Boosted ANFIS model using augmented marine predator algorithm with mutation operators for wind power forecasting

There are several major available renewable energies, such as wind power which can be considered one of the most potential energy resources. Thus, wind power is a vital green source of electric power generation. The prediction of wind power is a critical issue to decrease the uncertainty of the energy systems. It is an essential process to balance energy demand and supply. The main objective of the current paper is to present an efficient prediction tool to estimate wind power using time-series datasets. We develop an enhanced variant of the ANFIS (adaptive neuro-fuzzy inference system) using the advances of metaheuristic (MH) optimization algorithms. We propose a new variant of the marine predator algorithm (MPA), called MPAmu, using additional mutation operators to augment the MPA to prevent its premature convergence on local optima. The developed MPAmu is used to optimize the ANFIS parameters and to boost its configuration process. We use well-known datasets collected from wind turbines located in France to evaluate the proposed MPAmu-ANFIS model using several evaluation metrics. Additionally, we compare the developed MPAmu-ANFIS to the traditional ANFIS and several modified ANFIS models using different MH algorithms. More so, we compare the developed model to other time-series prediction models, such as support vector machine (SVM), feedforward neural network, and long short term memory (LSTM). The findings of the current paper reveal that the application of MPAmu contributes significantly to boosting the prediction accuracy of the traditional ANFIS.

Automated summary

This paper presents an efficient prediction tool for estimating wind power using time-series datasets. By developing an enhanced variant of ANFIS and utilizing metaheuristic optimization algorithms, the proposed model shows significant improvements in prediction accuracy compared to traditional ANFIS and other time-series prediction models.