Model -wise ADF test results.

Amidst a growing need for effective energy management, government policies increasingly rely on accurate electricity consumption forecasts to make informed decisions on renewable energy adoption. This study investigates the predictive capabilities of night light satellite imagery in forecasting electricity usage in India, aligning with Sustainable Development Goals 7 and 10. Utilizing data from the VIIRS satellite and NASA’s Black Marble product, the research employs various LSTM models to analyse electricity consumption trends. Additionally, state-wise analyses have been conducted by applying k-means clustering to capture spatial consumption variations. By demonstrating the strong correlation between night lights and electricity consumption, the study emphasizes the utility of satellite imagery for actionable insights into energy dynamics. The results emphasize the viability of night light data as a dependable indicator of electricity demand, with MAPE values below 10% and RMSE values below 20 MU. It also highlights the transformative impact of remote sensing technologies in advancing sustainable development agendas and highlights the pivotal role of night light imagery in energy forecasting initiatives.

在高效能源管理需求日益凸显的当下，各国政府愈发依赖精准的电力消费预测，为可再生能源推广的决策制定提供科学支撑。本研究以印度为研究区域，探究夜间灯光卫星影像在电力消费量预测中的应用潜力，契合联合国可持续发展目标7与10。本研究采用VIIRS卫星与美国国家航空航天局（NASA）的Black Marble产品数据，结合多种长短期记忆网络（LSTM）模型对电力消费趋势展开分析。此外，本研究通过k-means聚类算法开展分邦分析，以捕捉电力消费的空间异质性。本研究证实了夜间灯光与电力消费间存在强相关性，强调卫星影像可为能源动态分析提供可落地的决策参考价值。研究结果表明，夜间灯光数据可作为可靠的电力需求指示器，其平均绝对百分比误差（MAPE）低于10%，均方根误差（RMSE）低于20 MU。本研究同时凸显了遥感技术对推进可持续发展议程的变革性作用，以及夜间灯光影像在电力预测工作中的关键支撑价值。