Optimal Method for Equipment Capacity of Integrated Energy System Based on Improved Particle Swarm Optimization Algorithm

Abstract

The present invention relates to a method for optimizing the equipment capacity of an integrated energy system based on an improved particle swarm optimization algorithm, including: establishing an integrated energy system model composed of photovoltaic, wind power, CCHP, and electrical energy storage devices; determining the objective function and constraint conditions of the integrated energy system model; constructing an improved dynamic multi-swarm particle swarm optimization algorithm without velocity terms based on reverse learning and elite promotion; using the improved dynamic multi-swarm particle swarm optimization algorithm without velocity terms, combining with the objective function and constraint conditions of the integrated energy system model, to complete the solution of the equipment capacity configuration of the integrated energy system and obtain the optimal capacity configuration scheme. Compared with the prior art, by improving the particle swarm optimization algorithm, the present invention can not only effectively simplify the position update process of particles, but also avoid particles missing the global optimal solution, so as to quickly, accurately, and stably solve the optimal scheme for the equipment capacity configuration of the integrated energy system.

Description

Technical Field

[0001] The present invention relates to the technical field of comprehensive energy system equipment capacity optimization, and particularly to a comprehensive energy system equipment capacity optimization method based on an improved particle swarm algorithm. Background Technique

[0002] Traditional energy systems mainly adopt an independent energy supply mode, and the multi-energy collaborative comprehensive energy system introducing clean energy has become a new development direction. Due to the high randomness of the energy supply of renewable energy equipment, this makes the energy system with multi-energy coupling and complementarity more complex, and carrying out system equipment capacity optimization has now become a popular research topic. Many engineering optimization problems in reality can be abstracted into the mathematical expression of a multi-peak function, and the particle swarm optimization algorithm is a stochastic optimization method based on social behavior simulation, proposed by Kenney and Eberhart in 1995. The main idea of the particle swarm optimization algorithm is to initialize a group of random solutions for the system, and search for the optimal solution of the optimization problem through iteration. With its advantages such as simple structure and easy implementation, the particle swarm algorithm is currently widely used to solve optimization problems.

[0003] The PSO (Particle Swarm Optimization) algorithm originates from the bird flock predation model. The position of each particle represents a solution to the current optimization problem. All particles determine the exploration direction and distance through velocity, and the velocity will be adjusted in the solution space according to the individual historical optimal position and the population historical optimal position. In the process of each iterative evolution, the individual historical best position of the particle is updated through the "survival of the fittest rule". And the global best particle Gbest.

[0004] However, the existing PSO algorithm has problems such as weak search ability, poor solution dispersion, unstable convergence speed, and easy to fall into local optimum when dealing with complex and high-dimensional optimization problems. Summary of the Invention

[0005] The purpose of the present invention is to overcome the above-mentioned defects existing in the prior art and provide a comprehensive energy system equipment capacity optimization method based on an improved particle swarm algorithm. Through a dynamic multi-population particle swarm algorithm without velocity terms based on reverse learning and elite promotion, the equipment capacity configuration of the comprehensive energy system is optimized, and the purpose of fast, accurate, and stable solution optimization can be achieved.

[0006] The purpose of the present invention can be achieved through the following technical solutions: A comprehensive energy system equipment capacity optimization method based on an improved particle swarm algorithm, including the following steps:

[0007] S1. Establish a comprehensive energy system model composed of photovoltaic, wind power, CCHP (Combined Cooling Heating and Power), and electrical energy storage equipment;

[0008] S2. Determine the objective function and constraints of the comprehensive energy system model;

[0009] S3. Based on reverse learning and elite promotion, construct an improved dynamic multi-swarm particle swarm optimization algorithm without velocity terms;

[0010] S4. Use the improved dynamic multi-swarm particle swarm optimization algorithm without velocity terms, combined with the objective function and constraints of the comprehensive energy system model, to complete the solution of the equipment capacity configuration of the comprehensive energy system and obtain the optimal capacity configuration plan.

[0011] Further, the objective function of the comprehensive energy system model is specifically the minimization of the initial investment cost.

[0012] Further, the constraints of the comprehensive energy system model include equipment operation power constraints and electrical, cooling, and heating power constraints.

