当前位置：首页 > news >正文

基于遗传算法的电动汽车充电站选址优化：模型与MATLAB实现

news 2026/5/5 15:54:51

电动汽车充电站选址优化模型，采用遗传算法求解。该问题综合考虑了建设成本、用户充电便利性、电网负荷等多重因素。

一、问题数学模型

1.1 问题描述

在给定区域内，从 \(M\) 个候选站点中选择 \(N\) 个建设充电站，使得：

总建设成本最小
用户充电总距离最小
电网负荷均衡性最好

1.2 数学模型

决策变量：

\[x_j = \begin{cases} 1 & \text{在候选点} j \text{建设充电站} \\ 0 & \text{否则} \end{cases}\]

目标函数（多目标优化）：

\[\min F = w_1 \cdot f_1 + w_2 \cdot f_2 + w_3 \cdot f_3 \]

其中：

建设成本：\(f_1 = \sum_{j=1}^{M} c_j \cdot x_j\)
用户充电距离：\(f_2 = \sum_{i=1}^{K} \min_{j \in S} d_{ij} \cdot d_i\)
负荷均衡度：\(f_3 = \sqrt{\frac{1}{N} \sum_{j=1}^{M} (L_j - \bar{L})^2}\)

约束条件：

建设数量约束：\(\sum_{j=1}^{M} x_j = N\)
容量约束：\(\sum_{i \in A_j} d_i \leq C_j \cdot x_j, \quad \forall j\)
服务半径约束：\(d_{ij} \leq R_{\max}, \quad \forall i,j\)

二、 MATLAB代码

2.1 主程序：充电站选址优化

%% 基于遗传算法的电动汽车充电站选址优化
clear; clc; close all;%% 1. 参数设置与数据生成
% 算法参数
pop_size = 100;          % 种群规模
max_gen = 200;           % 最大迭代次数
pc = 0.8;                % 交叉概率
pm = 0.05;               % 变异概率
elite_rate = 0.1;        % 精英保留比例% 问题参数
num_candidates = 50;     % 候选站点数量
num_selected = 10;       % 需要建设的站点数量
num_demand_points = 200; % 需求点数量% 权重设置 (多目标权重)
w_cost = 0.4;           % 建设成本权重
w_distance = 0.4;       % 充电距离权重
w_balance = 0.2;        % 负荷均衡权重% 生成模拟数据
rng(42); % 设置随机种子确保可重复性
[candidate_coords, demand_coords, candidate_costs, demand_weights] = ...generate_charging_station_data(num_candidates, num_demand_points);% 计算距离矩阵
dist_matrix = pdist2(demand_coords, candidate_coords);%% 2. 遗传算法初始化
% 初始化种群 (二进制编码)
population = init_population(pop_size, num_candidates, num_selected);% 计算初始适应度
fitness = zeros(pop_size, 1);
for i = 1:pop_sizefitness(i) = calculate_fitness(population(i,:), candidate_costs, ...demand_weights, dist_matrix, w_cost, w_distance, w_balance);
end% 记录最优解
[best_fitness, best_idx] = min(fitness);
best_solution = population(best_idx, :);
fitness_history = zeros(max_gen, 1);%% 3. 遗传算法主循环
fprintf('开始遗传算法优化...\n');
for gen = 1:max_gen% 选择操作 (锦标赛选择)parents = tournament_selection(population, fitness, pop_size);% 交叉操作 (单点交叉)offspring = crossover(parents, pc, num_selected);% 变异操作 (位翻转变异)offspring = mutation(offspring, pm, num_selected);% 精英保留策略[~, sorted_idx] = sort(fitness);elite_count = round(elite_rate * pop_size);elite_pop = population(sorted_idx(1:elite_count), :);% 合并种群new_population = [elite_pop; offspring(1:pop_size-elite_count, :)];% 计算新种群适应度new_fitness = zeros(size(new_population, 1), 1);for i = 1:size(new_population, 1)new_fitness(i) = calculate_fitness(new_population(i,:), candidate_costs, ...demand_weights, dist_matrix, w_cost, w_distance, w_balance);end% 更新种群population = new_population;fitness = new_fitness;% 更新最优解[current_best, current_idx] = min(fitness);if current_best < best_fitnessbest_fitness = current_best;best_solution = population(current_idx, :);endfitness_history(gen) = best_fitness;% 显示进度if mod(gen, 20) == 0fprintf('第 %d 代，最优适应度: %.4f\n', gen, best_fitness);end
end%% 4. 结果分析与可视化
fprintf('\n========== 优化结果 ==========\n');
fprintf('最优适应度值: %.4f\n', best_fitness);
selected_sites = find(best_solution == 1);
fprintf('选中的充电站位置编号: %s\n', mat2str(selected_sites));% 计算各项目标值
[total_cost, total_distance, balance_degree] = ...evaluate_solution(best_solution, candidate_costs, demand_weights, dist_matrix);
fprintf('总建设成本: %.2f 万元\n', total_cost);
fprintf('用户总充电距离: %.2f km\n', total_distance);
fprintf('负荷均衡度: %.4f\n', balance_degree);% 可视化结果
visualize_results(candidate_coords, demand_coords, best_solution, ...demand_weights, dist_matrix);% 绘制收敛曲线
figure('Name', '遗传算法收敛曲线', 'Color', 'w');
plot(1:max_gen, fitness_history, 'b-', 'LineWidth', 2);
xlabel('迭代次数');
ylabel('适应度值');
title('遗传算法收敛曲线');
grid on;

