CN-122021314-A - Load prediction and cooperative adjustment method for virtual power plant

CN122021314ACN 122021314 ACN122021314 ACN 122021314ACN-122021314-A

Abstract

The application discloses a load prediction and collaborative adjustment method of a virtual power plant, which belongs to the technical field of power systems and energy Internet, and comprises the steps of collecting multi-source heterogeneous data in an aggregation range of the virtual power plant, preprocessing and characteristic engineering processing the data, constructing a characteristic set and a system operation state which represent a historical operation rule and a current operation state of the virtual power plant, carrying out rolling prediction on load demands, distributed energy output and adjustable resource capacity in a plurality of time scales in the future based on a data-driven prediction model, generating a net load prediction curve and a resource adjustment boundary, constructing a multi-target adjustment optimization model by combining electric market signals and operation constraint, generating an adjustment decision scheme of the virtual power plant in a preset scheduling period, and carrying out optimization updating on the prediction model and the adjustment model by executing feedback, so as to realize continuous optimization of the operation of the virtual power plant. The economical efficiency, the power grid friendliness and the user participation of the virtual power plant can be enhanced.

Inventors

TIAN ZEHAO
Zhang Maile
GAO JIANXIANG
Yue Benyong
WU XIAOFENG

Assignees

西安邮电大学
陕西交控绿色发展集团有限公司

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (8)

1. A method for load prediction and coordinated regulation of a virtual power plant, the method comprising: S1, collecting multi-source heterogeneous data in an aggregation range of a virtual power plant, and preprocessing the multi-source heterogeneous data to form a unified data set; S2, carrying out feature engineering processing on the unified data set, generating a feature set of load prediction and adjustment decision, and constructing a system running state of the virtual power plant at the current moment based on real-time state information data of the virtual power plant at the current moment; S3, based on the feature set and the system running state, calling a data-driven prediction model, performing rolling prediction on running conditions in a plurality of time scales in the future, outputting a prediction result, and generating a net load prediction curve and a resource adjustment boundary of the virtual power plant in a prediction time domain according to the prediction result; And S4, constructing an adjustment optimization model of the virtual power plant based on the net load prediction curve and the resource adjustment boundary and combining the power market signal and the operation constraint, and generating an adjustment decision scheme of the virtual power plant in a preset scheduling period by solving the adjustment optimization model by taking economic benefit maximization, power grid operation index optimization and user satisfaction as optimization targets.
2. The method for load prediction and co-regulation of a virtual power plant of claim 1, further comprising: The adjustment decision scheme is issued to a corresponding device control unit or a user interaction terminal through a communication network so as to execute the adjustment decision scheme, and the operation result of the virtual power plant is monitored in real time in the execution process, so that the actual response data of each device and the user response data from the user interaction terminal are obtained; The actual response data and the user response data are used as the actual values of model training, the execution effect of the adjustment decision scheme is evaluated, and feedback information representing economic benefits, power grid influence and user response deviation is generated; and based on the feedback information, optimizing and updating the data-driven prediction model and the adjustment optimization model.
3. The method for load prediction and collaborative adjustment of a virtual power plant according to claim 1, characterized by performing feature engineering processing on the unified data set, comprising: And carrying out statistical analysis, time correlation analysis and data transformation processing on the unified data set, extracting multi-dimensional characteristic information representing load change rules, environmental influence factors and user electricity behavior characteristics, and carrying out screening, combination and standardization processing on the multi-dimensional characteristic information to form a characteristic set for load prediction and adjustment decision, wherein the characteristic set comprises time sequence characteristics representing load time sequence change characteristics, environmental characteristics representing environmental and external condition influences and behavior characteristics representing the user electricity behavior characteristics.
4. The method for load prediction and collaborative adjustment of a virtual power plant according to claim 1, wherein the constructing a system operating state of the virtual power plant at a current time comprises: The method comprises the steps of acquiring real-time state information data of various devices and resources in a virtual power plant through a real-time data acquisition system, wherein the real-time state information data comprise current power grid power, charging states of an energy storage system, output power of distributed energy sources, load demand data and power grid frequency, performing time sequence alignment and cleaning pretreatment on the real-time state information data, and determining a system running state of the virtual power plant at the current moment by combining preset running constraint conditions and the pretreated real-time state information data, wherein the system state comprises current total load, current available capacity of adjustable resources, output conditions and adjustable potential of the distributed energy sources and charging/discharging capacity of the energy storage system.
5. The method for load prediction and collaborative adjustment of a virtual power plant according to claim 1, wherein the rolling prediction of operating conditions over a plurality of future time scales comprises: The method comprises the steps of integrating a characteristic set representing a historical operation rule with a system operation state representing a virtual power plant operation working condition at the current moment to serve as input of a prediction model, respectively executing prediction calculation on a plurality of continuous time intervals in the future according to a set prediction time scale, wherein different time scales correspond to different prediction granularities and prediction time domains, updating or recalling the prediction model based on the current system operation state in each prediction period to generate a load demand prediction result, a distributed energy output prediction result and an adjustable resource capacity prediction result in the corresponding time scale, and correcting subsequent predictions by using the prediction result and actual operation data of the previous period in the rolling propulsion process of the prediction period, so that a rolling prediction result set which is dynamically updated along with time is formed, and outputting the prediction result.
6. The method of claim 1, wherein constructing the tuning optimization model of the virtual power plant based on the net load prediction curve and resource tuning boundaries in combination with power market signals and operating constraints comprises: In time sequence within a predetermined scheduling period In order to optimize a time domain, the output or the adjustment quantity of each adjustable resource in each time period is taken as a decision variable, and the adjustment optimization model is constructed, wherein the optimization targets comprise: the economic benefit maximization is used for representing the difference between the transaction benefit and the operation cost of the virtual power plant in the electric power market; the grid operation index optimization is used for restraining or minimizing power deviation, load fluctuation of a virtual power plant grid connection point or adverse influence on safe operation of the grid; the user satisfaction is used for restraining or minimizing the influence on user comfort or energy consumption requirements caused by load adjustment; Objective function of the optimization objective Expressed as: Wherein, the The economic benefit objective function is represented by the term, Representing the relevant objective function of the operation index of the power grid, Representing a user satisfaction-related objective function, For the corresponding weight coefficient(s), Is a time period; and applying constraint conditions of resource output constraint based on the resource adjustment boundary, operation constraint based on the physical characteristics of equipment, power balance constraint based on the net load prediction curve and constraint conditions based on user energy demand or service protocol in the optimization process; And generating an adjustment decision scheme of each adjustable resource of the virtual power plant in the preset scheduling period by solving the adjustment optimization model.
7. The method of claim 1, wherein the heterogeneous multi-source data comprises historical and real-time data related to power loads, operational status data of distributed energy and energy storage systems, meteorological data, date type data, data related to power markets and grid operation.
8. The method for predicting and co-regulating loads of a virtual power plant according to claim 6, wherein the regulation decision scheme comprises a power setting state or an operating state of each controllable power generation or energy storage unit, a regulation command or an excitation signal of each flexible load.

