Skip to content
TheChatArchiveThe Chat Archive
gpt-4o-2024-08-06· OpenAIAi2 WildChat-4.8M (AI2)

can you run this code for 10 run. then plot the best pareto front from all of that run.…

  1. User

    can you run this code for 10 run. then plot the best pareto front from all of that run. use that plot and pick 1 point from the pareto front with the best Comfort Index and best Energy Consumption and display it. then use that point to do the Plot Settings Before and After Optimization.

    % run_MOEMA_MultiObjective.m % Comprehensive script to run MOEMA optimization for Smart Building Comfort and Energy Consumption

    %% ------------------ Initialization ------------------

    clc; clear; close all;

    % ------------------ Data Initialization ------------------

    % Data from UTHM % Columns: Time_instance, Temp_(F), Temp_(C), lux, IAQ dataSH = [
    1.0000 60.0000 15.5600 711.0000 610.0000 2.0000 65.0000 18.3300 717.0000 670.0000 3.0000 67.0000 19.4400 718.0000 650.0000 4.0000 66.0000 18.8900 717.0000 640.0000 5.0000 67.0000 19.4400 700.0000 600.0000 6.0000 63.0000 17.2200 713.0000 620.0000 7.0000 62.0000 16.6700 724.0000 670.0000 8.0000 81.0000 27.2200 710.0000 920.0000 9.0000 80.0000 26.6700 914.0000 920.0000 10.0000 81.0000 27.2200 900.0000 960.0000 11.0000 83.0000 28.3300 897.0000 900.0000 12.0000 80.0000 26.6700 895.0000 930.0000 13.0000 79.0000 26.1100 904.0000 950.0000 14.0000 82.0000 27.7800 903.0000 960.0000 15.0000 81.0000 27.2200 906.0000 970.0000 16.0000 80.0000 26.6700 898.0000 925.0000 17.0000 66.0000 18.8900 728.0000 610.0000 18.0000 63.0000 17.2200 703.0000 670.0000 19.0000 66.0000 18.8900 714.0000 630.0000 20.0000 66.0000 18.8900 721.0000 645.0000 21.0000 66.0000 18.8900 722.0000 650.0000 22.0000 67.0000 19.4400 703.0000 600.0000 23.0000 66.0000 18.8900 701.0000 670.0000 24.0000 81.0000 27.2200 728.0000 980.0000 25.0000 81.0000 27.2200 891.0000 930.0000 26.0000 82.0000 27.7800 906.0000 948.0000 27.0000 80.0000 26.6700 901.0000 965.0000 28.0000 81.0000 27.2200 901.0000 916.0000 29.0000 79.0000 26.1100 915.0000 900.0000 30.0000 83.0000 28.3300 914.0000 960.0000 31.0000 79.0000 26.1100 890.0000 970.0000 32.0000 81.0000 27.2200 912.0000 930.0000 33.0000 65.0000 18.3300 717.0000 610.0000 34.0000 67.0000 19.4400 714.0000 620.0000 35.0000 65.0000 18.3300 724.0000 670.0000 36.0000 67.0000 19.4400 726.0000 680.0000 37.0000 66.0000 18.8900 723.0000 650.0000 38.0000 64.0000 17.7800 726.0000 660.0000 39.0000 65.0000 18.3300 708.0000 640.0000 40.0000 80.0000 26.6700 712.0000 900.0000 41.0000 83.0000 28.3300 917.0000 910.0000 42.0000 79.0000 26.1100 905.0000 920.0000 43.0000 82.0000 27.7800 893.0000 980.0000 44.0000 80.0000 26.6700 920.0000 940.0000 45.0000 81.0000 27.2200 913.0000 950.0000 46.0000 81.0000 27.2200 905.0000 940.0000 47.0000 82.0000 27.7800 896.0000 970.0000 48.0000 80.0000 26.6700 898.0000 980.0000 ];

    % ------------------ MOEMA Parameters ------------------

    % Number of decision variables: 3 per period for 48 periods dim = 144;

    % Population size and number of iterations NPop = 150; % Population size T = 300; % Number of iterations

