应用神经网络决定多机系统快速汽门控制规律

APPLICATION OF NEURAL METWORK TO TURBINE FAST VALVING IN NMULTI-MACHINE POWER SYSTEM

提出了一种应用单隐层前向神经网络决定多机电力系统暂态稳定快速汽门控制规律的方法。神经网络采用故障前各发电厂单台机组的有功功率和开机台数等稳态量以及故障位置和类型等当地暂态信息作为输入，其输出为不控情况下故障后某一适当时段内发电机转子间的最大相对摇摆角以及汽门关闭持续时间的上、下限，前者用来与某一给定临界值进行比较以决定是否需要控制，后者则用于确定合适的关闭持续时间。为了说明这种神经网络结构的合理性，文中定性地证明了其输出与输入之间近似呈连续映射关系，并给出了一个６机２２节点试例系统的结果。

The application of single hidden layer feedforward neural network to turbine fast valving for im-proving multi-machine power system transient stability is proposed in this paper.The inputs of the neuralnetwork are the number of operating units and the active power of individual generator at each power plantunder prefault operating condition as well as the local information about fault type and location,while theoutputs are the maximum relative swing angle between generator rotors in an appropriate postfault periodwithout fast valving, the minimum dead time for preventing firot swing stability and the maximum deadtime determined by second swing stability. Fast valving requirement is decided by comparring the first out-put with the given critical value. The dead tune is determined by average of the second and the third out-put. The mappings of the inputs to outputs are analysed to illustrate the rationality of the proposed neuralnetwork structure. The results of a 6-machine test system are given.