CN-121980808-A - Super-multi-objective optimization method and system based on double-case co-evolution and double-distance indexes

CN121980808ACN 121980808 ACN121980808 ACN 121980808ACN-121980808-A

Abstract

The invention discloses a super-multi-objective optimization method and system based on double-file co-evolution and double-distance indexes, which comprises the steps of generating an initial population based on parameters of a power system, initializing parameters and a reference vector set, storing the population into a convergence file and a diversity file respectively, calculating ideal points and worst points based on the convergence file, updating the convergence file by combining the double-distance indexes, maintaining global distribution of the population through reference vector association operation, evaluating the population distribution state by combining information entropy to determine the local neighborhood size, calculating the local neighborhood density of an individual by utilizing parallel distances based on the local neighborhood distribution state, updating the diversity file, constructing a mating pool based on the double-file, generating a child population, and performing iteration until the non-dominant solution set in the diversity file is output. The invention enhances the convergence and diversity maintaining capability of the algorithm to the complex front edge, and can efficiently generate the power system scheduling scheme with multi-objective balance of power generation cost, transmission loss, line utilization rate and power generation adjustment.

Inventors

LI WEI
YANG NING
XU QING
LIU ZHITING
OuYang Sien
LI ZEYU

Assignees

江西理工大学

Dates

Publication Date: 20260505
Application Date: 20260204

Claims (9)

1. The super-multi-objective optimization method based on double-case co-evolution and double-distance indexes is characterized by comprising the following steps of: S1, generating an initial population based on parameters of a power system to be optimized, initializing algorithm parameters and a reference vector set, and storing the initial population into a convergence file and a diversity file respectively; S2, calculating ideal points and worst points based on the convergence file, and updating the convergence file by combining double-distance indexes; S3, carrying out association operation on individuals in the diversity archive and the reference vector set so as to maintain global distribution of the population; s4, evaluating the population distribution state of the diversity archives by combining information entropy, and determining the local neighborhood size according to the evaluation result; s5, calculating the local neighborhood density of the individual in the diversity file by using a parallel distance based on the local neighborhood size, and updating the diversity file based on the local neighborhood density; S6, constructing a mating pool based on the updated convergence file and the diversity file, and generating a child population through the mating pool; And S7, iteratively executing the steps S2-S6 until the termination condition is met, and outputting a solution set in the diversity file as an optimal scheduling scheme of the power system.
2. The double-case co-evolution and double-distance index-based super-multi-objective optimization method according to claim 1, wherein, In the step S2, ideal points and worst points are calculated based on the convergence file, and the convergence file is updated by combining the double-distance index, including: The method comprises the steps of calculating ideal points and worst points based on target values of individuals in a current population, calculating double-distance index values of each individual in the convergence file according to the ideal points and the worst points, screening the individuals in the convergence file based on the double-distance index values, and updating the convergence file.
3. The double-case co-evolution and double-distance index-based super-multi-objective optimization method according to claim 2, wherein, The calculation formula of the double-distance index in the S2 is as follows: ; Wherein, the Is an individual The Euclidean distance to the ideal point is calculated as: ; Is an individual The Euclidean distance to the worst point is calculated as: ; is in the current population Is set to be a minimum value of (c), Is in the current population Is set at the maximum value of (c), Is the first The ideal value of the individual target is set, Is the first The worst value of the individual targets is, Is the target number of times, Is an individual In the first place Objective function values on the individual objectives.
4. The double-case co-evolution and double-distance index-based super-multi-objective optimization method according to claim 1, wherein, And S4, evaluating the population distribution state of the diversity archives by combining information entropy, and determining the local neighborhood size according to an evaluation result, wherein the method comprises the following steps: And determining the local neighborhood size based on the ratio of the current population distribution entropy to the optimal distribution entropy.
5. The double-case co-evolution and double-distance index-based super-multi-objective optimization method according to claim 4, wherein, Calculating the current population distribution entropy And optimal distribution entropy The formula of (2) is: ; ; wherein L is the number of reference vectors, Is the first The number of reference vectors is chosen such that, Is the second and third of the diversity files Individual proportions of the association of the individual reference vectors, i.e , Is the reference vector The number of individuals to be associated with, Is the size of the files of the diversity, Is the probability value under ideal uniform distribution.
6. The double-case co-evolution and double-distance index-based super-multi-objective optimization method according to claim 4, wherein, Determining the local neighborhood size based on the ratio of the current population distribution entropy to the optimal distribution entropy, wherein the formula is as follows: ; Wherein, the In order to be a function of the rounding-off, As a function of the maximum value.
7. The double-case co-evolution and double-distance index-based super-multi-objective optimization method according to claim 1, wherein, And S5, calculating the local neighborhood density of the individuals in the diversity archive by using the parallel distance, wherein the method comprises the steps of calculating the parallel distance between the individuals based on the normalized fitness value of the individuals, and calculating the local neighborhood density of each individual based on the parallel distance and the local neighborhood size.
8. The double-case co-evolution and double-distance index-based super-multi-objective optimization method according to claim 1, wherein, The parameters of the power system to be optimized comprise the upper and lower limits of the output power and the cost coefficient of the generator set, the impedance and the capacity limit value of the power transmission line and the load demand of each node; The solution set is a non-dominant solution set on multiple targets of power generation cost, transmission loss, line utilization and power generation adjustment.
9. A dual-case co-evolution and dual-distance index-based ultra-multi-objective optimization system for implementing the method of any one of claims 1-8, comprising: the initialization module is used for generating an initial population based on parameters of the power system to be optimized, initializing algorithm parameters and a reference vector set, and storing the initial population into a convergence file and a diversity file respectively; the convergence file updating module is used for calculating ideal points and worst points based on the convergence file and updating the convergence file by combining the double-distance indexes; the global distributed maintenance module is used for carrying out association operation on individuals in the diversity archive and the reference vector set; the state evaluation module is used for evaluating the population distribution state of the diversity archives by combining information entropy and determining the local neighborhood size according to the evaluation result; the local diversity maintenance module is used for calculating the local neighborhood density of an individual in the diversity file by utilizing a parallel distance based on the local neighborhood size and updating the diversity file based on the local neighborhood density; The mating pool generation module is used for constructing a mating pool based on the updated convergence file and the diversity file, and generating a child population through the mating pool; and the iteration control module is used for controlling and iteratively executing the processing of the convergence file updating module, the global distributed maintenance module, the state evaluation module, the local diversity maintenance module and the mating pool generation module until the termination condition is met, and outputting a solution set in the diversity file as an optimal scheduling scheme of the power system.

