Setting the Time-of-Use Tariff Rates With NoSQL and Machine Learning to a Sustainable Environment

The electricity consumption will continue to increase despite the overall efforts and tendencies of changing the old appliances to less energy intensive ones. The advancements of Electric Vehicles (EV) and public mobility, electric heating, and the abundance of smart appliances that enhance the comfort of modern life lead to an increasing consumption trend. On the other hand, prosumers raising the quota of distributed generation and storage capacity will balance the electricity consumption trend. These changes at the consumption and generation level lead to the necessity to increase the awareness and incentive the consumers' behavior to flatten the consumption curve and improve the savings. Such objectives could be reached by properly setting the Time-of-Use (ToU) tariff rates to encourage the consumption at off-peak hours when the rates are lower and unstress the grid loading. In this paper, we propose a methodology for setting the Time-of-Use (ToU) tariff rates and peak/off-peak intervals using big data technologies and machine learning, and verify the assumptions considering the large volume of consumption data of over 4200 residential consumers recorded in a smart metering implementation trail period that took place in Ireland from January to December 2010. We calculate the contribution to the peak/off-peak of the total consumption and use it in setting the ToU tariff rates starting from the flat tariff. Then, the consumers' sensitivity to tariff change from flat to ToU is considered to identify the consumption change. The results show that using ToU instead of flat tariff, the peak is reduced in average by 5 to 7.5% and annual savings are around 4%. Also, by clustering the consumers a better allocation of the tariffs is possible. Thus, clustering is proposed considering the importance of the tariff allocation in Demand Side Management (DSM).