神经网络法在以网损为目标的配电网重构中的应用

Application of the Neural Network Method in the Reconfiguration of Power Distribution Networks Aiming at Transmission Losses

配电网重构是降低电网线损、提高经济性的重要手段。采用BP神经网络法进行配电网的重构,网络输入为电网负荷,输出为实现电网最小线损的开关开合状态,通过样本训练来实现对两者非线性关系的模拟。首先将电网负荷按负荷曲线特征及负荷水平分为三种类型和五种负荷水平;然后对不同负荷类型和水平下的最小线损采用最优化方法计算,作为网络训练的样本;以一个16节点系统为算例,验证了BP神经网络法在配电网的重构中的应用价值。为弥补以往研究的不足,探讨了神经网络结构对电网重构的影响,发现通过选择适当的输出神经元数目可以在不增加太多网络训练时间负担的前提下,提高神经网络在实际配电网中应用时的效率。

Reconfiguration of the power distribution network is an important means for the reduction of power network line loss and improvement of economical efficiency. In this paper, the BP neural network （BPN） approach is applied for the reconfiguration of the power distribution network. The network input is the load of the power grid, while the output is the open-close state of the switch for minimal line loss of the power network. The nonlinear relationship between the input and output values is simulated through sample training. First, network loads are categorized into three types and five load levels according to load curve characteristics and load level. The minimal line loss, calculated in the optimization method for different load types and load levels, is used as the sample for network training. Then, in a computational example of a 16-node system, We verify the application value Of the BP neural network method in distribution network reconfiguration. Finally, to make up for the deficiency in past research, this paper discusses the influence of neural network structure upon power network reconfiguration. It is found that selection of a proper number of output neurons can improve the application efficiency of BPN in distribution network reconfiguration, without increasing too much network training time.