Constrained Multi-objective Optimization Method Based On Improved Feasibility Rule And Its Application

The multi-objective optimization algorithm has become one of the research hotspots in the field of evolutionary computation.However,the multi-objective optimization problems in the actual science or engineering fields generally have constraint conditions.Some multi-objective optimization problems have relatively more constraint conditions and lead to smaller feasible regions.The general constraint processing method dealing with such constrained multi-objective optimization problems is not satisfactory.Therefore,this dissertation focuses on different types of constraints in multi-objective optimization problems,and applies the constraint multi-objective processing methods proposed in this paper to different multi-objective optimization problems in power grids.The main research contents are as follows:1)Constrained multi-objective approach to improve feasibility rulesFor the feasibility rule,the "best" infeasible solution can not be fully utilized,and an improved feasibility rule is proposed.The non-dominated rank information and the constraint violation degree information are integrated into the feasibility rule to improve the search ability to the non-dominated individual.The improved feasibility rules are integrated into the NSGA2 and IP-MOEA algorithm framework,respectively,to improve the constraint processing ability of the multi-objective optimization algorithm.2)A multi-objective evolutionary algorithm for discrete distribution of decision variablesFor the constraint problem of discontinuity of decision variables,we first design an individual generator that satisfies the interval discrete solution.Then,for the case that the conventional constraint violation calculation method cannot be applied to discrete constraints of decision variables,a method for calculating the violation degree of interval discrete constraints is proposed.3)A multi-objective optimization algorithm for variable constraintsAiming at the problem of variable constraints in multi-objective optimization problems,a multi-objective optimization algorithm for dealing with variable constraints is proposed.If the constraints have changed,the population before the constraints change as the initial population.Algorithm first archive the population before the constraints change and make full use of the original solution in the population to improve the efficiency of the algorithm.The algorithm one execution result saves the optimal solution for each constraint.4)Application of Constrained Multi-objective Optimization Algorithm in Power GridIn view of the randomness and load uncertainty of DG output in distribution networks,an interval model with the goal of minimizing investment operating costs,line loss costs,and power purchase costs of distributed generation is established.The IP-MOEA algorithm solves this model and achieves good results;Aiming at the multi-objective problem of discrete distribution of decision variables in power plant load distribution,a multi-objective evolutionary algorithm for discrete distribution of decision variables was proposed.The multi-objective optimization problem was solved and good results were obtained.