Abstract
<title>Abstract</title> <p> The rapid transformation of modern power systems through renewable energy integration, distributed energy resources, battery energy storage, electric vehicle charging, and dynamic electricity markets has significantly increased the need for accurate, reproducible, and standardized forecasting methodologies. Although artificial intelligence has demonstrated strong potential for smart-grid forecasting, objective comparison across forecasting models remains challenging due to inconsistent datasets, heterogeneous experimental protocols, and limited availability of open benchmarking frameworks. This paper presents <bold>BQEB ForecastBench</bold> , an open benchmarking framework for evaluating artificial intelligence models on smart-grid forecasting tasks using the publicly available <bold>BQEB-Data v1</bold> benchmark dataset. Developed as the forecasting evaluation component of the broader BIO-Quantum Energy Brain (BQEB) research initiative, the benchmark establishes standardized forecasting tasks, chronological data partitioning, transparent evaluation metrics, and reproducible experimental procedures. Two operationally significant forecasting problems are investigated: next-hour electricity load forecasting and day-ahead electricity price forecasting. Three baseline machine learning models—Linear Regression, Random Forest, and Gradient Boosting—are evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and the coefficient of determination (R²). Experimental results demonstrate that Linear Regression achieves the best overall performance for both forecasting tasks, obtaining an RMSE of <bold>9.94 MW</bold> and <bold>R² = 0.9288</bold> for load forecasting, and an RMSE of <bold>4.39 USD/MWh</bold> and <bold>R² = 0.2078</bold> for electricity price forecasting. The results further highlight the greater stochastic complexity of electricity price prediction compared with demand forecasting. By providing publicly accessible benchmark data, standardized evaluation protocols, and reproducible baseline results, BQEB ForecastBench establishes a reusable reference framework for future research in machine learning, deep learning, intelligent energy forecasting, and trustworthy artificial intelligence for next-generation smart-grid systems. </p>