资源环境科技发展态势分析平台

The benefits and costs of power plants, including their environmental impacts, depend on their technology and on how much electricity each plant actually generates. However, plant-level generation data are not reported in most countries. This technical note documents methods to estimate the annual electricity generation of power plants for the Global Power Plant Database. We use distinct estimation models for different fuel types, including wind, solar, hydropower (hydro), and gas power plants. The methodology combines statistical regression with machine learning techniques. Explanatory variables include plant-level characteristics such as plant size and fuel type, and country-level characteristics, such as country- and fuel-specific average generation per megawatt of installed capacity. We show that fuel-specific models can provide more accurate results for wind, solar, and hydro plants. Estimations for natural gas plants also improve, but the error remains high, especially for smaller plants.

Related Resources

• Github: Code to estimate generation by plant, given the model
• Github: Code that describes training and tuning of the model, including figures and tables presented in the technical note
• Dataset: Global Power Plant Database