    % Define lower and upper bounds for each variable % Each period has T, L, IAQ variables with specific bounds lb = zeros(1, dim); ub = zeros(1, dim);

    % Define bounds per variable type % T: Temperature [20, 25.56] % L: Light [720, 880] % IAQ: IAQ [700, 880]

    % Assign bounds for each variable in the decision vector for k = 1:48 idx = 3*(k-1) + 1; lb(idx) = 20; % Temperature lower bound ub(idx) = 25.56; % Temperature upper bound

    lb(idx+1) = 720;    % Light lower bound
    ub(idx+1) = 880;    % Light upper bound
    
    lb(idx+2) = 700;    % IAQ lower bound
    ub(idx+2) = 880;    % IAQ upper bound
    

    end

    % ------------------ Define Fitness Function ------------------

    % Define the fitness function handle with dataSH fitness_fcn = @(X) multi_objective_cost_function(X, dataSH);

    % ------------------ Run MOEMA ------------------

    % Run the MOEMA optimization [paretoFront, paretoSet, Trajectories, fitness_history, position_history, spacing] = MOEMA(NPop, T, lb, ub, dim, fitness_fcn);

    % ------------------ Plot the Pareto Front ------------------

    figure; plot(paretoFront(:,1), paretoFront(:,2), 'ro', 'MarkerSize',8, 'LineWidth',1.5); xlabel('-Average Comfort Index'); ylabel('Average Energy Consumption'); title('Pareto Front for Smart Building Optimization using MOEMA'); grid on; set(gca, 'FontSize', 12);

    % First, find the best balanced solution from Pareto front % Convert -CI_avg back to CI_avg CI_values = -paretoFront(:,1); EC_values = paretoFront(:,2);

    % Normalize the values CI_norm = (CI_values - min(CI_values)) / (max(CI_values) - min(CI_values)); EC_norm = (EC_values - min(EC_values)) / (max(EC_values) - min(EC_values));

    % Find the solution closest to ideal point (1,0) distances = sqrt((1-CI_norm).^2 + EC_norm.^2); [~, best_idx] = min(distances);

    % Get the best solution settings optimized_solution = paretoSet(best_idx, :); % 1x144 vector

    % Extract optimized settings for each period optimized_settings = reshape(optimized_solution, 3, 48)'; % 48x3 matrix: [T, L, IAQ]

    % Original settings from dataSH original_temp = dataSH(:,3); % Temperature in C original_lux = dataSH(:,4); % lux original_IAQ = dataSH(:,5); % IAQ

    % Optimized settings from the best balanced solution optimized_temp = optimized_settings(:,1); optimized_lux = optimized_settings(:,2); optimized_IAQ = optimized_settings(:,3);

    % Desired values desired_temp = 22.78; % As per cost function desired_lux = 800; % As per cost function desired_IAQ = 800; % As per cost function

    % Create a time vector for 48 periods time = dataSH(:,1); % Assuming time instance is from 1 to 48

    % Temperature Plot figure; plot(time, original_temp, 'b-', 'LineWidth',1.5); hold on; plot(time, optimized_temp, 'r--', 'LineWidth',1.5); yline(desired_temp, 'g:', 'LineWidth',1.5); xlabel('Time Instance'); ylabel('Temperature (°C)'); title('Temperature Before and After Optimization (Best Balanced Solution)'); legend('Original', 'Optimized', 'Desired', 'Location', 'best'); grid on; set(gcf, 'Position', [100, 100, 800, 400]); % Set window size

    % Light Level Plot figure; plot(time, original_lux, 'b-', 'LineWidth',1.5); hold on; plot(time, optimized_lux, 'r--', 'LineWidth',1.5); yline(desired_lux, 'g:', 'LineWidth',1.5); xlabel('Time Instance'); ylabel('Light Level (lux)'); title('Light Level Before and After Optimization (Best Balanced Solution)'); legend('Original', 'Optimized', 'Desired', 'Location', 'best'); grid on; set(gcf, 'Position', [200, 200, 800, 400]); % Set window size

