Knowledge based VRLA battery monitoring and health assessment

This paper discusses an approach for utilising on-line continuous measurements including planned tests and unplanned events for the assessment of real-time battery health. Both direct health dependent parameters such as charge and capacity, and indirect operational parameters such as voltage, temperature and discharge rate are taken into consideration. The aim is to (a) maximise the confidence in the real time measurements and trend analysis as well as planned short discharge tests, (b) reduce the frequency of medium discharge tests, and (c) eliminate the need for on-site regular full discharge tests. To meet these targets the various knowledge elements that contribute to the assessment of battery health should be taken into consideration. This includes assessing individual and collective knowledge on bloc voltage, temperature, rate, charge and capacity. They are also considered on both a real time and a trend basis. In addition bloc behaviour is analysed individually and in reference to peer blocs. The analysis is based on capacity, which dominates the interpretation of battery state of health. Behaviour of base parameters during normal operation provides an indication of the influence on the state of health. This allows for adjustments to the state of health between successive capacity estimations. The capacity estimation model is made adaptive resulting in an increase in reliability of estimation with battery age. Teaching information is captured from any planned or unplanned discharge. Both hard and soft computing are utilised for supporting the analysis and updating process