This paper deals with the problem of iterative learning control for a class of discrete singular systems with fixed initial shift. According to the characteristics of the discrete singular systems, a closed-loop learn...This paper deals with the problem of iterative learning control for a class of discrete singular systems with fixed initial shift. According to the characteristics of the discrete singular systems, a closed-loop learning algorithm is proposed and the corresponding state limiting trajectory is presented.It is shown that the algorithm can guarantee that the system state converges uniformly to the state limiting trajectory on the whole time interval. Then the initial rectifying strategy is introduced to the discrete singular systems for eliminating the effect of the fixed initial shift. Under the action of the initial rectifying strategy, the system state can converge to the desired state trajectory within the pre-specified finite time interval no matter what value the fixed initial shift takes. Finally, a numerical example is given to illustrate the effectiveness of the proposed approach.展开更多
以风电为代表的新能源发电装机容量占比逐年增长,精确的风电功率超短期预测对提高风能利用率、助力双碳实现有重要意义。该文提出一种基于多元注意力框架与引导式监督学习的闭环风电功率超短期预测策略,从特征筛选、模型优化、策略改良...以风电为代表的新能源发电装机容量占比逐年增长,精确的风电功率超短期预测对提高风能利用率、助力双碳实现有重要意义。该文提出一种基于多元注意力框架与引导式监督学习的闭环风电功率超短期预测策略,从特征筛选、模型优化、策略改良3个角度全面提高预测准确性与模型智能性。首先,采用动态权重特征选择算法、孤立森林算法以及最邻近节点算法筛选并处理数据,便于预测模型更好把握其中特征;其次,对长短期记忆(long short term memory,LSTM)基模型多角度优化,并根据基模型中不同信息的特点,构建关于LSTM的多元注意力框架(Multielement-attention-LSTM),将此框架用于对LightGBM集成学习模型的引导,并通过多种可视化方法提高了模型可解释性;最后,将Bland-Altman应用于模型输出与实际风电出力一致性检验,在预测数据与实际数据交互的基础上实现训练–预测闭环机制。仿真结果表明,所构建的Multielement-attention-LSTM框架具有提高模型预测精度的作用,且闭环更新机制具备合理性。展开更多
基金supported in part by the National Natural Science Foundation of China under Grant Nos.61374104 and 61773170the Natural Science Foundation of Guangdong Province of China under Grant No.2016A030313505
文摘This paper deals with the problem of iterative learning control for a class of discrete singular systems with fixed initial shift. According to the characteristics of the discrete singular systems, a closed-loop learning algorithm is proposed and the corresponding state limiting trajectory is presented.It is shown that the algorithm can guarantee that the system state converges uniformly to the state limiting trajectory on the whole time interval. Then the initial rectifying strategy is introduced to the discrete singular systems for eliminating the effect of the fixed initial shift. Under the action of the initial rectifying strategy, the system state can converge to the desired state trajectory within the pre-specified finite time interval no matter what value the fixed initial shift takes. Finally, a numerical example is given to illustrate the effectiveness of the proposed approach.
文摘以风电为代表的新能源发电装机容量占比逐年增长,精确的风电功率超短期预测对提高风能利用率、助力双碳实现有重要意义。该文提出一种基于多元注意力框架与引导式监督学习的闭环风电功率超短期预测策略,从特征筛选、模型优化、策略改良3个角度全面提高预测准确性与模型智能性。首先,采用动态权重特征选择算法、孤立森林算法以及最邻近节点算法筛选并处理数据,便于预测模型更好把握其中特征;其次,对长短期记忆(long short term memory,LSTM)基模型多角度优化,并根据基模型中不同信息的特点,构建关于LSTM的多元注意力框架(Multielement-attention-LSTM),将此框架用于对LightGBM集成学习模型的引导,并通过多种可视化方法提高了模型可解释性;最后,将Bland-Altman应用于模型输出与实际风电出力一致性检验,在预测数据与实际数据交互的基础上实现训练–预测闭环机制。仿真结果表明,所构建的Multielement-attention-LSTM框架具有提高模型预测精度的作用,且闭环更新机制具备合理性。
