| Summary: | 碩士 === 國立臺北科技大學 === 資訊工程系 === 107 === In this study, the operating parameters and operating data of the collected water-cooled system were used for predictive analysis. The research method uses the theory and technology of many different scholars to compile a set of COP (Coefficient of Performance) formula for calculating the performance of water-cooled. As long as the evaporator saturation temperature, the condenser saturation temperature, the compressor outlet temperature, the evaporation pressure and the condensing pressure are used, the refrigerant enthalpy value at the evaporator outlet on the Mollier diagram, the condenser outlet refrigerant enthalpy value, and the compressor are calculated. By exporting the refrigerant depreciation, it is possible to calculate the performance COP of water-cooled.
The calculated evaporator outlet refrigerant enthalpy value, condenser outlet refrigerant enthalpy value, and compressor outlet refrigerant enthalpy value data can be used to train the LSTM (Long Short Term Memory Network) prediction model. By observing the past time series data, LSTM can predict the performance index trend of the future ice water host, the evaporator outlet refrigerant enthalpy characteristic curve, the condenser outlet refrigerant enthalpy characteristic curve and the compressor outlet refrigerant enthalpy characteristic curve. We improved the LSTM network to improve the accuracy. The performance forecast of the ice water host is conducive to the development of future fault prediction, saving additional maintenance costs and energy consumption.
|
|---|
