基于改进粒子群算法的油料消耗量优化组合预测方法

POL Consumption Optimal Combination Forecast Method Based on Particle Swarm Optimization

油料消耗量预测是实施精确保障的关键环节,而预测精度是油料消耗量预测的重要评价标准。为提高预测精度,以GM(1,1)、平滑指数法(ES)及广义回归神经网络(GRNN)3种单一预测模型为基础,构建一种基于改进粒子群算法确定权重分配的优化组合预测模型。以某部队执行某任务的消耗量数据为依据,分别采用3种单一预测模型和基于改进粒子群算法的优化组合预测模型对油料消耗量进行拟合预测。结果表明:后者较前3种单一预测模型的拟合精度更高、预测误差更小,充分验证了该组合预测模型的可靠性和精确性。

POL consumption forecast is the key link in precise support, and forecast accuracy is the important evaluation index of POL consumption forecast. In order to improve forecast accuracy, this paper firstly establishes an optimal combina- tion forecast model with improved particle swarm optimization to determine weight distribution on the base of three single forecast models ： grey model（GM） （ 1 , 1 ） , exponential smoothing（ES） and general regression neural network （ GRNN ）. Then, it forecasts POL consumption with three single forecast models and optimal combination forecast model based on im- proved particle swarm optimization respectively according to the consumption data of a troop. The result shows that the optimal combination forecast model has higher fitting accuracy and the less forecast error compare with three single forecast models, which has proved the reliability and accuracy of this combination forecast model.