Decision tree-based approach for online management of pem fuel cells for residential application

• Main Author: Mohamed, Mohd Rusllim
• Format: Thesis
• Language: English; English; English
• Published: 2004
• Subjects: TK Electrical engineering. Electronics Nuclear engineering; TK2896-2985 Production of electricity by direct energy conversion
• Online Access: http://eprints.uthm.edu.my/7679/1/24p%20MOHD%20RUSLLIM%20MOHAMED.pdf; http://eprints.uthm.edu.my/7679/2/MOHD%20RUSLLIM%20MOHAMED%20COPYRIGHT%20DECLARATION.pdf; http://eprints.uthm.edu.my/7679/3/MOHD%20RUSLLIM%20MOHAMED%20WATERMARK.pdf; http://eprints.uthm.edu.my/7679/

This thesis demonstrates a new intelligent technique for the online optimal management of PEM fuel cells units for on site energy production to supply residential utilizations. Classical optimization techniques were based on offline calculations and cannot provide the necessary computational speed for online performance. In this research, a Decision Tree (DT) algorithm was employed to obtain the optimal, or quasi-optimal, settings of the fuel cell online and in a general framework. The main idea was to employ a classification technique, trained on a sufficient subset of data, to produce an estimate of the optimal setting without repeating the optimization process. A database was extracted from a previously�performed Genetic Algorithm (GA)-based optimization that has been used to create a suitable decision tree, which was intended for generalizing the optimization results. The approach provides the flexibility of adjusting the settings of the fuel cell online according to the observed variations in the tariffs and load demands. Results at different operating conditions are presented to confirm the high accuracy of the proposed generalization technique. The accuracy of the decision tree has been tested by evaluating the relative error with respect to the optimized values. Then, the possibility of pruning the tree has been investigated in order to simplify its structure without affecting the accuracy of the results. In addition, the accuracy of the DTs to approximate the optimal performance of the fuel cell is compared to that of the Artificial Neural Networks (ANNs) used for the same purpose. The results show that the DTs can somewhat outperform the ANNs with certain pruning levels.