    % IAQ Plot figure; plot(time, original_IAQ, 'b-', 'LineWidth',1.5); hold on; plot(time, optimized_IAQ, 'r--', 'LineWidth',1.5); yline(desired_IAQ, 'g:', 'LineWidth',1.5); xlabel('Time Instance'); ylabel('IAQ'); title('IAQ Before and After Optimization (Best Balanced Solution)'); legend('Original', 'Optimized', 'Desired', 'Location', 'best'); grid on; set(gcf, 'Position', [300, 300, 800, 400]); % Set window size

    % Display metrics for the best balanced solution fprintf('\nBest Balanced Solution Metrics:\n'); fprintf('Comfort Index: %.3f\n', CI_values(best_idx)); fprintf('Energy Consumption: %.2f\n', EC_values(best_idx));

    % ------------------ plot for the best pareto ------------------ figure; % Plot all Pareto front points in red plot(paretoFront(:,1), paretoFront(:,2), 'ro', 'MarkerSize',8, 'LineWidth',1.5); hold on; % Plot the balanced solution point in green plot(paretoFront(best_idx,1), paretoFront(best_idx,2), 'go', 'MarkerSize',12, 'LineWidth',2, 'MarkerFaceColor','g'); xlabel('-Average Comfort Index'); ylabel('Average Energy Consumption'); title('Pareto Front for Smart Building Optimization using MOEMA'); legend('Pareto Front', 'Balanced Solution', 'Location', 'best'); grid on; set(gca, 'FontSize', 12);

    % ------------------ Display Metrics ------------------

    disp(['Number of solutions in Pareto Front: ', num2str(size(paretoFront, 1))]); disp(['Minimum -Average Comfort Index: ', num2str(min(paretoFront(:,1)),4)]); disp(['Minimum Average Energy Consumption: ', num2str(min(paretoFront(:,2)),4)]); % Display the spacing metric fprintf('Spacing Metric: %.4f\n', spacing); %% ------------------ Functions ------------------

    % ---------- Multi-Objective Cost Function ---------- function objectives = multi_objective_cost_function(X, dataSH) % MULTI_OBJECTIVE_COST_FUNCTION Evaluates user comfort and energy consumption. % % This function is designed for multi-objective optimization with two objectives: % 1. Maximize User Comfort (to be minimized as -CI_avg) % 2. Minimize Energy Consumption (EC_avg) % % Inputs: % X - A vector of decision variables for 48 periods, structured as: % [T1, L1, IAQ1, T2, L2, IAQ2, ..., T48, L48, IAQ48] % dataSH - A 48x5 matrix where: % dataSH(:,3) = Temperature coefficients (Tc) % dataSH(:,4) = Light coefficients (Lc) % dataSH(:,5) = IAQ coefficients (IAQc) % % Output: % objectives - A 2-element vector: % [ -Average Comfort Index, Average Energy Consumption ]

    % Ensure that X has the correct length
    if length(X) ~= 144
        error('Decision vector X must have 144 elements (3 variables per 48 periods).');
    end
    
    % Weight coefficients for comfort components
    p1 = 1/3; % Temperature Weight
    p2 = 1/3; % Light Weight
    p3 = 1/3; % IAQ Weight
    
    % Desired setpoints for comfort
    Td = 22.78; % Desired Temperature (°C)
    Ld = 800;   % Desired Light Level (lux)
    Id = 800;   % Desired IAQ Level
    
    % Energy consumption weights
    pT = 9; % Temperature Energy Consumption Weight
    pL = 1; % Light Energy Consumption Weight
    pI = 1; % IAQ Energy Consumption Weight
    
    % Extract coefficients from dataSH for all 48 periods
    Tc = dataSH(:,3)'; % Temperature coefficients (1x48 vector)
    Lc = dataSH(:,4)'; % Light coefficients (1x48 vector)
    Ic = dataSH(:,5)'; % IAQ coefficients (1x48 vector)
    
    % Initialize vectors to store metrics for each period
    CI = zeros(48,1);    % Comfort Index for each period
    EC = zeros(48,1);    % Energy Consumption for each period
    
    % Loop through each of the 48 periods to compute metrics
    for k = 1:48
        % Extract decision variables for period k
        Tk = X(3*k-2); % Temperature setting
        Lk = X(3*k-1); % Light setting
        Ik = X(3*k);   % IAQ setting
        
