This presentation shows how the AMM project deployed in France, known as Linky, improves knowledge of residential consumption, through the combined effects of an appropriate customer panel and a modelling method, adapted to more frequent reading of consumer indices. This presentation is one of the first applications of the rollout of "smart meters" in France regarding settlement management. The historical consumption profiles are built up using a residential panel managed by ERDF. The simulation of more frequent reading of meter indices has enabled us to study and test various scenarios covering the quality of electricity consumption estimates, depending on various criteria: meter reading frequency, profiling choice (adjusted, dynamic, flat), and the size of the sample used to build up the profile (for dynamic profiling). The quality of the estimates is based on a MAPE criterion (Mean Absolute Percentage Error) between an actual curve and an estimated one, on a half-hourly consumption basis. Studies conducted by ERDF have demonstrated that meter reading frequency and quality of dynamic profiling are two major levers to increase the accuracy of consumption modelling for a given client portfolio. A higher meter reading frequency provides enhanced accuracy but reaches a threshold at 5.5% (meter reading every day) on the adjusted profile, whereas the quality of dynamic profiling and hence the sampling process enables us to obtain greater accuracy (3% instead of 5.5% for a daily meter reading frequency with 1,000 customers). We can also conclude that it is possible to take advantage of the Linky meter without waiting for the end of deployment, using dynamic profiling. In fact, the dynamic profile enables us to increase the accuracy of the system, and hence its quality. This study provides ERDF with strong arguments when se
展开▼
