Personalized Electricity Tariff Recommendation Method for Residential Customers Lacking Historical Metering Data Incorporating Customer Profiles and Behavioral Changes

Effective energy management from the demand side and smart meters play an important role in achieving carbon neutrality. The government and utilities in South Korea are working to expand the installation of smart meters and develop time-of-use (TOU) tariffs for residential customers. Although these efforts have expanded the selection of tariffs for customers, it has become increasingly difficult to determine electricity tariffs. Therefore, some studies have been conducted to recommend tariffs for residential customers with historical metering data. However, recommending tariffs for customers who have recently installed smart meters or have failed to obtain historical metering data is a challenging task. Therefore, this paper presents a systematic method to estimate energy consumption patterns and incorporate behavioral changes based on the input profiles of residential customers for personalized electricity tariff recommendations. The proposed method attempts to predict bills by estimating energy consumption patterns using customer profiles. It is designed to reflect the behavioral changes in each pattern caused by the TOU tariff in predicting bills. In addition, the bill prediction model uses deep learning-based matrix factorization with the estimated patterns to improve bill prediction performance. The proposed method increases the probability of selecting TOU tariffs that reduce bills. It can be used as an effective tool for recommending tariffs to residential customers, helping them reduce their bills based on the prediction results. An increase in the number of customers selecting TOU tariffs also contributes to improving the stability and reducing the capital investment cost of the power system through peak shaving.