Multi-agent architecture for optimal energy management of a smart micro-grid using a weighted hybrid BP-PSO algorithm for wind power prediction

Abstract

In this paper we present a multi-agent architecture based on wind power prediction using neural network (NN), this process aims to implement smart micro-grid with different generation units like wind turbines and fuel generators. In the proposed architecture this micro-grid can exchange electricity with the main grid therefore it can buy or sell electricity. The main objective is to find the optimal policy using average wind speed prediction for the next hour in order to maximise the benefit and minimise the cost. To forecast the wind speed and taking into account the convergent speed and convergent accuracy, we propose in this paper an NN based on hybrid weighted algorithm combining back-propagation (BP) algorithm with particle swarm optimisation (PSO) algorithm referred to as W-BP-PSO. Finally, for the simulation, the Java Agent Development Framework (JADE) platform is used to implement the approach and analyse the results.