        % --- Calculate Comfort Index Components ---
        Ti = 1 - ((Tk - Td) / Tc(k))^2;
        Li = 1 - ((Lk - Ld) / Lc(k))^2;
        Ii = 1 - ((Ik - Id) / Ic(k))^2;
        CI(k) = p1 * Ti + p2 * Li + p3 * Ii;
        
        % --- Calculate Energy Consumption ---
        ET = pT * abs(Tk - Tc(k));
        EL = pL * abs(Lk - Lc(k));
        EI = pI * abs(Ik - Ic(k));
        EC(k) = ET + EL + EI;
    end
    
    % --- Aggregate Metrics Over All Periods ---
    CI_avg = mean(CI);   % Average Comfort Index
    EC_avg = mean(EC);   % Average Energy Consumption
    
    % --- Prepare Objectives for Multi-Objective Optimization ---
    % Objective 1: Maximize User Comfort -> Minimize (-CI_avg)
    % Objective 2: Minimize Energy Consumption -> Minimize (EC_avg)
    objectives = [-CI_avg, EC_avg];
    

    end

    function [paretoFront, paretoSet, Trajectories, fitness_history, position_history, spacing] = MOEMA(NPop, T, lb, ub, dim, FITNESSFCN) % MOEMA: Multi-Objective Evolutionary Mating Algorithm % % Solves multi-objective optimization problems of the form: % min F(X) = [f1(X), f2(X)] subject to: lb <= X <= ub % % Inputs: % NPop - Population size % T - Number of iterations % lb - Lower bounds (1 x dim) % ub - Upper bounds (1 x dim) % dim - Number of decision variables % FITNESSFCN - Function handle for the multi-objective fitness function % % Outputs: % paretoFront - Set of non-dominated solutions in objective space % paretoSet - Set of non-dominated solutions in decision space % Trajectories - Best solution trajectory over iterations % fitness_history - History of fitness values % position_history - History of population positions % spacing - Spacing metric of the obtained Pareto front

    % Preallocation
    fitness_history = zeros(NPop, T, 2);    % Assuming 2 objectives
    position_history = zeros(NPop, T, dim);
    Trajectories = zeros(dim, T);
    
    % Generation of initial population
    pop = initialization(NPop, dim, ub, lb);
    pop_dad = pop(1:floor(NPop/2), :);      % Parent subset 1
    pop_mum = pop(floor(NPop/2)+1:end, :);  % Parent subset 2
    
    % Evaluation of objective function for initial population
    obj_dad = evaluate_population(pop_dad, FITNESSFCN);
    obj_mum = evaluate_population(pop_mum, FITNESSFCN);
    obj = [obj_dad; obj_mum];
    
    % Initialize archive with non-dominated solutions
    archive = update_archive([], [], pop, obj);
    
    % Evolutionary Loop
    for gen = 1:T
        % Shuffle parents
        k1 = randperm(size(pop_dad,1));
        k2 = randperm(size(pop_mum,1));
        k1x = pop_dad(k1, :);
        k2x = pop_mum(k2, :);
        
        % Generate children
        for k = 1:size(k1x,1)
            % Mating process
            imates = randn(1, dim);  % Mating factor
            p = rand(1, dim);        % Crossover probability
            r = rand();
            
            if mean(imates) >= 0  % Decide mating based on imates
                child = p .* k1x(k, :) + (1 - p) .* k2x(k, :);
            else
                child = p .* k2x(k, :) + (1 - p) .* k1x(k, :);
            end
            
            % Impact of best environment (using random non-dominated solution)
            if ~isempty(archive.pop)
                best_pop = archive.pop(randi(size(archive.pop,1)), :);
                K = rand(1, dim) < 0.8;  % Modification probability
                child = best_pop .* (1 - K) + child .* K;
            end
            
            % Ensure child is within bounds
            child = max(min(child, ub), lb);
            
            % Evaluate child
            obj_child = FITNESSFCN(child);
            
            % Update population using non-dominated sorting
            [pop_dad, obj_dad] = update_population(pop_dad, obj_dad, child, obj_child, k);
            [pop_mum, obj_mum] = update_population(pop_mum, obj_mum, child, obj_child, k);
            
