CN-122022959-A - Virtual power plant bidding method and system based on personalized incentive demand response

CN122022959ACN 122022959 ACN122022959 ACN 122022959ACN-122022959-A

Abstract

The application discloses a virtual power plant bidding method and a virtual power plant bidding system based on personalized excitation demand response, and relates to the technical field of joint scheduling operation of energy sources. And optimizing the daily market objective function by using the maximized virtual power plant operation profit and the maximized comprehensive satisfaction degree of the demand response users, and constructing a daily-real-time two-stage bidding optimization model of the virtual power plant by using a rolling time domain optimization strategy and taking the total deviation adjustment cost between the minimum and daily bidding and the scheduling plan as a real-time market objective function to obtain the real-time bidding and scheduling plan. The application effectively improves the demand response participation depth, the virtual power plant operation profit and the market competitiveness, and simultaneously realizes the balance of the benefits of operators and clients.

Inventors

WANG JINFENG
MAO LUMING
CAO MIN
WEN DONG
JIA RONGRONG
SHI RONG
Lei Ruixue
HUANG ZONGJUN
DI HONGYU
REN ZHENGMOU
SUN XIAOCHEN
ZHANG YUHAO
Dong Yuezhao
JIANG YANJUN
ZHENG NAN

Assignees

国网陕西省电力有限公司经济技术研究院
国网陕西省电力有限公司营销服务中心(计量中心)
陕西电力交易中心有限公司
国网陕西省电力有限公司西咸新区供电公司

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (10)

