Individual smart meter’s energy consumption forecasting for strategic decision making

Abstract

This paper analyzes the benefits of high frequency data obtained from smart meters readings, specifically from individual smart meter household’s energy consumption. The purpose is to learn the consumer’s behavior as leverage to improve the business strategy, the consumer’s experience and work towards a more efficient market. To tackle this, we performed exploratory data analysis techniques where we not only learned more about the customers, but we cleaned the data to perform load forecasting. For this last point we employed both statistical and machine learning techniques in order to help reach a consensus on the best option for this type of data. Results showed that customer characterization can be key for analyzing consumption behavior as well as a great strategy to improve forecasting. Also, the industry’s standard for forecasting performed very poorly compared to other techniques. From an industry standpoint this study shows how the use of data form smart meters can greatly benefit both the industry and the consumer. Energy consumption and, therefore, generation is a key player for the world economy whilst also being a scarce resource that we should learn to better manage; big data together with the right analytics tools can be a great place to start.