            % Exploration process
            if r < 0.2
                if ~isempty(archive.pop)
                    temp1 = archive.pop(randi(size(archive.pop,1)), :);
                    temp_child = rand(1, dim) .* temp1;  % Exploration child
                    temp_child = max(min(temp_child, ub), lb);
                    obj_temp_child = FITNESSFCN(temp_child);
                    
                    % Randomly decide where to place the exploration child
                    replace_idx_dad = randi(size(pop_dad,1));
                    replace_idx_mum = randi(size(pop_mum,1));
                    
                    pop_dad(replace_idx_dad, :) = temp_child;
                    obj_dad(replace_idx_dad, :) = obj_temp_child;
                    
                    pop_mum(replace_idx_mum, :) = temp_child;
                    obj_mum(replace_idx_mum, :) = obj_temp_child;
                end
            end
        end
        
        % Combine updated populations
        pop = [pop_dad; pop_mum];
        obj = [obj_dad; obj_mum];
        
        % Update archive with non-dominated solutions
        archive = update_archive(archive.pop, archive.obj, pop, obj);
        
        % Record fitness and positions
        fitness_history(:, gen, :) = obj;
        position_history(:, gen, :) = pop;
        
        % Update Trajectories with a random non-dominated solution
        if ~isempty(archive.pop)
            Trajectories(:, gen) = archive.pop(randi(size(archive.pop,1)), :)';
        end
        
        % Display iteration status
        disp(['Iteration #', num2str(gen), ' completed. Archive size: ', num2str(size(archive.pop,1))]);
    end
    
    % Return Pareto front and Pareto set
    paretoFront = archive.obj;
    paretoSet = archive.pop;
    
    % Calculate spacing metric
    spacing = spacing_metric(paretoFront);
    

    end

    % ---------- Spacing Metric Function ---------- function spacing = spacing_metric(obtained_pf) if size(obtained_pf, 1) < 2 spacing = Inf; return; end

    n = size(obtained_pf, 1);
    d = zeros(n, 1);
    
    % Calculate di for each solution
    for i = 1:n
        % Calculate distances to all other solutions
        min_dist = Inf;
        for j = 1:n
            if i ~= j
                % Calculate |f1(x) - f1(x')| + |f2(x) - f2(x')|
                dist = abs(obtained_pf(i,1) - obtained_pf(j,1)) + ...
                       abs(obtained_pf(i,2) - obtained_pf(j,2));
                min_dist = min(min_dist, dist);
            end
        end
        d(i) = min_dist;
    end
    
    % Calculate mean of distances
    d_bar = mean(d);
    
    % Calculate spacing metric SP
    spacing = sqrt(sum((d_bar - d).^2) / (n - 1));
    

    end

    % ---------- Initialization Function ---------- function pop = initialization(NPop, dim, ub, lb) % INITIALIZATION Generates an initial population within bounds pop = rand(NPop, dim) .* (ub - lb) + lb; end

    % ---------- Evaluate Population Function ---------- function obj = evaluate_population(pop, FITNESSFCN) % EVALUATE_POPULATION Evaluates the objective function for each individual num_individuals = size(pop,1); obj = zeros(num_individuals, 2); % Assuming 2 objectives parfor i = 1:num_individuals obj(i, :) = FITNESSFCN(pop(i, :)); end end

    % ---------- Update Population Function ---------- function [pop, obj] = update_population(pop, obj, child, child_obj, index) % UPDATE_POPULATION Updates the population based on dominance if dominates(child_obj, obj(index, :)) pop(index, :) = child; obj(index, :) = child_obj; elseif dominates(obj(index, :), child_obj) % Do nothing, parent dominates child else % Neither dominates; randomly decide to replace if rand() < 0.5 pop(index, :) = child; obj(index, :) = child_obj; end end end

    % ---------- Update Archive Function ---------- function archive = update_archive(archive_pop, archive_obj, pop, obj) % UPDATE_ARCHIVE Updates the archive with non-dominated solutions combined_pop = [archive_pop; pop]; combined_obj = [archive_obj; obj];

