Model Prediksi Angka ISPU CO dan PM10 dengan Faktor Meteorologis di DKI Jakarta Menggunakan LSTM

Abstract

Jakarta is one of the areas in Indonesia that has poor air quality. This makes air quality in Jakarta a problem that deserves serious attention. Currently, the air quality standard index that is officially used in Indonesia is the Air Pollutant Standard Index (ISPU). ISPU calculations are carried out on 7 parameters, namely PM10, PM2.5, NO2, SO2, CO, O3, and HC. Of all ISPU parameters, the parameters that have a negative impact in a relatively small range are CO and PM10. One way to solve the problem of air pollution in DKI Jakarta is to make temporal predictions of air quality using data from the past. This study aims to create PM10 and CO prediction models in DKI Jakarta using LSTM with meteorological parameters as predictors, namely wind speed, solar radiation, humidity and air temperature to see how these variables affect the model. The data used in this study are ISPU data and weather data from January 1, 2017 to March 31, 2021. This study has successfully built a prediction model for CO and PM10 indices in 5 air quality monitoring stations in DKI Jakarta using LSTM. The modeling was carried out using two scenarios: modeling using meteorological predictors and without meteorological predictors in five air quality monitoring stations in DKI Jakarta. The results obtained show that the use of meteorological predictors in the CO prediction model does not affect the model. Still, the use of meteorological predictors has an influence on the PM10 prediction model, where prediction models using meteorological predictors produce a smaller RMSE and a stronger correlation coefficient than modeling. PM10 without using a meteorological predictor. This is because PM10 has a stronger correlation with the meteorological variables used than CO. The prediction model for CO and PM10 obtained in this study has not been able to predict extreme values because of the outlier found in each SPKU and the model can only be used to predict CO and PM10 index on the next day. Therefore, a prediction model for the CO at SPKU DKI 2 and the PM10 at SPKU DKI 1 was made to be compared with the model before adding the number of lags. The RMSE obtained shows that the addition of lags in the model decreases the model’s performance.