Price Forecast Methodologies Comparison for Microgrid Control with Multi-Agent Systems

Multi-Agent systems offer a way to control distributed generation in microgrids, reliability and cost minimisation capabilities can be improved by price forecast methodologies that can be deployed without the need of external control signals. This paper presents and compares two suitable electricity price forecast methodologies for use in distributed control of Microgrids’ resources using Multi-Agents: Markov Chain Monte Carlo simulations with heuristic and numerical optimisation and price prediction with Non-linear Auto Regressive Artificial Neural Networks with different internal architectures. The methods are evaluated using MAPE and RMSE functions for the UK electricity market data. It was found that the proposed heuristic model has less error than the Neural Networks only when the price data contains outliers.