Time Series Synthetic Load Data Generation for Forecasting and Power System Studies

In this manuscript, we structure a method for generating synthetic load data using historical timeseries from the Independent System Operator of New England (ISO-NE) for IEEE benchmarking power systems. The method starts with collecting regional electricity demand data from multiple ISONE zones and analyzing each region’s load contribution. These contributions are then mapped to bus locations within a standard IEEE test system. By comparing the regional weights from ISONE with those in the IEEE system, and appropriate scaling factors are determined for applying to match the timeseries with the test network. Load values are distributed for individual buses based on regional contribution and adjusted to show realistic system behavior. In addition, we include the exogenous weather variables to increase future applications. The dataset contains five years of hourly resolution load and weather data. The outcomes of our dataset provides foundation for data-driven studies, such as load forecasting, renewable integration analysis, timeseries power or optimal power flow forecasting and state estimation. Existing general methods do not address bus level of spatial and temporal data generation required for realistic power system modeling. Therefore, this framework supports regionally separated data-driven testing environments for validating energy system models and predictive learning techniques.