Long Short-term Memory Forecasts of Biomass Power of an Installed Waste-water Treatment Plant

In this study, the parameters including power generations as well as waste-water parameters obtained from a waste-water treatment facility were forecasted, using long short-term memory (LSTM) method. Accordingly, a total of 445 data, that is found in the data folder of power generation (P) and data cluster involving of physical and chemical parameters has been utilized, in the estimations depending on LSTM tool. In this context, every instantaneous data of these data clusters formed of 445 reading points corresponds to daily mean energy generation (P) obtained from the waste water treatment facility’s gas turbines as well as corresponds to physical and chemical parameters involving of conductivity (σ), degree of acidity (pH), temperature (T), and the daily cumulative volumetric flow of the generated methane gas that is burned in the gas generator (Q). All in all, based on the power generation (P) outcomes, the statistical error results have indicated that the best forecasting result was obtained for LSTM at the hidden layer number of HL=100. Namely, at this HL number, error results of mean absolute error (MAE), root mean square error (RMSE), and the correlation coefficient (R) were reported to correspond 2.4956 MWh/day, 3.7002 MWh/day, and 0.9541, respectively.&nbsp;&nbsp;<br>