1. The virtual power plant bidding method based on personalized incentive demand response is characterized by comprising the following steps of: Acquiring historical operation data of a virtual power plant and historical participation demand response conditions of demand response users; the historical operation data is subjected to data extraction to obtain renewable energy source output data and load demand data; generating a renewable energy output prediction curve and a load demand prediction curve according to renewable energy output data and load demand data by using a time sequence prediction model; According to the load shedding capacity and the bid price, adopting an unsupervised combined clustering algorithm based on density clustering and neighbor classification to perform cluster analysis on demand response users, and obtaining a cluster analysis result; according to the renewable energy output prediction curve, the load demand prediction curve and the clustering analysis result, constructing a daily market bidding and scheduling optimization model by taking the maximized virtual power plant operation profit and the maximized comprehensive satisfaction degree of the demand response user as daily market objective functions to obtain a daily bidding and scheduling plan; acquiring real-time ultra-short-term renewable energy output data, load demand data and real-time bidding price; And according to the real-time ultra-short-term renewable energy output data, the load demand data and the real-time bidding price, the total deviation adjustment cost between the minimum and the daily bidding and scheduling plans is used as a real-time market objective function, the daily market bidding and scheduling optimization model is subjected to real-time scheduling optimization through a rolling time domain optimization strategy, the daily-real-time market bidding and scheduling optimization model is obtained, and the real-time bidding and scheduling plans are obtained.
2. The virtual power plant bidding method based on personalized incentive demand response according to claim 1, wherein the specific content of the cluster analysis result is obtained by adopting an unsupervised combined cluster algorithm based on density clustering and neighbor classification to perform cluster analysis on demand response users according to the load shedding capability and the bidding price, and the method comprises the following steps: constructing a two-dimensional feature vector according to the load shedding capacity and the bid price to form a clustering data set; Processing the clustering data set by adopting an improved OPTICS algorithm, calculating the core distance and the reachable distance of each data point in the clustering data set, and generating a reachable distance map; Dividing the density connected domain by adaptively selecting cluster radius parameters according to the reachable distance graph, and aggregating the demand response users in the same density connected domain into the same category to obtain the category of the demand response users; and for the new demand response users, extracting the load shedding capacity and the bidding price of the new demand response users to form new two-dimensional feature vectors, classifying the new users into the demand response user categories by adopting a K nearest neighbor algorithm, and obtaining cluster analysis results, namely, the demand response users of each category.
3. The virtual power plant bidding method based on personalized incentive demand response of claim 2, wherein constructing a daily market bidding and scheduling optimization model with maximized virtual power plant operation profit and maximized comprehensive satisfaction of demand response users as daily market objective functions according to renewable energy output prediction curves, load demand prediction curves and cluster analysis results, and obtaining specific contents of daily bidding and scheduling plans comprises: Taking the comprehensive satisfaction degree of the maximized virtual power plant operation profit and the maximized demand response user as a daily market objective function, and combining a renewable energy output prediction curve, a load demand prediction curve and a cluster analysis result, and carrying out optimization solution by adopting an improved epsilon-constraint method under the constraint condition of meeting the power system operation constraint and the daily market objective function to obtain a pareto optimal solution set; And selecting an optimal scheduling scheme according to the pareto optimal solution set by utilizing a fuzzy clustering technology to obtain a daily market bidding and scheduling optimization model, wherein the optimal scheduling scheme is a daily bidding and scheduling plan.
4. The virtual power plant bidding method based on personalized incentive demand response according to claim 3, wherein the real-time market bidding and scheduling optimization model is obtained by real-time scheduling optimization of the day-ahead market bidding and scheduling optimization model through a rolling time domain optimization strategy with the total deviation adjustment cost of the minimum and day-ahead bidding and scheduling plans as a real-time market objective function according to real-time ultra-short-term renewable energy output data, load demand data and real-time bidding price, and the specific contents of the real-time bidding and scheduling plans are obtained comprise: According to the real-time ultra-short-term renewable energy output data, the load demand data and the real-time bidding price, rolling optimization is carried out in each window at 15min intervals, and a real-time ultra-short-term renewable energy output prediction curve and a load demand prediction curve are generated through a prediction model; Based on the day-ahead market bidding and scheduling optimization model, according to the real-time ultra-short-term renewable energy output prediction curve and the load demand prediction curve, the total deviation adjustment cost between the minimum day-ahead bidding and scheduling plan is used as a real-time market objective function, under the constraint condition that the real-time market objective function is met, the day-ahead bidding and scheduling plan is optimized, the transaction electric quantity of the day-ahead market and the real-time market is adjusted, the day-ahead real-time market bidding and scheduling optimization model is obtained, and the real-time bidding and scheduling plan is obtained.
5. A virtual power plant bidding method based on personalized incentive demand response as recited in claim 3, wherein the expression for maximizing the virtual power plant operation profit is: ; Wherein, the To maximize virtual power plant operating profits, t is the day-ahead schedule period index, N t is the total number of schedule periods, In order to respond to the amount of electricity sold to the consumer to the demand, In order to respond to the price of the electrical energy sold to the demand-responsive user, To sell the revenue generated by the electricity to demand response customers, For the amount of electricity sold to the market in the daytime, To sell the price of electrical energy to the market in the daytime, For revenue generated by the current market, For the decision variables of the electricity sold outwards, For the contract to violate the amount of electricity, To penalize the price of the penalty for the violation, The penalty revenues for the incentive type demand to respond to contract violations, In order to demand a response amount of electricity, In order to customize the stimulus to be applied, To customize incentive costs provided to different customer categories, For call decision variables of the demand response resource, For the operating cost of the micro gas turbine, Is a decision variable for starting and stopping the micro gas turbine unit, For the amount of electricity purchased from the market in the past, To purchase electricity from the market in the past, To cost of electricity purchased from the market in the past, For decision variables to purchase electricity from the market, For the cost of the renewable energy owner, Decision variables for renewable energy generation unit scheduling.
6. A virtual power plant bidding method based on personalized incentive demand response as recited in claim 3, wherein the expression for maximizing the comprehensive satisfaction of the demand response users is: ; Wherein, the To maximize the overall satisfaction of demand response users, t is the index of the day-ahead schedule period, N t is the total number of schedule periods, c is the demand response user category, N class is the total number of demand response user categories, i is the number of demand response users, N c is the total number of class c demand response users, The load of the i-th demand user in class c at time t is reduced, For the i-th demand user's customized incentive in class c at time t, Costs are not satisfied for the demand user.
7. The method of claim 5, wherein the minimizing the total deviation from the day-ahead bidding and scheduling plan adjusts costs by: ; Wherein, the To minimize the total bias from the day-ahead bidding and scheduling plans, the cost is adjusted, t is the day-ahead scheduling period index, To optimize the initial time of the scheduling period for the deep learning based rolling horizon, The length of the scheduling period is optimized for the deep learning based rolling horizon, To purchase the electricity for the real-time market, In order to purchase electricity prices in real-time markets, In order to purchase electricity costs from the real-time market, For the electricity bias cost of the market in the future and in real time, For the current deviation cost of the market purchase in the future and in real time, The offset costs are scheduled for the micro gas turbines, For renewable energy source output bias costs, The offset cost is scheduled for the demand response resource, For the real-time market to sell the electric quantity, For the real-time market selling price of electricity, Is the income of selling electricity to the real-time market.
8. The method of claim 4, wherein the constraint of the day-ahead market objective function and the constraint of the real-time market objective function each include an electric power supply-demand balance constraint, a distributed energy output constraint, an energy storage system operation constraint, a customer load shedding capability constraint, an incentive rate boundary constraint and a stakeholder benefit constraint.
9. The method of claim 4, wherein the daily bidding and scheduling schedule comprises a power bidding curve of the daily market of the virtual power plant, a scheduling schedule of internal resources, and an incentive rate of the user in response to the category demand; the real-time bidding and scheduling plan comprises an electric energy bidding curve of a virtual power plant in a real-time market and a real-time scheduling plan of internal resources; the internal resources comprise distributed power sources, energy storage systems and resources responding to demands.
10. A virtual power plant bidding system based on personalized incentive demand response, for implementing a virtual power plant bidding method based on personalized incentive demand response as claimed in any one of claims 1-9, comprising: the first data acquisition module is used for acquiring historical operation data of the virtual power plant and historical participation demand response conditions of demand response users; The data analysis module is used for extracting data from the historical operation data to obtain renewable energy output data and load demand data; the data prediction module is used for generating a renewable energy output prediction curve and a load demand prediction curve according to the renewable energy output data and the load demand data by utilizing the time sequence prediction model; The cluster analysis module is used for carrying out cluster analysis on the demand response users by adopting an unsupervised combined cluster algorithm based on density clustering and neighbor classification according to the load shedding capacity and the bid price to obtain a cluster analysis result; the day-ahead market module is used for constructing a day-ahead market bidding and scheduling optimization model according to the renewable energy output prediction curve, the load demand prediction curve and the clustering analysis result and by taking the maximized virtual power plant operation profit and the maximized comprehensive satisfaction degree of demand response users as objective functions, so as to obtain a day-ahead bidding and scheduling plan; The second data acquisition module is used for acquiring real-time ultra-short-term renewable energy output data, load demand data and real-time bidding price; And the real-time market module is used for carrying out real-time scheduling optimization on the day-ahead market bidding and scheduling optimization model by taking the total deviation adjustment cost between the minimum and day-ahead bidding and scheduling plans as an objective function according to the real-time ultra-short-term renewable energy output data, the load demand data and the real-time bidding price, so as to obtain a day-ahead real-time market bidding and scheduling optimization model and a real-time bidding and scheduling plan.