[0013] Further, the specific working process of the improved dynamic multi-swarm particle swarm optimization algorithm without velocity terms is as follows:

[0014] Initialize the population;

[0015] Dynamically divide the population to obtain different subgroups;

[0016] For different subgroups, execute corresponding evolutionary strategies respectively;

[0017] Update the individual extreme values, merge the subpopulations, and update the population extreme value;

[0018] Execute the elite promotion evolutionary strategy for the individual extreme values;

[0019] Execute the differential evolution strategy for the population extreme value;

[0020] Judge whether the termination condition is reached. If the termination condition is reached, output the optimal solution; otherwise, return and continue the iteration.

[0021] Further, the initialization of the population specifically includes: initializing the particle positions, the initial individual extreme value is the current particle itself, and the global extreme value in the current population is determined through the feasibility rule.

[0022] Further, the different subgroups include the first subgroup and other subgroups.

[0023] Further, the evolutionary strategy corresponding to the first subgroup is:

[0024]

[0025] Among them, is the position of the i-th particle in the (t + 1)-th generation population, represents the individual historical best position of the i-th particle in the t-th generation population, c1 and c2 are learning factors, specifically non-negative constants, and r1 and r2 are random numbers between [0, 1]. is the new solution obtained by the particle X i through generalized opposition-based learning. is the new solution of the t-th generation population obtained by the particle X i through generalized opposition-based learning, and γ is a random number between [0, 1]. and are the minimum and maximum boundary values of the solution space in the d-th dimension respectively. is in the solution space of the d-th dimension, and X i,d is the particle X i in the solution space of the d-th dimension.

[0026] Furthermore, the evolutionary strategy corresponding to the other subgroups is:

[0027]

[0028] Among them, is the position of the i-th particle in the k-th subgroup of the (t + 1)-th generation population, c1, c2, and c3 are learning factors, non-negative constants, and r1, r2, and r3 are random numbers between [0, 1]. is the individual historical best position of the particle . is the individual historical best position of the randomly selected particle , and Gbest t is the global best particle of the t-th generation population. is the population best particle of the (k - 1)-th subgroup of the t-th generation population.

[0029] Furthermore, the elite promotion evolutionary strategy is specifically:

[0030]

[0031] Among them, is the new individual historical best solution optimized by the particle X i through the elite promotion strategy, and Pbest q , Pbest w , Pbest e are the individual best solutions randomly selected from the population respectively, α is a random number between [0, 1], and d represents the d-th dimension.

[0032] Further, the termination condition is specifically that the current iteration number is greater than or equal to the set maximum iteration number.

[0033] Compared with the prior art, aiming at the problems of poor accuracy and slow convergence speed existing in the existing particle swarm algorithm when solving the equipment capacity configuration of the integrated energy system, based on the particle swarm without velocity term, by dynamically dividing the population and adopting different evolutionary strategies for different subgroups, in the way of reverse learning and information interaction between subgroups, the loss of population diversity and premature convergence of the algorithm during the search process are prevented; for the individual historical optimal particles, the elite promotion evolutionary strategy is adopted to improve the quality of the individual optimal particles, thereby improving the search accuracy of the algorithm; for the global optimal particles of the population, the differential evolution strategy is adopted to enhance the local refined search ability of the algorithm, further improving the optimization accuracy and stability of the algorithm. Therefore, the optimal integrated energy system capacity configuration scheme can be solved quickly, accurately and stably. Description of the Drawings

[0034] Figure 1 It is a schematic flow chart of the method of the present invention;

[0035] Figure 2 It is a schematic diagram of the working process of the improved particle swarm algorithm in the present invention. Detailed Embodiments

[0036] The present invention will be described in detail below with reference to the drawings and specific embodiments.

