基于DE—BP算法的空调负荷预测研究

Air conditioning load prediction based on DE—BP algorithm

建立了基于BP神经网络理论的空调系统负荷预测模型。针对BP神经网络参数优化过程中容易陷入局部最优的缺陷,采用差异演化算法(differential evolution algorithm,DE)对其进行优化,以提高预测精度。结合具体实例进行空调冷负荷预测,并与采用遗传算法、蚁群算法、粒子群算法对BP神经网络进行参数优化的仿真实验结果对比表明,由DE—BP算法所具有较好的预测性能。

Based on BP neural network theory,the model to predict air conditioning load was established. In order to optimize the behavior of BP neural network,DE algorithm was introduced into classic BP neural network. Using this algorithm to predict a real example and compare with BP model optimization method based on GA（ Genetic Algorithm） ,ACO（ Ant Colony Optimization） and POS （ Partial Swarm Optimization） demonstrate an improvement of generalization performance.