Machine Learning Ready Dataset for Optimal PV-BESS-Grid System under Variable Demand Scenarios

Current dataset presents a comprehensive collection of performance results for a Hybrid Renewable Energy System (HRES) designed and simulated under the hot and dry climatic conditions of Rajasthan, India. The dataset was developed as part of the research project titled \u201cCharacterization of Electrical Load Demand to Design the Optimal Photovoltaic and Battery Energy Storage System (BESS) using Advanced Machine Learning Technique.\u201d The purpose of this dataset is to support reproducible research, enable benchmarking of predictive algorithms, and provide a foundation for future studies in renewable energy optimization and machine learning applications. The dataset is organized into four separate Excel files, each corresponding to a different random variability (RV) case of load demand: 5% Day-to-Day \/ 5% Time Step, 10%\/10%, 15%\/15%, and 20%\/20%. These RV cases capture the stochastic behavior of demand and generation fluctuations, thereby offering realistic conditions for model training and evaluation. Each file contains results from an optimized HRES configuration, integrating solar photovoltaic (SPV), lithium-ion battery storage (Mohit100LI), a grid connection, and an inverter system.The parameters included in the dataset cover both technical and economic aspects of the system. Technical indicators include PV installed capacity, maximum power point tracking (SPV-MPPT), inverter output (rectifier and inverter mean outputs), annual renewable fraction, and storage performance metrics such as nominal capacity, usable capacity, autonomy, and annual throughput. Economic indicators include Net Present Cost (NPC), Levelized Cost of Energy (LCOE), initial capital investment, and operating cost. Additionally, grid-related data such as annual energy purchased and sold are provided, reflecting the dynamic interaction between the HRES and external power supply. For example, under the VR (5\/5) case, the dataset records a renewable fraction of approximately 35%, with PV production exceeding 1.35 million kWh per year and grid purchases around 2.39 million kWh. Similar structured outputs are available for the other RV scenarios, allowing researchers to compare the sensitivity of HRES performance across different demand fluctuations.The dataset is particularly suited for applications in advanced machine learning, where predictive models can be trained to estimate optimal system sizing, cost-performance trade-offs, and demand-side management strategies. It also serves as a benchmark dataset for evaluating stochastic optimization techniques and forecasting algorithms in renewable energy studies. By making present dataset publicly available, the current study contributes to advancing open science in energy research. It enables academics, engineers, and policymakers to develop, test, and validate models for sustainable energy system planning under uncertainty, with practical relevance to regions experiencing similar climatic and demand variability conditions.