[0037] Embodiment

[0038] As Figure 1 shown, an integrated energy system equipment capacity optimization method based on an improved particle swarm algorithm includes the following steps:

[0039] S1. Establish an integrated energy system model composed of a photovoltaic power generation, a wind power generation, a CCHP, and an electric energy storage device;

[0040] S2. Determine the objective function and constraint conditions of the integrated energy system model. Among them, the objective function of the integrated energy system model is specifically to minimize the initial investment cost;

[0041] The constraint conditions of the integrated energy system model include equipment operation power constraints and electric, heat, and cold power constraints;

[0042] S3. Based on reverse learning and elite promotion, construct an improved dynamic multi-population particle swarm algorithm without velocity term. Among them, the specific working process of the improved dynamic multi-population particle swarm algorithm without velocity term is as Figure 2 shown, including:

[0043] Initialize the population;

[0044] Dynamically divide the population to obtain different subgroups;

[0045] Execute corresponding evolutionary strategies for different subgroups respectively;

[0046] Update the individual extreme value, merge the sub-populations, and update the population extreme value;

[0047] Execute the elite promotion evolutionary strategy for the individual extreme value;

[0048] Execute the differential evolution strategy for the population extreme value;

[0049] Judge whether the termination condition is reached. If the termination condition has been reached, output the optimal solution; otherwise, return to continue the iteration;

[0050] S4. Use the improved dynamic multi-population particle swarm optimization algorithm without velocity terms, combine the objective function and constraint conditions of the integrated energy system model, complete the solution of the equipment capacity configuration of the integrated energy system, and obtain the optimal capacity configuration plan.

[0051] Apply the above technical solution to the actual situation. The specific process is as follows:

[0052] First, perform system modeling on each device of the integrated energy system, estimate the relationship between the investment cost and power of each device, collect the illumination, wind power, cooling, heating and electricity load data of the typical day in the area where the system is located, and analyze the output of each device under the typical day conditions;

[0053] After that, with the minimum initial investment cost as the objective function, under the conditions of operating power constraints and the balance constraints of cooling, heating and electricity loads, use the improved particle swarm optimization algorithm to solve the optimal capacity configuration of the system.

[0054] This technical solution adopts the regional dynamic population division strategy to simulate the dynamic immigration and emigration strategy of the population, and uses the niche concept in the theory of evolution to enhance the population diversity. The implementation method is as follows: Randomly generate a particle in the solution space, select the K particles with the closest Euclidean distance to this particle (K is the number of particles in the sub-population) to generate the sub-population k. By adopting different evolutionary strategies for different sub-groups to ensure the population diversity and avoid the algorithm falling into local optimum; In addition, the reverse learning method is also used to update the particle position, effectively broadening the population search space; The elite promotion strategy is adopted to guide the particle exploration, which helps to improve the development and exploration ability of the algorithm.

[0055] Thus, a dynamic multi-population particle swarm optimization algorithm without velocity terms based on reverse learning and elite promotion is realized. This algorithm is based on the particle swarm optimization without velocity terms, and the exploration of particles is based on the individual historical best position Based on the random linear combination between the individual historical best particle pbest of each particle and the global best particle Gbest, the parameter settings of the algorithm are minimized. Before each evolutionary iteration of the population, the population is divided according to the regional dynamic population division strategy; different evolutionary strategies are adopted for different subgroups; for the individual historical best particle, an elite promotion strategy is adopted for optimization, and for the population historical best particle, a differential evolution strategy is adopted, where:

[0056] 1) The position update formula of the particle swarm without velocity term is expressed as follows:

[0057]

[0058] In the formula, c1 and c2 are learning factors, which are non-negative constants and can adjust the particle's self-cognition ability and group-cognition ability. When the learning factor is larger, the particle will accelerate the search for the target area, but it is easy to miss the global optimal solution; when its value is smaller, the particle can search within a smaller target area range, but it is easy to fall into the local optimal solution. r1 and r2 are random numbers between [0, 1]. is the individual historical best position, and Gbest is the global best particle.

[0059] 2) The multi-population PSO algorithm is a local PSO algorithm based on a special neighborhood topology structure, which helps to improve the efficiency of individual information exchange. A biological population generally refers to individuals of the same species living in a certain area. The dynamic population division technology aims to simulate the immigration and emigration phenomena in the biological population to promote the evolution of particle position updates.

[0060] 3) Opposition-based learning (OBL), as an effective mechanism to broaden the search space, simultaneously searches the current solution and its opposite solution, thereby quickly expanding the solution space and approaching the global optimal solution faster.