Description

Load prediction and cooperative adjustment method for virtual power plant Technical Field The application relates to the technical field of power systems and energy Internet, in particular to a load prediction and cooperative regulation method of a virtual power plant. Background With the large-scale access of distributed energy sources, energy storage systems and flexible loads, the traditional power system operation mode mainly comprising a centralized power source faces the problems of increased scheduling complexity, increased operation uncertainty and the like. Virtual power plants are taken as a technical form for aggregating and uniformly scheduling dispersed energy resources through information communication and control technologies, and become an important means for improving new energy consumption capability and flexibility of an electric power system. In the operation of the existing virtual power plant, load prediction is usually modeled based on historical load data, multisource information such as distributed energy output, user behavior and real-time operation state is difficult to fully fuse, prediction accuracy is limited, meanwhile, partial adjustment strategies focus on a single time scale or a single optimization target, comprehensive consideration of electric power market signals, power grid operation constraint and user response characteristics is lacked, and collaborative optimization adjustment among multiple resources is difficult to realize. In addition, the existing method is insufficient in feedback utilization after adjustment execution, lacks a dynamic correction and continuous learning mechanism, and influences the stability and economy of long-term operation of the virtual power plant. Therefore, a virtual power plant load prediction and collaborative adjustment method capable of fusing multi-source data, having multi-time scale prediction capability and realizing prediction, adjustment and feedback closed-loop optimization is needed. Disclosure of Invention Aiming at the technical problems in the background technology, the invention provides a load prediction and cooperative adjustment method for a virtual power plant. In order to solve the technical problems, the technical scheme of the invention is as follows: a method of load prediction and coordinated regulation of a virtual power plant, the method comprising: S1, collecting multi-source heterogeneous data in an aggregation range of a virtual power plant, and preprocessing the multi-source heterogeneous data to form a unified data set; S2, carrying out feature engineering processing on the unified data set, generating a feature set of load prediction and adjustment decision, and constructing a system running state of the virtual power plant at the current moment based on real-time state information data of the virtual power plant at the current moment; S3, based on the feature set and the system running state, calling a data-driven prediction model, performing rolling prediction on running conditions in a plurality of time scales in the future, outputting a prediction result, and generating a net load prediction curve and a resource adjustment boundary of the virtual power plant in a prediction time domain according to the prediction result; And S4, constructing an adjustment optimization model of the virtual power plant based on the net load prediction curve and the resource adjustment boundary and combining the power market signal and the operation constraint, and generating an adjustment decision scheme of the virtual power plant in a preset scheduling period by solving the adjustment optimization model by taking economic benefit maximization, power grid operation index optimization and user satisfaction as optimization targets. Further, the method further comprises: The adjustment decision scheme is issued to a corresponding device control unit or a user interaction terminal through a communication network so as to execute the adjustment decision scheme, and the operation result of the virtual power plant is monitored in real time in the execution process, so that the actual response data of each device and the user response data from the user interaction terminal are obtained; The actual response data and the user response data are used as the actual values of model training, the execution effect of the adjustment decision scheme is evaluated, and feedback information representing economic benefits, power grid influence and user response deviation is generated; and based on the feedback information, optimizing and updating the data-driven prediction model and the adjustment optimization model. Further, performing feature engineering processing on the unified data set, including: And carrying out statistical analysis, time correlation analysis and data transformation processing on the unified data set, extracting multi-dimensional characteristic information representing load change rules, environmental influence fact