Event-based with linear prediction electricity meters reading in building automation

Abstract

The most popular method in remote data reading from sensors and meters is the time-driven scheme. However, event-driven schemes seem more efficient, as they reduce the number of samples as well as energy consumption, while achieving the same ability of reconstruction of real continuous waveform from discrete samples with the same accuracy. Even greater benefits can be achieved using event-driven with prediction data reading schemes. In the event-driven schemes the number of needed samples depends strongly on the dynamics of change of measured value. In the paper the influence of using an event-driven with linear prediction scheme on quantity and quality of data was tested by comparing with other adequate methods of sampling. The investigation was based on real data of energy usage by different loads in the sample office building. For the assessment the `influence and benefit factors' proposed by authors were used. Also an algorithm and a computer program for data analysis were proposed.