On-line fault-diagnosis study: Model-based fault diagnosis for ultracapcitors

The failure of ultracapacitors was significantly accelerated by elevated temperature or increased voltage. The voltages between different cells will be enlarged after a number of deep charging or discharging cycles due to the capacitance difference between the capacitor cells. This will accelerate the aging of the weak ultracapacitors and affect the output capability. In order to improve the reliability of system, a correct and timely failure prediction is essential. At present, the method of manual diagnostics is using more for maintaining the system of ultracapacitors. However, this is a time consuming procedure. And it also needs to provide specialist knowledge and the experience of maintenance personnel. In this paper, a model-based fault diagnosis system using the least squares parameter estimation method was proposed. The circuit model of ultracapacitor we used was a second order model, which has significantly higher precision (the error between the simulated and real results below to five percent). And the identification model derived from the difference equation of the circuit. Then the parameter of the identification model, which estimated by using least squares method, was used for the interpretation of the fault. Several faults of ultracapcitors have been identified on the system. The experiments are carried out to diagnose the types of faults. Experiments show the effectiveness of the proposed technique for real time applications.