Classification and Profiling of Electricity Consumption Patterns using Bayesian Networks

Electricity waste is becoming more prevalent due to the lack of awareness of consumers about their energy-wasting habits. The main reason is the absence of a feedback mechanism to make consumers aware of the energy waste that occurs. Looking at this need, this study aims to determine a mechanism that can identify whether consumers are saving or wasting energy. Energy consumption patterns were analyzed based on usage trends and building categories. The mechanism used applies Bayesian Network techniques in analyzing energy consumption patterns, especially in the classification or profiling of energy consumption. Driven by this mechanism, a monitoring system was also developed based on the Bayesian Network to further study the energy consumption of buildings based on different building profiles and user profiles. This study requires several activities to achieve the set objectives. Among the important activities include the data preparation process followed by data pre-processing. Following this phase, a classification process is then carried out i.e., classification analysis for Usage Trends, Building Profiles and User Profiles. Next, several parameters were set before the classification model was trained and tested. This study used a Naïve Bayes classifier to train and test the data set. After the Bayesian model is trained, it will be tested for its accuracy. If the accuracy is low, the parameter setting process will be repeated to adjust the best settings. The classification test will display the accuracy of the classification results based on ‘good classify’ and ‘bad classify’. In the analysis made, it was found that more correct classifications were successfully performed compared to incorrect classifications. This is by looking at the percentage of ‘good classify’ outputs that is higher than ‘bad classify’ outputs. Based on the generated models, the energy consumption monitoring system can use to analyze energy consumption behavior and in turn, can provide insights to resolve existing issues.