Analysis of Data Quality Impact on Load Forecasting

Accurate load forecasting is essential for the stable operation and economic efficiency of power systems, particularly with the rise and renewable energy integration. However, the quality of load data plays a critical role in determining the accuracy of forecasting models. This paper explores the significant impact of data quality on load forecasting performance. Common issues such as missing values, outliers, and noise can distort historical load patterns, leading to suboptimal forecasts. To this end, this study examines the effects of data quality degradation on performance of various forecasting models, including traditional time-series models like ARIMA, Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNN). Specifically, we evaluate the performance of these models when trained on datasets with varying degrees of data corruption. Moreover, the study proposes methods for improving data quality, including imputation techniques for missing values, outlier detection and removal methods, and noise reduction strategies. Specifically, advanced imputation methods, such as K-Nearest Neighbors (KNN) and LSTM-based imputation, demonstrate superior performance in restoring data integrity compared to simpler methods like mean imputation. Finally, experimental results show the importance of ensuring high-quality data collection and processing in load forecasting.