Description

Virtual power plant bidding method and system based on personalized incentive demand response Technical Field The application relates to the technical field of joint scheduling operation of energy sources, in particular to a virtual power plant bidding method and system based on personalized incentive demand response. Background With the development of large-scale access of renewable energy sources (Renewable Energy Source, RES) and internet of things (Internet of Things, ioT) technology, virtual power plants (Virtual Power Plant, VPP) can get rid of geographic space and power grid topology constraints, aggregate widely distributed resources for centralized management based on advanced communication technology, and realize optimal scheduling of the distributed resources by means of energy market transaction, so that the virtual power plants (Virtual Power Plant, VPP) have become research hotspots. However, with the continuous diversification of VPP aggregation bodies, how to effectively coordinate complementarity of these multi-source operation bodies, deeply dig and fully utilize flexible potential of massive distributed resources, and further improve market competitiveness thereof has become an important difficulty to be solved. Currently, operation of VPP faces two major core challenges, namely, inherent intermittence and strong uncertainty of RES (e.g., photovoltaic, wind power) output and load, leading to prediction difficulties, directly affecting accuracy and market risk of bidding strategies, and how to effectively stimulate internal Demand Response (DR) resources, particularly stimulating demand response (IDR), to provide flexible regulation capability. The prior VPP bidding strategy research has the defects that a single deep learning model is adopted in a prediction model, the processing capacity of a non-stable and nonlinear sequence is limited, the prediction precision is unstable in complex weather and market scenes, an IDR incentive mechanism generally adopts a fixed rate or a simple incremental rate, the diversity of different types of clients in terms of load reduction cost, response will and capability cannot be fully considered, the incentive efficiency is low, the flexibility of a demand side cannot be deeply excavated, and the profitability and market competitiveness of the VPP are limited. Disclosure of Invention The application aims to provide a virtual power plant bidding method and a system based on personalized incentive demand response, which solve the defects of a virtual power plant bidding strategy in the prior art in terms of user incentive and dynamic adjustment, realize differential incentive and dynamic scheduling and improve the market competitiveness and operation robustness of a virtual power plant. In order to achieve the above purpose, the application provides a virtual power plant bidding method based on personalized incentive demand response, which comprises the following steps: Acquiring historical operation data of a virtual power plant and historical participation demand response conditions of demand response users; the historical operation data is subjected to data extraction to obtain renewable energy source output data and load demand data; generating a renewable energy output prediction curve and a load demand prediction curve according to renewable energy output data and load demand data by using a time sequence prediction model; According to the load shedding capacity and the bid price, adopting an unsupervised combined clustering algorithm based on density clustering and neighbor classification to perform cluster analysis on demand response users, and obtaining a cluster analysis result; according to the renewable energy output prediction curve, the load demand prediction curve and the clustering analysis result, constructing a daily market bidding and scheduling optimization model by taking the maximized virtual power plant operation profit and the maximized comprehensive satisfaction degree of the demand response user as daily market objective functions to obtain a daily bidding and scheduling plan; acquiring real-time ultra-short-term renewable energy output data, load demand data and real-time bidding price; And according to the real-time ultra-short-term renewable energy output data, the load demand data and the real-time bidding price, the total deviation adjustment cost between the minimum and the daily bidding and scheduling plans is used as a real-time market objective function, the daily market bidding and scheduling optimization model is subjected to real-time scheduling optimization through a rolling time domain optimization strategy, the daily-real-time market bidding and scheduling optimization model is obtained, and the real-time bidding and scheduling plans are obtained. Preferably, according to the load shedding capability and the bid price, adopting an unsupervised combined clustering algorithm based on density clustering and neighbor clas