【作者】 王淑玲；

【导师】 邢棉；

【作者基本信息】 华北电力大学（河北） ， 应用数学， 2008， 硕士

【摘要】 电力系统负荷预测对电力系统的运行、控制和计划都有非常重要的影响,其预测精度直接影响到了电网及各发电厂的经济效益。应用神经网络进行电力负荷预测已经非常普遍。BP神经网络在应用中存在一些缺陷:BP算法收敛速度慢、易陷入局部极小值;确定神经网络结构费时费力,影响了模型的预测精度和适应性。本文提出了遗传禁忌混合算法,将该算法应用于训练神经网络,形成了遗传禁忌神经网络模型和改进的遗传禁忌神经网络模型。负荷预测试验表明,通过与标准的遗传算法和禁忌算法相比,遗传禁忌神经网络提高了预测精度,尤其是改进的遗传禁忌神经网络,能同时优化结构和权值,且在神经网络模型的适应性方面也有较大的提高。更多还原

【Abstract】 Load forecasting of electric power system has all-important effect on power system, such as operation, control and plan. The prediction accuracy has direct influence on economy benefits of the grid and power plants. Applying neural network to load forecasting of electric power system has become prevalent. The BP neural network is easy to stick in local optimization and has slow convergence speed. It takes much time and effort to determine the structure of neural network, which makes the neural network have the disadvantages of low prediction precision and worse applicability. This paper proposed the mixed Genetic Tabu algorithm(GATS). The Genetic Tabu neural network model and the improved genetic tabu neural network model were formed, which were trained by GATS. Applying them to the short-term load forecasting, the results were that the two models have higher prediction precision comparing with normal genetic algorithm and tabu search, especially, the improved genetic tabu neural network model obtained the perfect structure and the optimum weights at the same time, forever, it posessed better applicability of model.更多还原