The Research Of Chiller Load Allocation Problem Based On MOEA/D

The optimization chiller problem(OCL)is a type of optimization problem in engineering.To obtain certain load demands,while considering certain constraints,the OCL problem is a typical constraint optimization problem.In the classical OCL problem,two objectives should be considered simultaneously,i.e.,minimization of the energy consumptions,and maximization of the system performance.Therefore,the problem is a typical multi-objective optimization problem.Because of the complexity of the objective features,the traditional optimization algorithms cannot be directly applied to solve these types of optimization problems.With the development of evolutionary algorithms,multi-objective evolutionary algorithms are widely investigated for solving these types of constrained optimization problems,with the features of global search capabilities and parallel performances.Based on the framework of MOEA/D algorithm,this study developed an improved multi-objective optimization algorithm to solve the OCL problems.The main contributions of this study are as follows.(1)A multi-objective method is presented to solve the constrained optimization problem.The multi-objective method is used to process the constraint conditions to expand the search scope and avoid falling into the local optimal.The multi-objective method can also be used to find more optimal solution sets at the same time,so that the decision maker can select and set according to different tolerances for errors.(2)An improved decomposition strategy is developed to deal with the problem of large differences in the value range of the objective function.In this study,the weight vector is used to aggregate multiple objective functions.Because different objective functions have different range of values,therefore,individual objective functions have greater influence in the final solution process,and individual targets almost have none results.To eliminate this type of effect,the decomposition strategy was standardized and transformed to eliminate the impact of the unified value range.(3)The biased dominance relationship is proposed to speed up the handling of theconstraints.Through transforming the constraints functions to objective functions,the search scope will be enlarged and the local optimum will be avoided.However,the computational complexity will be increased.To decrease the computational complexity,the biased and dominant approach is utilized to speed up the search process,and therefore more computing power will be used for the original target search process.