RELIABILITY CONTROL METHOD FOR DATA FROM AMR METERING DEVICES, BASED ON A FORECASTING POWER CONSUMPTION MODEL

Tymchuck Aleksey Igorevich

Purpose. Methods that are currently used in AMR to control the reliability of data, for monitoring the reliability of data from metering devices have disadvantages in terms of redundancy of electricity metering devices, redundant computing power and growth in the volume of system databases, which increases AMR implementation and maintenance cost. The reliability control of these metering devices using a mathematical apparatus will eliminate these shortcomings. Target. The target is to develop a method for monitoring the reliability of these metering devices in AMR that does not have the disadvantages of currently used methods, as well as to evaluate its effectiveness by creating on its basis an information system for AMR of an apartment building. Methods: the following methods were used to develop the technique and the information system that implements it – the Box-Jenkins method for building ARIMA models, the Foster-Stewart method to check if power consumption time series have a trend, the analysis of autocorrelation and partial autocorrelation functions to check power consumption time series for the presence of periodic interdependencies, the Dolado-Jenkins-Sausville-Riviero method and the extended Dickey-Fuller test to test power consumption time series for stationarity, the gradient descent method for training a forecasting power consumption neural network model. Novelty. The novelty of the method lies in the use of a forecasting power consumption model for reliability control of data from AMR metering devices. Results. The use of this method will reduce the search time in the AMR for electricity metering devices that transmit false data, which in turn will reduce power losses and the cost of the system maintenance Practical relevance. The developed method can be put into practice in an information system that interacts with the AMR databases. The forecasting power consumption model is built separately for each type of AMR control object. This article provides an example of using the method for an apartment building equipped with an AMR system.

Keywords:information system, power consumption forecasting, neural networks, ARIMA, AMR, data reliability control