Abstract
This paper describes development of failure prediction system for GHPs which is connected with the remote monitoring equipments. The remote monitoring center preserves and analyzes operating data to detect failure. The system developed has two distinct features. One is that it can diagnose GHPs when they are following cooling or heating load suitably. This can attain preventive maintenance prior to peak load season. The other feature is that diagnosis algorithm is designed to apply to various types of GHPs. The service using this prediction system has started in spring of this year by Tokyo Gas, and we are going to develop and improve algorithm every year.