The prediction accuracy of the MLP ANN AE model was evaluated using R2, RMSE, and MSE shown in Table 3. The optimum MSE of 0.00075 was determined during the validation phase of the AE model. The highest MSE of 0.0025 was determined during the testing phase of the model. Although the testing phase produced the lowest performance, the model performance was still acceptable. This means that the distance between the data points and the regression line are close to each other. The difference between the highest and lowest MSE was 70%, which was a significant difference. The relationship between MSE and number of iterations is shown in Figure 8. Similar results were obtained by the following scholars (Kundu et al. 2013; Tumer & Edebali 2015; MG et al. 2018; Bekkari & Zeddouri 2019; Saleh & Kayi 2021). Saleh & Kayi (2021) applied ANN algorithm to model the prediction of COD and the model produced MSE of 0.000441. MG et al. (2018) applied ANN algorithm in modelling WWTP performance and the model produced MSE value of 0.311. Similarly, Kundu et al. (2013) applied ANN algorithm to model the biological removal of organic carbon and nitrogen in wastewater and the model produced MSE of 2.81. Bekkari & Zeddouri (2019) applied ANN algorithm to predict effluent COD in wastewater and the model produced MSE of 0.056. Tumer & Edebali (2015) applied ANN algorithm to model the biological treatment process and the model produced MSE of 0.0041.
Table 3

MLP ANN energy consumption model performance

TrainingValidationTesting
R2 0.963 0.963 0.939 
MSE 0.0012 0.00075 0.0025 
RMSE 0.0353 0.0274 0.05 
TrainingValidationTesting
R2 0.963 0.963 0.939 
MSE 0.0012 0.00075 0.0025 
RMSE 0.0353 0.0274 0.05 
Figure 8

Mean squared error performance at each iteration.

Figure 8

Mean squared error performance at each iteration.

Close modal
Close Modal

or Create an Account

Close Modal
Close Modal