【摘要】 针对燃料电池城市客车车速预测精度低的问题，提出改进海鸥优化算法(ISOA)和长短期记忆神经网络(LSTM)相结合的车速预测模型。以标准工况驾驶循环数据库为训练集，以中国典型城市公交循环工况为测试集，使用引入莱维飞行、柯西变异等策略改进后的海鸥优化算法，确定LSTM最优参数，建立基于城市道路的ISOA-LSTM燃料电池城市客车车速预测模型，与LSTM模型、SOA-LSTM模型和GWO-LSTM模型进行对比。结果表明：基于ISOA-LSTM的车速预测模型的均方根误差为1.965,平均绝对误差为1.570,决定系数为0.983,预测精度更高。更多还原

【Abstract】 To address the issue of low prediction accuracy in fuel cell city buses speed forecasting, a speed prediction model combining improved Seagull Optimization Algorithm(ISOA) and Long Short-Term Memory neural network(LSTM) is proposed. The standard driving cycle database is used as the training set, while the China typical city bus driving cycle serves as the test set. The seagull optimization algorithm improved by introducing Levy flight, Cauchy mutation and other strategies is employed to determine the optimal parameters of the LSTM. An ISOA-LSTM fuel cell city bus speed prediction model was established, and compared with LSTM model, SOA-LSTM model and GWO-LSTM model. Experimental results demonstrate that the proposed ISOA-LSTM model achieves superior prediction accuracy, with a root mean square error(RMSE) of 1.965, mean absolute error(MAE) of 1.570, and coefficient of determination(R~2) of 0.983.更多还原