Handling Missing Data Using Standardized Load Profile (SLP) and Support Vector Regression (SVR)

In recent years, the research and application of data mining techniques encountered many difficulties and major challenges, including the lack of attribute values of data. There are many different reasons for this problem: the device is broken, the data is refused to protect the privacy, data entry mistakes or incidents occur during data transmission. In particular, the lack of data for electricity load research and forecasting is one of the problems for the electricity industry. Currently, the power companies are doing this by interpolating from the measured values of previous days and hours manually, which significantly affects the results of data analysis during the load forecasting process. The paper proposes a method of processing missing data by building a Standardized Load Profile (SLP) based on past load data, combining machine learning algorithms SVR (NN/RD) to rebuild the load curve, thereby we can estimate the data missed or not recorded during the measurement.