Search

CN-122020374-A - Electric power spot shipment model construction method based on multiple intelligent agents and ADMM

CN122020374ACN 122020374 ACN122020374 ACN 122020374ACN-122020374-A

Abstract

The invention provides an electric power spot delivery clear model construction method based on multiple intelligent agents and ADMM, which comprises the steps of constructing a multiple intelligent agent framework, conducting independent optimization on each intelligent agent according to preset cost parameters and risk quantification models, conducting scene reduction by adopting Monte Carlo simulation to generate a scene of wind and light output based on probability distribution characteristics of wind and light output, extracting high probability typical scenes to serve as optimization input, conducting distributed iterative optimization on SCUC and SCED models of grid coordination intelligent agents by utilizing an ADMM algorithm, wherein the grid coordination intelligent agents calculate clear price vectors according to the quantitative quotation information of each intelligent agent, embedding risk quantification results into objective functions of thermal power and energy storage intelligent agents, and conducting iterative update on dual variables of the ADMM algorithm until supply and demand balance deviation and clear price deviation are smaller than preset thresholds, completing convergence calculation of the clear models, and outputting optimal quotation strategies of each market main body.

Inventors

  • GUO MENGJIE
  • PENG YI
  • NI HUI
  • XING TIANLONG

Assignees

  • 北京清大科越股份有限公司

Dates

Publication Date
20260512
Application Date
20260122

Claims (10)

  1. 1. The utility model provides a power spot delivery model construction method based on multiple agents and ADMM, which is characterized by comprising the following steps: S1, constructing a multi-agent framework comprising a wind power agent, a photovoltaic agent, a thermal power agent, an energy storage agent, a load agent and a power grid coordination agent, wherein each agent is independently optimized according to preset cost parameters and a risk quantization model; S2, based on probability distribution characteristics of wind and light output, generating a wind and light output scene by adopting Monte Carlo simulation, performing scene reduction through a clustering algorithm, and extracting a high-probability typical scene as an optimized input; S3, performing distributed iterative optimization on SCUC and SCED models of the power grid coordination intelligent agents by using an ADMM algorithm, wherein the power grid coordination intelligent agents calculate clear electricity price vectors according to the amount quotation information of each intelligent agent, and embed risk quantization results into objective functions of the thermal power and energy storage intelligent agents; s4, iteratively updating the dual variables of the ADMM algorithm until the supply and demand balance deviation and the power price deviation are smaller than the preset threshold value, completing convergence calculation of the price model, and outputting the optimal quotation strategy of each market subject.
  2. 2. The method of claim 1, wherein S1 comprises: S11, optimizing objective functions of the wind power intelligent agent and the photovoltaic power intelligent agent respectively comprise maximized income items And minimizing cost terms , wherein, And Refers to the electricity market price and the new energy generating capacity at the moment t, Initial construction investment cost and depreciation, financial cost and operation and maintenance cost; S12, optimizing an objective function of the thermal power intelligent agent to comprise maximizing a benefit item And minimizing cost terms , wherein, Refers to the output of the thermal power at the moment t, The electricity price of the thermal power is indicated; 、 And The method comprises the steps of a thermal power unit cost quadratic term coefficient, a cost primary term coefficient and no-load cost.
  3. 3. The method of claim 1, wherein S2 comprises: s21, adopting a Weibull distribution function for wind speed probability distribution , Modeling is performed, wherein, For wind speed frequency, k is a shape parameter, c is a scale parameter, For the average wind speed, Is a gamma function; S22, calculating Euclidean distance through K-means clustering algorithm in scene cut , wherein, Is that P is the number of attributes, Belongs to one data set, and clusters the scene of wind and light output into K typical scenes.
  4. 4. The method of claim 1, wherein S3 further comprises: S31, the SCUC model contains system load balance constraint , wherein, Representing the output of the genset i in the province during the period t, Representing the planned power of the tie-line j during period t, NT is the total number of tie-lines, System load for period t; s32, the SCED model comprises new energy station output constraint Wherein E is a new energy station set, The new energy station i predicts the output in the period t.
  5. 5. The method of claim 1, wherein S4 comprises: S41, adopting a dual variable iteration formula , The updating is performed, wherein, Representing the Lagrangian multiplier vector in the ADMM optimization algorithm for agent i in iteration k+1, t is the 96-point period, Penalty factors for supply and demand bias; s42, the supply and demand deviation calculation formula is , Wherein And Representing the generated power vector and the load power vector of node i in the kth iteration respectively, And representing a supply and demand deviation vector of the node i in the kth iteration, wherein t is a 96-point period.
  6. 6. An electric power spot shipment clear model construction device based on multiple agents and ADMM, characterized by comprising: the multi-agent framework construction module is used for constructing a multi-agent framework comprising a wind power agent, a photovoltaic agent, a thermal power agent, an energy storage agent, a load agent and a power grid coordination agent, and each agent is independently optimized according to preset cost parameters and a risk quantification model; the wind-light output scene generation and reduction module is used for generating a wind-light output scene by adopting Monte Carlo simulation based on the probability distribution characteristic of wind-light output, reducing the scene by a clustering algorithm, and extracting a high-probability typical scene as an optimized input; the ADMM distributed iterative optimization module is used for carrying out distributed iterative optimization on SCUC and SCED models of the power grid coordination intelligent agents by utilizing an ADMM algorithm, wherein the power grid coordination intelligent agents calculate clear electricity price vectors according to the quantitative quotation information of each intelligent agent, and the risk quantization results are embedded into objective functions of the thermal power and the energy storage intelligent agents; and the dual variable iteration convergence module is used for iteratively updating the dual variables of the ADMM algorithm until the supply and demand balance deviation and the clearing electricity price deviation are smaller than a preset threshold value, completing convergence calculation of the clearing model and outputting the optimal quotation strategy of each market main body.
  7. 7. The apparatus of claim 6, wherein the multi-agent framework building module is further to: the optimized objective functions of the wind power intelligent agent and the photovoltaic power intelligent agent respectively comprise the maximum benefit item And minimizing cost terms , wherein, And Refers to the electricity market price and the new energy generating capacity at the moment t, Initial construction investment cost and depreciation, financial cost and operation and maintenance cost; the optimized objective function of the thermal power agent comprises maximizing the benefit item And minimizing cost terms , wherein, Refers to the output of the thermal power at the moment t, The electricity price of the thermal power is indicated; 、 And The method comprises the steps of a thermal power unit cost quadratic term coefficient, a cost primary term coefficient and no-load cost.
  8. 8. The apparatus of claim 6, wherein the scene generation and reduction module is further configured to: The probability distribution of wind speed adopts a Weibull distribution function , Modeling is performed, wherein, For wind speed frequency, k is a shape parameter, c is a scale parameter, For the average wind speed, Is a gamma function; scene cut calculation of Euclidean distance by K-means clustering algorithm , wherein, Is that P is the number of attributes, Belongs to one data set, and clusters the scene of wind and light output into K typical scenes.
  9. 9. A computer device comprising a processor and a memory; Wherein the processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory for implementing the multi-agent and ADMM-based power off-the-shelf model building method according to any one of claims 1-5.
  10. 10. A non-transitory computer-readable storage medium having stored thereon a computer program, wherein the program when executed by a processor implements the multi-agent and ADMM-based power off-the-shelf model building method of any of claims 1-5.