[0061] 4) Due to the limited search ability and information utilization rate of particles, algorithms using a single search operator can only perform well on some problems. The information contained in elite particles can not only guide particle exploration but also help improve the exploitation and exploration capabilities of the algorithm. The present invention uses an elite promotion strategy. By using two new search operators and utilizing the useful information contained in other population members, the individual best position of each particle is further evolved. The idea of the elite promotion strategy is: make full use of the historical best particles of other individuals, and at the same time consider that opposition-based learning helps to avoid the population falling into the local optimal point. The implementation process is as follows: If a certain probability is satisfied, the following formula is executed for the individual optimal particle:

[0062]

[0063] In the formula, represents particle X iThe historical optimal solution of the new individual optimized by the elite promotion strategy; Pbest q , Pbest w , Pbest e represents the individual optimal solution randomly selected from the population; α is a random number between [0, 1]; d represents the d-th dimension.

[0064] 5) The characteristic of the differential evolution strategy is that it introduces a differential mutation mode for iterative search, providing good global optimization performance and stability. The mutation vector Z d is formed as follows:

[0065] Z d = Gbest d + F(Pbest r1，d - Pbest r2，d )

[0066] In the formula, Gbest is the optimal particle of the population; Pbest r1 and Pbest r2 represent randomly selected two individual extreme values; F is the scaling factor. The finest granularity in the crossover stage is each dimension of the vector. In setting the crossover condition, if the random number rand ≤ C R then the crossover operation is performed. C R is the preset crossover probability. At the same time, to ensure that the trial particle U is different from the original vector Gbest, a number rn i randomly selected from the set {1, 2,..., D} is introduced. When d is equal to rn i , the crossover condition is satisfied. Z i,d represents the i-th dimension of the mutation vector Z d . The crossover stage is operated according to the following formula:

[0067]

[0068] The differential evolution algorithm is used to update the optimal solution of the population, making the algorithm have good global optimization performance and stability.

[0069] To sum up, this technical solution is based on the particle swarm optimization algorithm, dynamic population division, reverse learning, elite promotion strategy and differential evolution technology, and proposes an improved particle swarm optimization algorithm, which can effectively solve multi-dimensional complex function optimization problems, has better algorithm search accuracy than other comparison algorithms, and can ensure good algorithm stability and convergence speed. It effectively solves the problems such as poor convergence accuracy and easy to fall into local optimum that are prone to occur in the practical application of the particle swarm algorithm, and applies this algorithm to the optimization of the equipment capacity of the integrated energy system, and can quickly, accurately and stably solve the optimal capacity configuration scheme.

[0070] In this embodiment, first, the particle positions are initialized. The extreme value of the initial generation individuals is the current particle itself, and the global extreme value (i.e., the population extreme value) in the current population is determined through the feasibility rules.

[0071] After that, the population is dynamically divided. A particle is randomly generated in the solution space, and the K particles with the closest Euclidean distance to this particle are selected to generate the sub-population k, where K is the number of particles in the sub-population.

[0072] Then, corresponding evolutionary strategies are executed for different populations. Among them, the first population updates its position according to the following formula:

[0073]

[0074] In the formula, represents the position of the i-th particle in the (t + 1)-th generation population; represents the individual historical optimal position of the i-th particle in the t-th generation population, denoted as c1 and c2 represent learning factors, which are non-negative constants. r1 and r2 are random numbers between [0, 1]. For particle X i , the new solution obtained through generalized opposition-based learning for particle X i is the new solution of the t-th generation population for particle X obtained through generalized opposition-based learning. The solution space of the new solution is represented as follows:

[0075]

[0076] In the formula, γ is a random number between [0, 1], and represent the minimum and maximum boundary values of the solution space in the d-th dimension, is in the solution space of the d-th dimension, and X i,d is the solution space of particle X i in the d-th dimension.

[0077] The particles of other sub-populations update their positions according to the following formula:

[0078]

[0079] In the formula, is the position of the i-th particle in the k-th sub-group of the (t + 1)-th generation population. c1, c2, and c3 are learning factors and are non-negative constants. r1, r2, and r3 are random numbers between [0, 1], is the individual historical optimal position of particle , is the individual historical optimal position of the randomly selected particle , and Gbest t is the globally best particle of the t-th generation population, It is the population best particle of the (k - 1)-th subpopulation of the t-th generation population.

[0080] The historical optimal solution of the new individual optimized by the elite promotion strategy.

[0081] Update the individual extreme value, merge the subpopulations and update the population extreme value.

