Consumption Prediction on Steam Power Plant Using Data Mining Hybrid Particle Swarm Optimization (PSO) and Auto Regressive Integrated Moving Average (ARIMA)

Abstract

PT. PLN (Persero) is Indonesia state-owned company engaged in the electricity sector. In serving the needs of electricity in Indonesia, about 49.91% power plant is a Coal Fired Steam Power Plant (PLTU). To maintain coal operations, PLN has an rregulation to maintain the average coal stock limit of 22–26 days of operation; in addition PLN has also conduct a Coal Online Application (BBO) which records coal stock conditions in each power plant. Even though PLN have already create a regulations and online applications to monitor the power plants, coal stocks in several power plants are still experiencing a stock crisis caused by volatility in coal power plant needs due to changing capacity factor (CF) and specific fuel consumption (SFC). Therefore PLN need a model to estimate coal demand at the power plant so that it can help mmanagement make decisions early before the stock crisis.This study predicts coal consumption from time series CF data, SFC data; coal received and coal consumption taken from BBO applications for the period 2013 to 2017. The algorithm used is a hybrid between the Autoregressive Integrated Moving Average (ARIMA) model and the Particle Swarm Optimization (PSO). The results of applying data mining to predict coal needs at the coal-fired steam power plant Indramayu using ARIMA compared to hybrid ARIMA and PSO, has been proven increasing its accuracy by decreasing Mean Absolute Percentage Error (MAPE) from 9.03% to 4.69%.