Solar photovoltaic power forecasting for microgrid energy management system using an ensemble forecasting strategy

ABSTRACT

Solar photovoltaic (PV) power production is always unpredictable due to intermittent solar irradiance and weather parameters which pose a major challenge in microgrid energy management systems (M-EMS). An accurate solar PV power forecasting strategy is therefore important for the competent M-EMS that can lead to financial gains for customers and the power industry. In this paper, an ensemble forecasting strategy has been proposed to predict the solar PV power for M-EMS based on the combination of four different forecasting methods, i.e. tunicate swarm algorithm (TSA)-based least-square support vector machine (LSSVM), TSA-based multilayer perceptron neural network (TSA-MLPNN), whales optimization algorithm (WOA)-based LSSVM, and WOA-based MLPNN (WOA-MLPNN). The output of each forecasting model is aggregated through the Bayesian model averaging (BMA) method. The historical data of solar PV power and influencing factors obtained from the N-78 Griffith University Nathan campus have been employed to analyze and verify the performance of the proposed strategy. The simulation is carried out using MATLAB-2018 and the comparison of proposed ensemble strategy has been drawn with several competitive strategies. The simulation results indicate that the proposed forecasting strategy minimizes the root means square error by 29%, 33%, 27%, 33%, 58%, 12%, 26% and 43% in comparison with TSA-MLPNN, TSA-LSSVM, particle swarm optimization (PSO)-based artificial neural network (ANN), PSO-LSSVM, backpropagation neural network, ensemble strategy-1, ensemble strategy-2, and ensemble strategy-3, respectively.

KEYWORDS:

• Microgrid
• multilayer perceptron neural network
• least-square support vector machine
• solar PV power forecasting
• whales optimization algorithm

Acknowledgment

This work was funded by the Griffith University International Postgraduate Research Scholarship (GUIPRS) and Griffith University Postgraduate Research Scholarship (GUPRS) through Griffith University, Queensland, Australia, and the Researchers Supporting Project No. (RSP-2021/363), King Saud University, Riyadh, Saudi Arabia.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The work was supported by the Griffith University [GUIPRS and GUPRS]; King Saud University [Researchers Supporting Project No. (RSP-2021/363)]