[0082] Execute the elite promotion evolution strategy on the individual extreme value.

[0083] Execute the differential evolution strategy on the global extreme value (i.e., the population extreme value).

[0084] Judge whether the termination condition is reached. If the termination condition has been reached (the current iteration number is greater than or equal to the set maximum iteration number), then output the optimal solution. Otherwise, return and continue the iteration.

Claims

1. An equipment capacity optimization method for integrated energy systems based on an improved particle swarm algorithm, characterized in that, It includes the following steps: S1. Establish a comprehensive energy system model composed of photovoltaic, wind power, CCHP, and electrical energy storage devices; S2. Determine the objective function and constraint conditions of the comprehensive energy system model; S3. Based on reverse learning and elite promotion, construct an improved dynamic multi-population particle swarm optimization algorithm without velocity terms; S4. Use the improved dynamic multi-population particle swarm optimization algorithm without velocity terms, combined with the objective function and constraint conditions of the comprehensive energy system model, to complete the solution of the equipment capacity configuration of the comprehensive energy system and obtain the optimal capacity configuration plan; The specific working process of the improved dynamic multi-population particle swarm optimization algorithm without velocity terms is as follows: Initialize the population; Dynamically divide the population to obtain different sub-populations, including the first sub-population and other sub-populations; For different sub-populations, execute corresponding evolutionary strategies respectively; Update the individual extreme values, merge the sub-populations, and update the population extreme value; Execute the elite promotion evolutionary strategy for the individual extreme values; Execute the differential evolution strategy for the population extreme value; Judge whether the termination condition is reached. If the termination condition is reached, output the optimal solution; otherwise, return and continue the iteration; The evolutionary strategy corresponding to the first sub-population is: Among them, is the position of the i-th particle in the (t + 1)-th generation population, represents the individual historical best position of the i-th particle in the t-th generation population. c1 and c2 are learning factors, specifically non-negative constants, and r1 and r2 are random numbers between [0, 1]. is the new solution obtained by the particle X i through generalized opposition-based learning. is the new solution of the t-th generation population obtained by the particle X i through generalized opposition-based learning. γ is a random number between [0, 1]. and are the minimum and maximum boundary values of the solution space in the d-th dimension, respectively. is in the solution space of the d-th dimension, and X i,d is the particle X i in the solution space of the d-th dimension; The evolutionary strategy corresponding to the other sub-populations is: Among them, is the position of the i-th particle in the k-th subgroup of the (t + 1)-th generation population. c1, c2, and c3 are learning factors and non-negative constants, and r1, r2, and r3 are random numbers between [0, 1]. is the particle 's individual historical optimal position. is the randomly selected particle 's individual historical optimal position, and Gbest t is the global best particle of the t-th generation population. is the population best particle of the (k - 1)-th subgroup of the t-th generation population.

2. The method for optimizing the equipment capacity of an integrated energy system based on an improved particle swarm algorithm according to claim 1, wherein The objective function of the comprehensive energy system model is specifically the minimization of the initial investment cost.

3. An optimization method for the equipment capacity of an integrated energy system based on an improved particle swarm algorithm according to any one of claims 1 to 2, characterized in that, The constraint conditions of the comprehensive energy system model include equipment operation power constraints and electrical, heating, and cooling power constraints.

4. A method for optimizing the equipment capacity of an integrated energy system based on an improved particle swarm algorithm according to claim 1, characterized in that The specific initialization of the population includes: initializing the particle positions, the initial individual extreme value is the current particle itself, and the global extreme value in the current population is determined through the feasibility rule.

5. According to the method for optimizing the equipment capacity of a comprehensive energy system based on an improved particle swarm optimization algorithm according to claim 1, the elite promotion evolutionary strategy is specifically: Among them, For particle X i The historical optimal solution of the new individual optimized by the elite promotion strategy, Pbest q , Pbest w , Pbest e are the optimal solutions of individuals randomly selected from the population respectively. α is a random number between [0, 1], and d represents the d-th dimension.

6. According to the method for optimizing the equipment capacity of a comprehensive energy system based on an improved particle swarm optimization algorithm according to claim 1, the termination condition is specifically that the current iteration number is greater than or equal to the set maximum iteration number.

Images