基于混沌CSO优化时序注意力GRU模型的超短期风电功率预测

High-precision wind power prediction is crucial for the grid-connected operation of wind power. In order to extract the time information implied in the wind power input sequence, a prediction model based on gated recurrent units is established; and a temporal attention mechanism is introduced on the input side of the model to improve the sensitivity of the model to key historical time nodes by weighting with the input. To accelerate the convergence of the model, the dynamic chaotic crisscross optimization algorithm is used to optimize the weights and thresholds of the prediction model in the early stage of training; at the same time, by constructing a regular term with multiple indicators acting together and jointly with the parameters to be optimized as the objective fitness function, the problem of model generalization in the optimization process is avoided. Experiments are carried out with the measured data of a certain wind farm with collection intervals of 1 hour and 10 minutes, and the results show that the performance of the proposed combined prediction method is better than other comparison models, and its effectiveness is verified.

高精度的风电功率预测对风电的并网运营至关重要。为提取风电功率输入序列隐含的时间信息,建立以门控循环单元为基础的预测模型;并在模型输入侧引入时序注意力机制,通过与输入进行加权的方式提高模型对关键历史时间节点的敏感性。为加速模型收敛,在训练的早期利用动态混沌纵横交叉算法优化预测模型的权值和阈值;同时,通过构造多指标共同作用并联合待优化参数的正则项作为目标适应度函数,以避免优化过程中模型泛化性问题的出现。以某风电场采集间隔为1h和10min的实测数据进行实验,结果表明所提组合预测方法性能优于其他对比模型,并对其有效性进行了验证。