2.2 数据生成函数

function [candidate_coords, demand_coords, candidate_costs, demand_weights] = ...generate_charging_station_data(num_candidates, num_demand_points)% 生成候选站点坐标 (均匀分布)candidate_coords = rand(num_candidates, 2) * 100;% 生成需求点坐标 (聚类分布，模拟居民区)demand_coords = zeros(num_demand_points, 2);cluster_centers = [20, 20; 80, 20; 50, 80; 20, 80; 80, 80];cluster_sizes = [0.3, 0.2, 0.25, 0.15, 0.1] * num_demand_points;idx = 1;for c = 1:length(cluster_sizes)n_points = round(cluster_sizes(c));for i = 1:n_pointsif idx > num_demand_pointsbreak;enddemand_coords(idx, :) = cluster_centers(c, :) + randn(1,2)*10;idx = idx + 1;endend% 生成候选站点建设成本 (万元)candidate_costs = 50 + 150 * rand(num_candidates, 1);% 生成需求点权重 (充电需求强度)demand_weights = 0.5 + 1.5 * rand(num_demand_points, 1);
end

2.3 适应度计算函数

function fitness = calculate_fitness(solution, candidate_costs, ...demand_weights, dist_matrix, w_cost, w_distance, w_balance)% 计算单个解的适应度值% 1. 建设成本total_cost = sum(candidate_costs(solution == 1));% 2. 用户充电距离selected_sites = find(solution == 1);if isempty(selected_sites)fitness = inf;return;end% 为每个需求点分配最近的充电站dist_to_selected = dist_matrix(:, selected_sites);[min_dist, ~] = min(dist_to_selected, [], 2);total_distance = sum(min_dist .* demand_weights);% 3. 负荷均衡度% 计算每个充电站的负荷station_loads = zeros(length(selected_sites), 1);for i = 1:length(selected_sites)% 找到分配给该站点的需求点[~, assigned_station] = min(dist_to_selected, [], 2);station_idx = (assigned_station == i);station_loads(i) = sum(demand_weights(station_idx));endif length(selected_sites) > 1avg_load = mean(station_loads);balance_degree = sqrt(sum((station_loads - avg_load).^2) / length(selected_sites));elsebalance_degree = 0;end% 4. 归一化处理max_cost = sum(candidate_costs);max_distance = sum(max(dist_matrix, [], 2) .* demand_weights);norm_cost = total_cost / max_cost;norm_distance = total_distance / max_distance;norm_balance = balance_degree / (sum(demand_weights) / length(selected_sites));% 5. 加权求和得到适应度fitness = w_cost * norm_cost + w_distance * norm_distance + w_balance * norm_balance;
end

