; continuously monitoring deviation conditions of actual operation data and predicted values of each resource by adopting a sliding time window method, and automatically correcting aggregation parameters when the accumulated deviation exceeds a preset error threshold value to obtain corrected aggregation mechanism parameters for the next round of prediction calculation; And recalculating the overall running state evaluation index of the virtual power plant through the corrected aggregation mechanism pa