Forecasting Battery Capacity and Feasibility Using the Gaussian Process Regression Method

Abstract

The 110VDC batteries at the 150kV South Surabaya Substation have a shortage in the number of units. Therefore, they require extra supervision to ensure that protection and control equipment relying on DC power sources can operate normally during rectifier system outages, preventing potentially severe disruptions at the substation. The objective of this study is to use Matlab's forecasting degradation method for battery performance using Regression Learner, aimed at facilitating operators at the 150kV South Surabaya Substation. The research focuses on forecasting battery performance degradation using Gaussian Process Regression (GPR) with datasets obtained from observed discharging and charging tests compiled in Excel format. Data analysis techniques involve building a GPR model using Matlab software and comparing forecasted results with discharging test data over two years from PT. PLN (Persero). The study concludes that a 71% battery efficiency qualifies as sufficiently reliable backup power during AC or rectifier disruptions. This ensures continuous operation of protection and control equipment during blackouts, thereby preventing operational disruptions and serious safety issues.