2.4 遗传算法操作函数

%% 初始化种群
function population = init_population(pop_size, num_candidates, num_selected)population = zeros(pop_size, num_candidates);for i = 1:pop_size% 随机选择num_selected个位置设为1idx = randperm(num_candidates, num_selected);population(i, idx) = 1;end
end%% 锦标赛选择
function parents = tournament_selection(population, fitness, pop_size)tournament_size = 3;parents = zeros(size(population));for i = 1:pop_size% 随机选择tournament_size个个体candidates = randperm(pop_size, tournament_size);[~, best_idx] = min(fitness(candidates));parents(i, :) = population(candidates(best_idx), :);end
end%% 单点交叉
function offspring = crossover(parents, pc, num_selected)[pop_size, num_candidates] = size(parents);offspring = parents;for i = 1:2:pop_size-1if rand() < pc% 随机选择交叉点crossover_point = randi([1, num_candidates-1]);% 执行交叉parent1 = parents(i, :);parent2 = parents(i+1, :);child1 = [parent1(1:crossover_point), parent2(crossover_point+1:end)];child2 = [parent2(1:crossover_point), parent1(crossover_point+1:end)];% 修复解：确保选中站点数量正确child1 = repair_solution(child1, num_selected);child2 = repair_solution(child2, num_selected);offspring(i, :) = child1;offspring(i+1, :) = child2;endend
end%% 位翻转变异
function offspring = mutation(offspring, pm, num_selected)[pop_size, num_candidates] = size(offspring);for i = 1:pop_sizeif rand() < pm% 随机选择两个位置进行位翻转idx1 = randi(num_candidates);idx2 = randi(num_candidates);while idx1 == idx2idx2 = randi(num_candidates);end% 执行变异offspring(i, idx1) = 1 - offspring(i, idx1);offspring(i, idx2) = 1 - offspring(i, idx2);% 修复解offspring(i, :) = repair_solution(offspring(i, :), num_selected);endend
end%% 修复解：确保选中站点数量正确
function solution = repair_solution(solution, num_selected)current_selected = sum(solution);if current_selected > num_selected% 随机移除多余的站点selected_idx = find(solution == 1);remove_idx = randperm(length(selected_idx), current_selected - num_selected);solution(selected_idx(remove_idx)) = 0;elseif current_selected < num_selected% 随机添加缺失的站点unselected_idx = find(solution == 0);add_idx = randperm(length(unselected_idx), num_selected - current_selected);solution(unselected_idx(add_idx)) = 1;end
end

2.5 结果评估与可视化

function [total_cost, total_distance, balance_degree] = ...evaluate_solution(solution, candidate_costs, demand_weights, dist_matrix)% 评估解的质量selected_sites = find(solution == 1);% 1. 建设成本total_cost = sum(candidate_costs(selected_sites));% 2. 用户充电距离dist_to_selected = dist_matrix(:, selected_sites);[min_dist, assigned_station] = min(dist_to_selected, [], 2);total_distance = sum(min_dist .* demand_weights);% 3. 负荷均衡度station_loads = zeros(length(selected_sites), 1);for i = 1:length(selected_sites)station_idx = (assigned_station == i);station_loads(i) = sum(demand_weights(station_idx));endif length(selected_sites) > 1avg_load = mean(station_loads);balance_degree = sqrt(sum((station_loads - avg_load).^2) / length(selected_sites));elsebalance_degree = 0;end
endfunction visualize_results(candidate_coords, demand_coords, solution, ...demand_weights, dist_matrix)% 可视化优化结果figure('Name', '充电站选址优化结果', 'Color', 'w', 'Position', [100, 100, 1200, 500]);% 子图1：站点分布subplot(1, 2, 1);hold on; grid on; axis equal;% 绘制所有候选站点scatter(candidate_coords(:,1), candidate_coords(:,2), 50, 'k', 'o', 'filled');% 绘制选中的充电站selected_sites = find(solution == 1);scatter(candidate_coords(selected_sites,1), candidate_coords(selected_sites,2), ...100, 'r', 's', 'filled', 'LineWidth', 2);% 绘制需求点（大小表示需求强度）demand_sizes = 20 + 80 * (demand_weights - min(demand_weights)) / ...(max(demand_weights) - min(demand_weights));scatter(demand_coords(:,1), demand_coords(:,2), demand_sizes, 'b', '^', 'filled', ...'MarkerFaceAlpha', 0.3);% 绘制服务关系dist_to_selected = dist_matrix(:, selected_sites);[~, assigned_station] = min(dist_to_selected, [], 2);colors = lines(length(selected_sites));for i = 1:length(selected_sites)station_demands = find(assigned_station == i);if ~isempty(station_demands)for j = 1:min(20, length(station_demands)) % 只绘制部分连接线避免混乱plot([candidate_coords(selected_sites(i),1), demand_coords(station_demands(j),1)], ...[candidate_coords(selected_sites(i),2), demand_coords(station_demands(j),2)], ...'Color', colors(i,:), 'LineWidth', 0.5, 'LineStyle', '--');endendendlegend('候选站点', '选中站点', '需求点', 'Location', 'best');title('充电站选址与需求点分配');xlabel('X坐标 (km)'); ylabel('Y坐标 (km)');% 子图2：负荷分布subplot(1, 2, 2);station_loads = zeros(length(selected_sites), 1);for i = 1:length(selected_sites)station_idx = (assigned_station == i);station_loads(i) = sum(demand_weights(station_idx));endbar(1:length(selected_sites), station_loads, 'FaceColor', [0.2, 0.6, 0.8]);hold on;avg_load = mean(station_loads);plot([0, length(selected_sites)+1], [avg_load, avg_load], 'r--', 'LineWidth', 2);xlabel('充电站编号');ylabel('负荷强度');title('各充电站负荷分布');legend('实际负荷', '平均负荷', 'Location', 'best');grid on;
end

