요약2 |
In this study, Python is used to predict chiller energy consumption and improve the performance of forecasting models. The forecasting model used a random forest model and an artificial neural network model. To improve the performance of the forecasting model, the accuracy was evaluated by adjusting the number of inputs and the training data size. As a result, for the random forest model, the prediction performance allowed by the criteria was shown from the number of input variables to seven, and the CvRMSE improved the prediction performance by up to 23.91% by increasing the number of inputs. The training data size was shown to have acceptable predictive performance for the criterion at 80% and increased the training data size, improving the predictive performance by up to 14.08%. For artificial neural network (ANN) models, the predictive performance allowed by the criterion was shown to have a predictive performance with four inputs, and the CvRMSE improved by up to 14.90% by increasing the number of inputs. The training data size was shown to have acceptable predictive performance for the criterion at 70% and the maximum increase in the training data size resulted in improved predictive performance by up to 11.99% for CvRMSE. Comparing the two models, the artificial neural network model has better predictive performance than the random forest model, and the model for improving predictive performance is also more advantageous for the use of input variables and the adjustment of training data size. |