A novel prediction of the PV system output current based on integration of optimized hyperparameters of multi-layer neural networks and polynomial regression models

Abstract

The renewable energy system has yielded substantial enhancements to worldwide power generation. Therefore, precise prediction of long-term renewable energy conductivity is vital for grid system. This study introduces a new predictive output current for the photovoltaic (PV) system using actual experimental data. This research proposes three key contributions: The IMGO method is enhanced using several hybrid tactics to improve local search capabilities and increase exploration of significant regions within the feature space. Subsequently, the architecture of the multilayer feedforward artificial neural network is developed. The IMGO is employed to determine the appropriate hyperparameters of the model, ranging from the number of neurons in the hidden layers and learning rate. The Bayesian regularization backpropagation procedure is applied to update the weights and bias of the network. The proposed IMGO_MFFNN model is ultimately combined with Polynomial regression model to improve the predictability of the PV system. The experimental results demonstrated that the proposed IMGO algorithm is very effective in addressing complex problems with high accuracy, capability, and speedy convergence. The proposed hybrid IMGO_PMFFNN model proved its superior correlation evaluations, surpassing the performance of ant lion optimizer based on random forest (ALO_RF) model, two stages of ANN (ALO_2ANN) model, long short-term memory (LSTM), gated recurrent unit (GRU), extreme learning machine (ELM), least square support vector machine (LSSVM), and convolutional neural network (CNN) models. The MATLAB code of the IMGO is free available at: https://www.mathworks.com/matlabcentral/fileexchange/177214-improved-mgo-method.