PREDICTIVE ANALYTIC MODEL FOR ELECTRICAL CHILLER SYSTEM (ECC)

Abstract

District cooling plant is a very complex system consisting of various equipment. One of them is an electric centrifugal chiller that is widely used in the industry to supply chilled water to thermal energy storage (TES). This paper investigates the predictive analytics model of the electric chiller system using real operation data. Data pre-processing was performed to remove the outliers. Further to that, the machine learning model was used to develop an artificial neural network (ANN), whereby the best model with the lowest RMSE was obtained. Then, the ANN was used to correlate between the input and output variables to find the critical parameters which contributed to the Coefficient of Performance (COP). This model could also predict the output from the actual data of the COP. Lastly, the Shewhart control chart was utilised in the root cause analysis model (RCA) to detect the anomalies based on five critical parameters at the early stage before its failure. The supervised learning algorithms using the feedforward ANN model demonstrates the most accurate predictions compare to all models.

Keywords

• chiller
• matlab
• rmse
• artificial neural network
• root cause analysis model
• shewhart control chart