Description

Super-multi-objective optimization method and system based on double-case co-evolution and double-distance indexes Technical Field The invention belongs to the technical field of power system optimization scheduling, and particularly relates to a super-multi-objective optimization method and system based on double-file co-evolution and double-distance indexes. Background In the field of power system optimization scheduling, power grid operators need to coordinate various physical devices such as a generator, a transmission line, a transformer and the like at the same time, and optimization targets generally comprise minimization of power generation cost, minimization of transmission loss, minimization of line utilization and minimization of power generation adjustment. There are substantial conflicts between these goals, for example, pursuing the lowest power generation cost may result in excessive line loads, while ensuring safe line operation may require increased power generation costs and increased unit adjustment frequency. This complex trade-off relationship between multiple objectives makes power system optimization scheduling essentially a complex super-multiple objective optimization problem. When the traditional single-objective optimization methods such as weighted summation and the like are used for processing the problems, the prior experience of a dispatcher is often relied on to convert the multi-objective problem into the single-objective problem, and a series of diversified weighing schemes are difficult to obtain by the method, so that the finally formulated dispatching scheme can not adapt to the running working condition of the real-time change of a power grid, and the economic running level and the safety stability margin of the system are affected. To address this challenge, researchers have proposed a variety of super-multi-objective evolutionary algorithms, mainly including three types of methods: (1) An algorithm based on improving the dominance relation. The selection pressure is enhanced by loosening the dominant conditions, but when the high-dimensional target space of the power system is processed, the search process is extremely easy to be premature due to the excessively strong selection pressure, a wide scheduling strategy cannot be explored, the finally obtained solution diversity is insufficient, the solution diversity is reflected to the actual power grid operation, namely, the solution diversity is single, and the solution diversity is difficult to cope with various load demands and operation conditions. (2) A decomposition algorithm based on the reference vector. The method guides the searching direction through a preset reference vector, but when the complex and irregular Pareto front edge common in the power system is processed, the fixed reference vector system is difficult to flexibly adapt to the actual operation constraint of the power distribution network, so that the algorithm convergence pressure is insufficient, the real high-efficiency operation interval of the system cannot be accurately approximated, and the generated scheduling scheme may not perform well at one end of economy or safety. (3) Algorithm based on performance index. The algorithm guides searching by using indexes, but a single fixed index easily causes the optimization process to be in local optimum, reflects the engineering, namely, the scheduling scheme is in a certain fixed operation mode, and cannot find innovative scheduling strategies which can obviously improve the comprehensive operation efficiency of the system, and the introduction of a plurality of indexes can obviously increase the calculation cost, so that the requirement of a large-scale power system on the real-time performance of the optimization calculation is difficult to meet. The existing double-file algorithm framework provides a basis for balancing convergence and diversity, but the main disadvantage of the framework in the application of the power system is that the convergence file updating strategy has insufficient adaptability to the complex front shape of the power grid, and a diversity file maintenance mechanism lacks a targeted design on the operation characteristics of the system. This limitation results in optimization results often biased towards locally optimal regions, making it difficult to generate a high quality set of scheduling schemes that both guarantee economy and meet safety constraints. Therefore, development of a novel optimization method capable of deeply fusing the operation characteristics of the power system and having better performances in terms of convergence speed and distribution quality is needed to meet the actual demands of the modern power grid on efficient and reliable scheduling decisions. Disclosure of Invention In order to solve the technical problems, the invention provides a super-multi-objective optimization method and system based on double-file co-evolution and double-distance i