一种基于预测-优化-闭环的多能源系统调度导向学习方法
By constructing a prediction model and clustering algorithm based on BiLSTM, and combining the error rate-additional cost curve, closed-loop optimization of prediction and scheduling is achieved, which solves the problem of the influence of prediction error asymmetry in multi-energy systems and improves the economy and resilience of the system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- XIAMEN UNIV
- Filing Date
- 2026-04-15
- Publication Date
- 2026-07-17
AI Technical Summary
The existing forecast-optimization-open-loop model cannot effectively distinguish the asymmetric impact of renewable energy output forecasting errors, resulting in suboptimal economic efficiency of multi-energy system dispatching decisions and a lack of a collaborative optimization mechanism between forecasting and dispatching.
A bidirectional long short-term memory network (BiLSTM) is used to construct a renewable energy output prediction model. Typical operating scenarios are extracted through clustering algorithms, and an error rate-additional cost curve is constructed to establish a closed-loop optimization framework for prediction and scheduling. The model parameters are updated using the backpropagation algorithm and gradient descent method to achieve closed-loop optimization of the prediction model.
It significantly reduced the operating costs of multi-energy systems, improved system flexibility and energy storage efficiency, enhanced resilience to the uncertainties of renewable energy, and reduced system operating costs by approximately 2.9%.
Smart Images
Figure CN122134156B_ABST