Skip to Main Content
Predictive reliability and availability calculations of power supply systems are generally too optimistic compared to values observed in the field. This is mainly due to lack of data for subset, and use of inappropriate reliability blocks diagrams (RBD). France Telecom R&D has tried to improve the methodology. Battery and public mains interrupt statistics have been studied, and then RBDs for equipment and energy chain are proposed to reach predictive reliability and unavailability values. First of all, a precise battery reliability model that fits field data is necessary. On the basis of recent studies and field failures observations made by France Telecom, a VRLA batteries string reliability model has been developed, taking into account: life time related to temperature of the cells; maintenance strategy of the strings; failures modes (including capacity loss, cell short circuits, open circuit equivalent to high impedance); and finally string failure criterion relative to mains power interrupt and power supply failures durations. In a second step, quality of the French electricity supplier is studied for high voltage (HTA). Failure rates and mean time to repair (MTTR) for very short, short and long power outages are analysed. Lognormal distributions laws are established for various kinds of built-up areas (Parisian area, 100,000 residents, 10,000 to 100,000 residents, less than 10,000 residents). Failure rates of the power supply chain equipment are then studied. MTTRs lognormal distributions of equipments placed in front of batteries are proposed, taking into account travel delay and repair time. At last, RBDs are proposed for usual 48 V DC architectures, allowing global predictive reliability and availability calculation for several autonomy batteries values.