CN-122026503-A - Multi-mode resource collaborative optimization scheduling method, system and equipment for virtual power plant

CN122026503ACN 122026503 ACN122026503 ACN 122026503ACN-122026503-A

Abstract

The invention discloses a multi-mode resource collaborative optimization scheduling method, a system and equipment of a virtual power plant, and relates to the relevant technical field of virtual power plant optimization scheduling, comprising the steps of uniformly modeling and dynamically aggregating diversified distributed resources in the virtual power plant to construct a standardized resource pool; the method comprises the steps of generating a prediction sequence of various future scenes based on external market and environment information, forming a state observation space, making decisions by utilizing a deep reinforcement learning agent, synchronously generating a real-time scheduling instruction and a combined bidding strategy, executing the scheduling instruction, submitting a market bid, and carrying out continuous iteration and optimization according to actual market response and environment feedback. The technical problems of insufficient multi-mode resource coordination, response assessment and scheduling disconnection and multi-market coordination mechanism deletion in the prior art are solved, and the technical effects of improving the economical efficiency, reliability and market competitiveness of the operation of the virtual power plant by generating a coordinated scheduling instruction and multi-market combined bidding strategy through the deep learning agent are achieved.

Inventors

Zou chan
Xiong Pengyu
ZHONG YONGHUA
LI DEZHONG
YANG FAN
HU WEIMING
JIANG FANGQIN
YOU YING
GU HONGYAN

Assignees

湖南大唐先一科技有限公司

Dates

Publication Date: 20260512
Application Date: 20251210

Claims (9)

1. The multi-mode resource collaborative optimization scheduling method of the virtual power plant is characterized by comprising the following steps of: uniformly modeling and dynamically aggregating the multi-mode distributed resources in the virtual power plant to generate a standardized resource pool for representing aggregate response capability; generating a prediction sequence of a plurality of future scenes based on external market and environment information; taking the state of the standardized resource pool and the prediction sequence as state observation spaces, making a decision through a deep reinforcement learning agent, and outputting a scheduling instruction and a combined bidding strategy for a plurality of electric power markets; and issuing the scheduling instruction to the bottom physical equipment for execution, submitting the combined bidding strategy to a plurality of electric power markets at the same time, and performing continuous iterative optimization of the strategy according to the response.
2. The method for collaborative optimization scheduling of multi-modal resources for a virtual power plant according to claim 1, wherein the method for unified modeling and dynamic aggregation of multi-modal distributed resources within the virtual power plant to generate a normalized resource pool characterizing aggregate response capability comprises: Modeling independent response capability of multi-source heterogeneous resources in a virtual power plant, constructing a unified feature extraction model based on a graph neural network, and extracting space-time dynamic features of the multi-source heterogeneous resources; Establishing a unified evaluation index system, weighting each index in the evaluation index system by adopting an entropy method based on the space-time dynamic characteristics, and mapping and aggregating the multi-source heterogeneous resources into a resource pool with three standard types which can be interrupted, translated and adjusted according to different weight combinations to generate the standardized resource pool.
3. The method for collaborative optimization scheduling of multi-modal resources for a virtual power plant according to claim 2, wherein the evaluation metric system includes response time, duration, upper and lower power limits, ramp rate, average response time.
4. The method for collaborative optimization scheduling of multi-modal resources in a virtual power plant according to claim 1, wherein the deep reinforcement learning agent adopts a deep Q network algorithm, and the action space comprises power setting and adjusting instructions for multi-modal distributed resources and bidding strategies for a plurality of electric markets; The reward function of the deep reinforcement learning agent is in a multi-objective optimization form, and the multi-objective comprises market total benefits, punishment violations, operation cost and conditional risk value, wherein the market total benefits are the sum of electric energy market benefits, auxiliary service market benefits and carbon transaction benefits, and the conditional risk value is used for quantifying tail risks under multi-market coupling.
5. The method for collaborative optimization scheduling of multi-modal resources for a virtual power plant according to claim 4, wherein the conditional risk value is used to measure multi-market coupled risk, including a desired revenue term and a risk adjustment term.
6. The method for collaborative optimization scheduling of multi-modal resources for a virtual power plant according to claim 1, wherein issuing the scheduling instructions to an underlying physical device for execution while submitting the joint bidding strategy to a plurality of power markets, performing continuous iterative optimization of the strategy based on the responses, comprises: each market is cleared based on master-slave games, wherein a virtual power plant is used as a leader to report in advance, and an electric energy market, an auxiliary service market and a carbon transaction market are used as followers to respectively clear the market and return to clear prices; The virtual power plant forms an experience tuple together with the state and the action in the deep reinforcement learning intelligent agent according to the returned price information, the acquired actual resource output and the user response behavior information, and stores the experience tuple in an experience playback buffer area; and updating the network parameters of the deep reinforcement learning agent based on the experience tuples in the experience playback buffer zone, and performing continuous iterative optimization of the strategy.
7. The method for collaborative optimization scheduling of multi-modal resources for a virtual power plant according to claim 6, wherein master-slave gaming further includes physical constraints, market constraints, and time-coupled constraints; Wherein the physical constraints include power balance constraints, reserve capacity constraints, and carbon emission constraints, and the market constraints include electric energy-auxiliary service capacity coupling constraints and cross-market arbitrage constraints.
8. A system for collaborative optimal scheduling of multi-modal resources for a virtual power plant, the system being configured to implement the method for collaborative optimal scheduling of multi-modal resources for a virtual power plant of any one of claims 1 to 7, the system comprising: The electric field resource aggregation module is used for carrying out unified modeling and dynamic aggregation on the multi-mode distributed resources in the virtual power plant to generate a standardized resource pool for representing the aggregation response capability; The prediction sequence generation module is used for generating prediction sequences of various future scenes based on external market and environment information; The bidding strategy output module is used for taking the state of the standardized resource pool and the prediction sequence as state observation spaces, making decisions through a deep reinforcement learning agent, and outputting scheduling instructions and a combined bidding strategy for a plurality of electric power markets; And the strategy optimization module is used for issuing the scheduling instruction to the bottom physical equipment for execution, submitting the combined bidding strategy to a plurality of electric power markets, and performing continuous iterative optimization of the strategy according to the response.
9. An electronic device, the electronic device comprising: a memory for storing executable instructions; the processor is configured to implement the multi-mode resource collaborative optimization scheduling method of the virtual power plant according to any one of claims 1 to 7 when executing the executable instructions stored in the memory.