    [is_dominated, ~] = pareto_dominance(combined_obj);
    non_dominated = ~is_dominated;
    
    archive.pop = combined_pop(non_dominated, :);
    archive.obj = combined_obj(non_dominated, :);
    

    end

    % ---------- Pareto Dominance Function ---------- function [is_dominated, domination_count] = pareto_dominance(objectives) % PARETO_DOMINANCE Determines domination relationships among solutions % % Outputs: % is_dominated - Logical array indicating if a solution is dominated % domination_count - Number of times a solution is dominated

    n_solutions = size(objectives,1);
    is_dominated = false(n_solutions,1);
    domination_count = zeros(n_solutions, 1);
    
    for i = 1:n_solutions
        for j = 1:n_solutions
            if i ~= j
                if dominates(objectives(j,:), objectives(i,:))
                    is_dominated(i) = true;
                    domination_count(j) = domination_count(j) + 1;
                end
            end
        end
    end
    

    end

    % ---------- Dominates Function ---------- function result = dominates(obj1, obj2) % DOMINATES Checks if obj1 dominates obj2 result = all(obj1 <= obj2) && any(obj1 < obj2); end

  2. Assistant

    I’m unable to execute code or run simulations directly. However, I can guide you on how to structure your script to perform the tasks you mentioned based on the MATLAB script you've provided.

    To run your MOEMA optimization 10 times and plot/compare the Pareto fronts, you would typically employ a loop to repeat the entire optimization process and collect the results. Here's how you could modify the script to achieve this:

    1. Initialize arrays to store results from multiple runs:

      all_pareto_fronts = cell(10,1);
      all_pareto_sets = cell(10,1);
      
    2. Modify the MOEMA Loop to run 10 times:

      for run = 1:10
          % Call your existing MOEMA function with necessary inputs
          [paretoFront, paretoSet, ~, ~, ~, ~] = MOEMA(NPop, T, lb, ub, dim, fitness_fcn);
          
          % Store the results
          all_pareto_fronts{run} = paretoFront;
          all_pareto_sets{run} = paretoSet;
      end
      
    3. Identify the best Pareto front regarding a certain criterion (e.g., Comfort Index or Energy Consumption):

      To determine the best Pareto front, you could calculate some metrics like mean or hypervolume, or choose a strategy that fits your project needs. Here is just an example by selecting the Pareto front with the best individual solution.

      best_index = 1; % Initialize, modify with actual logic 
      best_metric_value = inf; % Initialize, modify with actual logic 
      
      for i = 1:10
          % Here the logic to choose the best front should be incorporated
          % It might be a complex criterion based on specific objectives
          current_metric_value = someMetric(all_pareto_fronts{i}); % Define 'someMetric' as per requirement
          if current_metric_value < best_metric_value
              best_index = i;
              best_metric_value = current_metric_value;
          end
      end
      
      % Extract the best front and set
      best_pareto_front = all_pareto_fronts{best_index};
      best_pareto_set = all_pareto_sets{best_index};
      
    4. Plot the best Pareto front:

      figure;
      plot(best_pareto_front(:,1), best_pareto_front(:,2), 'ro', 'MarkerSize',8, 'LineWidth',1.5);
      xlabel('-Average Comfort Index');
      ylabel('Average Energy Consumption');
      title('Best Pareto Front from Multiple Runs');
      grid on;
      
    5. Select the optimal point and Plot settings before/after optimization:

      Use the logic from your initial script:

      % Determine the best point using your utility function
      [best_point, best_point_index] = selectBestPoint(best_pareto_front); 
      
      optimized_solution = best_pareto_set(best_point_index, :);
      optimized_settings = reshape(optimized_solution, 3, 48)';  % 48x3 matrix: [T, L, IAQ]
      
      % Plotting similar to your script logic
      plotSettingsComparison(dataSH, optimized_settings);
      
    6. Additional Functions Placeholder:

      Implement additional utility functions like someMetric, selectBestPoint, or plotSettingsComparison as necessary.

    This approach allows you to repeatedly execute the optimization process, accumulate results, and systematically determine which result best suits your objectives prior to visualization. You’ll need to replace placeholder logic with specific analyses that align with your decision criteria.