Inflow Prediction System and Grindulu Hydropower Production Optimization using The Bi-LSTM Seq2Seq Quantile Regression Model

Abstract

The Grindulu PS Hydroelectric Power Plant is one of the hydroelectric power plants that relies on the flow of the Grindulu River as its primary energy source. Due to the irregularity of water discharge, particularly at its lowest point during the dry season, the Grindulu PS Hydroelectric Power Plant is planned to be constructed using variable speed drive (VSD) technology [1], [2]. The objective of this study is to analyze the hydrological and hydraulic data of the Grindulu River to determine the annual water discharge profile, particularly during the dry season; to analyze the minimum potential for electricity generation from the Grindulu PS Hydroelectric Power Plant based on the lowest available discharge; and to analyze the Variable Speed Drive (VSD) system of the Grindulu PS Hydroelectric Power Plant in optimizing electricity generation under minimum discharge conditions during the dry season. This study is a quantitative-predictive analysis aimed at modeling inflow predictions at the proposed site for the Grindulu Hydropower Plant, which is planned to utilize a pumped-storage system. The prediction model used is the Bi-Directional Long Short-Term Memory Sequence-to-Sequence (Bi-LSTM Seq2Seq) with a Quantile Regression (QR) approach to generate inflow estimates at various levels of uncertainty (q10, q50, and q90) [3]. The research results show that the flow characteristics of the Grindulu watershed are dominated by relatively constant low-flow conditions throughout the year, particularly during the dry season, as indicated by the q50 and q90 flow values, which are both 1.021 m³/s. This indicates that the flow is dominated by baseflow and that high-flow events occur only for limited periods. With a minimum flow potential and a reservoir system design flow of 242 m³/s, a generation capacity of +1,000 MW can be achieved by utilizing the water stored in the upper reservoir. At the Grindulu River Basin hydroelectric power plant, the VSD system allows for more flexible and efficient operation of the pump-turbines, ensuring that power generation remains normal even under minimum flow conditions. The analysis results indicate that the hydraulic technical parameters that need to be considered in supporting the implementation of the VSD system under minimum flow conditions are water flow, effective head, reservoir water level fluctuations, turbine efficiency, and flow stability [2].