This dataset contains historical records of electricity demand in the Dominican Republic from January 2021 to December 2024, with hourly resolution. It was compiled to support short-term load forecasting of the National Interconnected Electric System (SENI). The dataset includes the total system demand in megawatts (MW), along with a set of exogenous variables commonly used in forecasting models. These variables include weather data retrieved from Open-Meteo (such as temperature and humidity), time-lagged demand features, and calendar-based indicators (e.g., weekends, holidays, month, hour). All data were collected from open sources, including the official website of the electricity market and system operator, the Organismo Coordinador (OC), as well as public meteorological APIs. The dataset is structured and cleaned to be directly usable for time series modeling applications. It can be reused by researchers, utility planners, and data scientists for benchmarking forecasting models, developing predictive tools, or supporting energy planning tasks in tropical, developing power systems. The data is provided in CSV format.

Data-constrained island power systems with increasing shares of variable renewable energy (VRE) face growing challenges in maintaining reliability while preserving market efficiency. Existing reserve sizing practices typically rely on either fixed deterministic rules or data-intensive probabilistic methods, both presenting practical limitations in Small Island Developing States (SIDS). This paper develops a market-based framework for the co-optimization of energy and operating reserves in low-inertia island power systems, in which reserve requirements are determined from historically observed extreme generation or load deviations that represent operationally validated high-risk system conditions, while reserve allocation and pricing emerge from the co-optimization process. By relying on observed operational variability, the proposed approach avoids explicit probabilistic uncertainty modeling while retaining sensitivity to system stress conditions. The approach is evaluated using a stylized island power system representative of Caribbean SIDS. Results show that reserve requirements are highly sensitive to operating conditions, reaching up to 26.7% of demand under high variability and significantly exceeding conventional fixed reserve criteria. The framework reduces non-served energy, improves reserve allocation efficiency, and generates scarcity-consistent reserve prices under stressed conditions. These findings demonstrate that the proposed methodology provides a practical intermediate solution between deterministic and probabilistic reserve sizing approaches while remaining suitable for data-constrained island power systems.

Data-constrained island power systems with increasing shares of variable renewable energy (VRE) face growing challenges in maintaining reliability while preserving market efficiency. Existing reserve sizing practices typically rely on either fixed deterministic rules or data-intensive probabilistic methods, both presenting practical limitations in Small Island Developing States (SIDS). This paper develops a market-based framework for the co-optimization of energy and operating reserves in low-inertia island power systems, in which reserve requirements are established using historically observed extreme generation or load deviations that represent operationally validated high-risk system conditions, while reserve allocation and pricing emerge from the co-optimization process. By relying on observed operational variability, the proposed approach avoids explicit probabilistic uncertainty modeling while retaining sensitivity to system stress conditions. The approach is evaluated using a stylized island power system representative of Caribbean SIDS. Results show that reserve requirements are highly sensitive to operating conditions, reaching up to 26.7% of demand under high variability and significantly exceeding conventional fixed reserve criteria. The framework reduces non-served energy, improves reserve allocation efficiency, and generates scarcity-consistent reserve prices under stressed conditions. These findings demonstrate that the proposed methodology provides a practical intermediate solution between deterministic and probabilistic reserve sizing approaches while remaining suitable for data-constrained island power systems.