Short-Term Missing Load Data Generation Method for Security Risk Analysis Based on Adaptive Rate Correction and Bidirectional ARIMA Model

Load forecasting is fundamental to the security risk analysis of the power system operation. However, the historical data required for load forecasting often suffers from missing values due to sensor failures, communication disruptions, and other factors, leading to a decline in forecasting accuracy. The ARIMA-based time series model has unique advantages in generating missing data. However, current research on ARIMA mainly focuses on data generation in a single data dimension and often overlooks the correction of anomalies in the original load data, which severely reduces the accuracy of data generation. To address this, a method for generating short-term missing load data based on adaptive rate correction and bidirectional ARIMA models is proposed. First, an adaptive rate detection model is used to detect anomalous data, and the weight coefficients are optimized by minimizing the Mean Squared Error (MSE) to correct the anomalies. Then, a bidirectional ARIMA model is established for missing data generation. Finally, using historical load data from a city as an example, the effectiveness of the proposed method is verified.