RESERVOIR OPERATING GUIDELINES FOR SERIES RESERVOIR IN CIRATA, SAGULING, AND JATILUHUR USING STOCHASTIC MODEL

• Main Author: Kurnia Utami, Dian
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/53161
• Institution: Institut Teknologi Bandung

The reservoir operating guidelines is made as a guide in running the reservoir. In a cascade or series reservoir, the operating pattern of the upstream reservoir will affect the operating pattern of the downstream reservoir. In West Java, there are three series reservoirs located in the Citarum river, namely Saguling, Cirata, and Ir. H. Djuanda (Jatiluhur). In 2019 there was a drought which impacted the operation of the three reservoirs so that the water surface conditions were below the normal operating limits for each reservoir. This study was conducted to obtain an overview of future discharge availability using the stochastic method and to simulate future reservoir operating guidelines. The stochastic method used in this study is ARIMA with the help of RStudio. RStudio itself is free software for computing statistics and graphics. The operating guidelines used for these three reservoirs is set at SNI Pd T-21-2004-A about Operation of Electric - Electric - Multipurpose Cascade Reservoirs. In general, the operation of a single reservoir is the same as a series reservoir, the difference is in the principle of equal sharing where the operation of the cascade reservoir is proportional to the effective volume of each reservoir to the total effective volume of all reservoirs. The operating pattern of each reservoir is made to describe three operations, namely dry conditions (dependable discharge 35%), normal conditions (dependable discharge 50%) and wet conditions (dependable discharge 65%). In these three reservoirs, the general operating pattern is to collect water from December to June (the reservoir will be close to full in May and full in June) and then the water level will shrink until November. Historical discharge data of Nanjung Station from 1919-2018 which was obtained from PUSAIR. In stochastic calculations, stationary data is needed. One of the first steps to getting stationary data is selecting the data to be used, where data that has the same characteristics tend to be selected. In this study, the data used were from 1974 - 2018 (540 monthly data). In the calculation of determining the ARIMA Model, training data and testing data ratio is 8: 2 because the data is quite long. Training data is set from January 1974 to December 2009 (432 monthly data) and testing data from the January 2010 to December 2018 (108 monthly data). After conducting preliminary tests (Augmented Dickey-Fuller Test, observations on the Plot Auto Correlation Function and Partial Auto Correlation Function, coefficient tests, Ljung-Box Test and Kolmogorov - Smirnov Test) on the training data it is found that the possible model is ARIMA (1,0,0) (1,0,1) (12). The model is then used to produce data with the same time span as the testing data and with a 95% confidence level of the forecast results, a correlation value of 0.51 and a Nash-Sutcliffe Efficiency (NSE) 0.084 value is obtained. The NSE value is close to 0, which means that the value generated by the model has the same mean value accuracy as the historical data. With the experimental results on the training data, the same ARIMA model is used for forecasting the next 5 years. The equation obtained is Y_t=6.4368+0.5593.Y_(t-1)+0.999.Y_(t-12)+a_t-0.9723a_(t-12) and from the result obtained, it shown that the result has not been able to describe the peak discharge. The results are then processed to obtain dependable discharge for dry, normal, and wet conditions. From the calculation of the operating pattern, there is a shortage in November 2020. However, the value is smaller than the river demand. So that in November this operating pattern was still able to supply the needs of PJT II. With the assumption that the water used by PJT II will return to the river, this is condition is permissible. In general, the elevation of each reservoir guidelines is close to full in May and will return to full condition in June. However, in Jatiluhur Reservoir these conditions are difficult to fulfil. However, the Jatiluhur Reservoir is designed as much as possible so that the reservoir is close to full in May. The lack of water generated in the calculation of the reservoir operation pattern is due to the prediction of the amount of incoming water that cannot describe the peak discharge of the incoming water. There is a difference in energy generated between the existing operating patterns and the calculated results. In general, the energy produced has increased. This is probably because the reservoir water level is kept relatively stable and the outflow discharge is not that different from the pattern shown in the data in the existing operating guidelines.