Air pollution forecasting application based on deep learning model and optimization algorithm

Abstract Air pollution monitoring is constantly increasing, giving more and attention to its consequences on human health. Since Nitrogen dioxide (NO 2 ) sulfur (SO are the major pollutants, various models have been developed predicting their potential damages. Nevertheless, providing precise predictions almost impossible. In this study, a new hybrid intelligent model based long short-term memory (LSTM) multi-verse optimization algorithm (MVO) has predict analysis air obtained from Combined Cycle Power Plants. proposed model, forecaster engine amount of produced NO SO by Plant, where MVO used optimize LSTM parameters in order achieve lower forecasting error. addition, evaluate performance, applied using real data Plant Kerman, Iran. The datasets include wind speed, temperature, , for five months (May–September 2019) with time step 3-h. tested two different types input parameters: type (1) includes lagged values output variables ); (2) just ). results show that higher accuracy than other combined benchmark (ENN-PSO, ENN-MVO, LSTM-PSO) considering network variables. Graphic abstract