Description

Multi-mode resource collaborative optimization scheduling method, system and equipment for virtual power plant Technical Field The invention relates to the technical field of virtual power plant optimization scheduling, in particular to a method, a system and equipment for collaborative optimization scheduling of multi-mode resources of a virtual power plant. Background The output of the distributed resources has the characteristics of intermittence, volatility, dispersivity and the like, brings great challenges to the real-time balance, safety and stability and economic operation of the power system, and the virtual power plant has remarkable potential in the aspects of improving the flexibility of a power grid, promoting the consumption of new energy, participating in the power market and the like as an important means for aggregating the distributed energy resources, flexible loads and energy storage systems. At present, the scheduling and market participation of the virtual power plant are mainly concentrated on single type resources or markets, such as charge-discharge optimization for an energy storage system or bidding strategies facing an electric market, however, as the types of distributed resources are increasingly diversified, a plurality of modes such as photovoltaics, wind power, energy storage, electric automobiles and flexible loads are covered, the cooperative complementary potential of the traditional scheduling is difficult to fully mine, meanwhile, an electric market mechanism is increasingly complex, multiple types of markets such as an energy market and an auxiliary service market coexist, the bidding strategies of the market are difficult to maximize the operation income of the virtual power plant, and in addition, the external environment and market conditions have obvious uncertainty, such as renewable energy output, electric price fluctuation, load change and the like, so that the virtual power plant is required to have the capability of efficiently predicting and dynamically responding to multi-source information. Therefore, in the related technology at the present stage, the technical problems of insufficient multi-mode resource coordination, disjoint response evaluation and scheduling and missing multi-market coordination mechanism exist. Disclosure of Invention The multi-mode resource collaborative optimization scheduling method, the system and the equipment for the virtual power plant solve the technical problems of insufficient multi-mode resource collaborative, response evaluation and scheduling disjoint and multi-market collaborative mechanism deficiency in the prior art, and achieve the technical effects of generating collaborative scheduling instructions and multi-market joint bidding strategies through deep learning agents and improving the economical efficiency, reliability and market competitiveness of the operation of the virtual power plant. The application provides a multi-mode resource collaborative optimization scheduling method of a virtual power plant, which comprises the steps of carrying out unified modeling and dynamic aggregation on multi-mode distributed resources in the virtual power plant to generate a standardized resource pool representing aggregate response capability, generating a prediction sequence of various future scenes based on external market and environment information, carrying out decision making by taking the state of the standardized resource pool and the prediction sequence as state observation spaces through deep reinforcement learning agents, outputting scheduling instructions and a combined bidding strategy oriented to a plurality of electric power markets, issuing the scheduling instructions to bottom physical equipment for execution, submitting the combined bidding strategy to the plurality of electric power markets, and carrying out continuous iterative optimization of the strategy according to the response. In a possible implementation manner, the multi-mode resource collaborative optimization scheduling method of the virtual power plant further performs the following processing of modeling the independent response capability of the multi-source heterogeneous resources in the virtual power plant, constructing a unified feature extraction model based on a graph neural network, extracting the space-time dynamic features of the multi-source heterogeneous resources, establishing a unified evaluation index system, weighting each index in the evaluation index system by adopting an entropy method based on the space-time dynamic features, mapping and aggregating the multi-source heterogeneous resources into an interruptible, translatable and adjustable three standard type resource pool according to different weight combinations, and generating the standardized resource pool. In a possible implementation manner, the multi-mode resource collaborative optimization scheduling method of the virtual power plant further performs the following pro