Research On Multi-objective Optimization Based On Neurodynamic Algorithm In Microgrid

Micro-grid(MG)is a small power distribution system composed of a group of distributed power supply,load,and energy storage equipment,which is a supplement and development to the traditional power grid.Since the micro-grid relies on renewable energy to reduce the dependence of power system on traditional energy and has the advantages of providing power for users and high system stability,therefore,it has attracted extensive attention from all countries.Because most optimization problems in micro-grid are complex problems with multiple variables,multiple objectives and multiple constraints,so,Therefore,this paper focuses on the multi-objective problem in micro-grid and how to deal with the multi-objective optimization.First at all,we build a grid-connected microgrid scenario.At the same time,an incentive-based demand response mechanism is proposed to optimize the electricity consumption in each time period.In order to effectively solve the multi-objective problem,We transform it into a single objective optimization problem by weighting different objective functions.The optimal value is obtained by using the projection neural network.The introduction of particle swarm optimization(pso)is mainly to update the weight value and obtain the pareto front.Finally,the Lyapunov function is used to prove that the proposed algorithm can converge to the optimal solution.Numerical simulation in microgrid also shows the feasibility of the algorithm.Inspired by collaborative neurodynamic optimization algorithm for solving multiobjective problems,a distributed algorithm framework is proposed to solve the multiobjective problems with global equality constraints and local inequality constraints.Here,we propose to transform redundant global equality constraints into local equality constraints.According to the properties of multi-objective,it can be inferred that the result still satisfies the global equation.In this framework,each agent first needs to find its own maximum and minimum.Then the influence of singular value is eliminated by normalization function.Next,we select weights for each agent according to its importance,and of course,the number of weighted samples directly affects the number of pareto solutions.Finally,a distributed algorithm is proposed to solve the weighted multi-objective problem.In addition,the experimental results not only prove the correctness of the frame and the algorithm,but also show that the weighted solutions after normalization are more evenly distributed than before.