Scholarlink

To improve power and load forecasting accuracy and optimize multi-objective scheduling in renewable energy systems, we propose a Multi-Task Temporal Decomposition-based Power and Load Forecasting Model (MTD-PLF) and a coordinated control strategy. The MTD-PLF model extracts seasonal and trend features from time-series data using grouped convolution and attention mechanisms, while multi-task learning enables high-precision joint forecasting of photovoltaic power, wind power, and load demand. Based on the forecasts, a multi-objective optimization model is designed to minimize operational costs, maximize renewable energy utilization, and reduce load fluctuations. An improved Non-dominated Sorting Genetic Algorithm III (NSGA-III) identifies Pareto-optimal solutions, and the Analytic Hierarchy Process (AHP) selects the optimal scheduling strategy. Experiments on simulated data from Anqing and Lu’an, Anhui Province, for January 2025 show that MTD-PLF outperforms benchmark models in Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). The proposed scheduling strategy ensures efficient renewable energy integration, cost reduction, and load stability, offering a novel approach for the secure and efficient operation of renewable energy systems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-task Temporal Decomposition-Based Power and Load Forecasting Model for Coordinated Control of Distributed Energy Resources

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Li Zhang,
  • Jiwen Wang,
  • Wei Zhang,
  • Wei Wang,
  • Lu Chen,
  • Zheng Xu,
  • Wei Fang

摘要

To improve power and load forecasting accuracy and optimize multi-objective scheduling in renewable energy systems, we propose a Multi-Task Temporal Decomposition-based Power and Load Forecasting Model (MTD-PLF) and a coordinated control strategy. The MTD-PLF model extracts seasonal and trend features from time-series data using grouped convolution and attention mechanisms, while multi-task learning enables high-precision joint forecasting of photovoltaic power, wind power, and load demand. Based on the forecasts, a multi-objective optimization model is designed to minimize operational costs, maximize renewable energy utilization, and reduce load fluctuations. An improved Non-dominated Sorting Genetic Algorithm III (NSGA-III) identifies Pareto-optimal solutions, and the Analytic Hierarchy Process (AHP) selects the optimal scheduling strategy. Experiments on simulated data from Anqing and Lu’an, Anhui Province, for January 2025 show that MTD-PLF outperforms benchmark models in Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). The proposed scheduling strategy ensures efficient renewable energy integration, cost reduction, and load stability, offering a novel approach for the secure and efficient operation of renewable energy systems.