Machine Learning Approach to Predict the Performance of a Stratified Thermal Energy Storage Tank at a District...

Area of Science:

• Energy Systems Engineering
• Artificial Intelligence in Energy
• Thermal Engineering

Background:

• District cooling plants rely heavily on thermal energy storage (TES) tanks for efficient energy management.
• Accurate monitoring of TES performance is essential for optimizing plant operations.
• Existing methods for TES performance evaluation include numerical and analytical approaches.

Purpose of the Study:

• To explore the prediction of TES tank thermocline thickness using machine learning models.
• To compare the performance of Artificial Neural Network (ANN), Support Vector Machine (SVM), and K-Nearest Neighbor (KNN) for thermocline thickness prediction.
• To identify the most accurate model for predicting TES tank thermocline thickness.

Main Methods:

• Collected one year of temperature data from 14 sensors in a district cooling plant.
• Developed ANN, SVM, and KNN models using 70% of the 263 selected data points for training and 30% for testing.
• Applied data normalization, moving average, and median filters; utilized K-fold cross-validation and hyperparameter optimization.

Main Results:

• The optimal ANN architecture (14-10-1) achieved an R-Squared value of 0.9 and minimum mean square error.
• Prediction accuracy: ANN (92%), SVM (89%), and KNN (96.3%).
• KNN demonstrated superior performance compared to ANN and SVM in predicting thermocline thickness.

Conclusions:

• K-Nearest Neighbor (KNN) models exhibit higher accuracy in predicting TES tank thermocline thickness.
• Machine learning, particularly KNN, offers a promising approach for enhancing TES performance monitoring in district cooling.
• The findings provide valuable insights for optimizing energy management in district cooling systems.