Distributed neurodynamic algorithm for multi-objective problem optimization and its applications to isolated micro-grid energy management

Abstract

In this paper, a novel distributed framework is proposed to solve the multi-objective problem which composed of several conflicting objective functions. In order to overcome the impact of inconsistent units of measurement for each objective function, a normalized function is applied to eliminate it. It will split a multi-objective optimization problem into three subproblems: the problem of maximization, the problem of minimization, the problem after normalization. Linear weighting method is adopted to assign weights to each normalized objective function and convert them into a single objective problem. Specifically, each objective function which regarded as an agent will calculate its maximum and minimum values respectively and finally the Pareto optimal solution is obtained through a distributed neuro-dynamic algorithm. The whole process is distributed and independent for each agent only needs its own rather than the global information. Furthermore, the convergence of the proposed algorithm can be proved through Lyapunov function. Eventually, a numerical example and a multi-objective optimization problem in micro grid are used to prove the credibility of the framework.