Description

Electric power spot shipment model construction method based on multiple intelligent agents and ADMM Technical Field The invention belongs to the technical field of power systems, and particularly relates to a power spot delivery model construction method based on multiple intelligent agents and ADMM. Background The power system optimization and market clearing model is used as a core support technology of a novel power system and is widely applied to market mechanism design in a new energy large-scale grid-connected scene. Along with the promotion of 'double carbon' targets, the permeability of wind-solar intermittent power supplies is continuously improved, and the traditional centralized clearing model taking thermal power as a main component faces reconstruction challenges. Specifically, the prior art system covers the whole process from load prediction to unit combination, including key links such as SCUC safety restraint unit combination, SCED safety restraint economic dispatch and the like, but has the core defect that a new energy uncertainty modeling and market main body risk quantification mechanism cannot be effectively fused. Disclosure of Invention The present invention aims to solve at least one of the technical problems in the related art to some extent. Therefore, a first object of the present invention is to provide a method for constructing an electric power spot model based on multiple agents and ADMM. A second object of the present invention is to provide an electric power spot model building apparatus based on multiple agents and ADMM. A third object of the invention is to propose a computer device. A fourth object of the present invention is to propose a non-transitory computer readable storage medium. To achieve the above objective, an embodiment of a first aspect of the present invention provides a method for constructing a power spot-delivery model based on multiple agents and ADMM, including: S1, constructing a multi-agent framework comprising a wind power agent, a photovoltaic agent, a thermal power agent, an energy storage agent, a load agent and a power grid coordination agent, wherein each agent is independently optimized according to preset cost parameters and a risk quantization model; S2, based on probability distribution characteristics of wind and light output, generating a wind and light output scene by adopting Monte Carlo simulation, performing scene reduction through a clustering algorithm, and extracting a high-probability typical scene as an optimized input; S3, performing distributed iterative optimization on SCUC and SCED models of the power grid coordination intelligent agents by using an ADMM algorithm, wherein the power grid coordination intelligent agents calculate clear electricity price vectors according to the amount quotation information of each intelligent agent, and embed risk quantization results into objective functions of the thermal power and energy storage intelligent agents; s4, iteratively updating the dual variables of the ADMM algorithm until the supply and demand balance deviation and the power price deviation are smaller than the preset threshold value, completing convergence calculation of the price model, and outputting the optimal quotation strategy of each market subject. In one embodiment of the present invention, the S1 includes: S11, optimizing objective functions of the wind power intelligent agent and the photovoltaic power intelligent agent respectively comprise maximized income items And minimizing cost terms, wherein,AndRefers to the electricity market price and the new energy generating capacity at the moment t,Initial construction investment cost and depreciation, financial cost and operation and maintenance cost; S12, optimizing an objective function of the thermal power intelligent agent to comprise maximizing a benefit item And minimizing cost terms, wherein,Refers to the output of the thermal power at the moment t,The electricity price of the thermal power is indicated;、 And The method comprises the steps of a thermal power unit cost quadratic term coefficient, a cost primary term coefficient and no-load cost. In one embodiment of the present invention, the S2 includes: s21, adopting a Weibull distribution function for wind speed probability distribution ,Modeling is performed, wherein,For wind speed frequency, k is a shape parameter, c is a scale parameter,For the average wind speed,Is a gamma function; S22, calculating Euclidean distance through K-means clustering algorithm in scene cut , wherein,Is thatP is the number of attributes,Belongs to one data set, and clusters the scene of wind and light output into K typical scenes. In one embodiment of the present invention, the step S3 further includes: S31, the SCUC model contains system load balance constraint , wherein,Representing the output of the genset i in the province during the period t,Representing the planned power of the tie-line j during period t, NT is the total n