三、工程实践建议与扩展方向

3.1 实际应用中的关键考虑

多目标优化改进：
- 使用 NSGA-II（非支配排序遗传算法）处理真正的多目标优化
- 采用 Pareto前沿 分析，提供多个非劣解供决策者选择

约束处理增强：

% 添加电网容量约束
function penalty = check_grid_constraints(solution, grid_capacity)total_power = sum(solution .* station_power_demand);if total_power > grid_capacitypenalty = 1000 * (total_power - grid_capacity);elsepenalty = 0;end
end

动态需求考虑：
- 考虑不同时段的充电需求变化
- 引入时间维度，优化充电站容量配置

3.2 算法性能优化

并行计算加速：

% 使用parfor并行计算适应度
parfor i = 1:pop_sizefitness(i) = calculate_fitness(population(i,:), ...);
end

局部搜索增强：
- 在遗传算法中嵌入 2-opt 或 变邻域搜索 提升局部搜索能力
- 采用 自适应参数调整：根据进化过程动态调整交叉和变异概率

3.3 实际数据接口

% 从实际GIS数据读取
function [coords, costs] = read_real_data(filename)data = readtable(filename);coords = [data.Longitude, data.Latitude];costs = data.ConstructionCost;% 坐标转换（如需要）% coords = lla2utm(coords);
end% 从电力系统数据读取负荷信息
function load_profile = read_load_profile(power_system_data)% 处理实际负荷曲线load_profile = power_system_data.Demand;
end

3.4 扩展为多类型充电站

% 考虑快充站和慢充站混合配置
classdef ChargingStationpropertiesType        % 'fast' 或 'slow'Power       % 充电功率 (kW)Cost        % 建设成本Capacity    % 最大服务车辆数end
end% 在适应度函数中考虑不同类型
function fitness = calculate_fitness_multi_type(solution, stations, ...)% solution: 三维矩阵 [位置 × 类型 × 数量]% 计算混合类型充电站的综合适应度
end

参考代码基于遗传算法的充电站选址优化求解方法实例 www.youwenfan.com/contentcnu/65103.html

四、运行与调试建议

首次运行：
- 直接运行主程序，观察收敛曲线
- 调整权重参数 w_cost, w_distance, w_balance 观察不同偏好下的选址方案
参数调优：
- 种群大小 pop_size: 50-200
- 迭代次数 max_gen: 100-500
- 交叉概率 pc: 0.7-0.9
- 变异概率 pm: 0.01-0.1
结果验证：
- 多次运行验证算法稳定性
- 与穷举法（小规模问题）或商业求解器结果对比