摘要
火电单元机组是一种复杂的多变量对象,常规方法难以建立它的非线性数学模型。该文利用一种多输入多输出的连续小波神经网络对单元机组负荷数学模型建模问题进行了研究。网络隐层采用框架小波函数,输出层采用线性函数,采用BP算法对网络进行训练,并利用自适应的学习速率和动量参数加快网络训练的收敛速度。网络的训练结果和测试结果均表明,小波网络输出值与实际模型输出值之间的误差在允许范围内,小波神经网络可以较好地逼近单元机组负荷数学模型。
Thermal power unit is a complex object with multi-variables. It is difficult to build its nonlinear mathematic model in an usual way. This paper presents a simulation study on load modeling of a thermal power unit by a kind of multi-input-multi-output continual WNN model. The linear function and wavelet basis function satisfying the frame condition are employed as an activation function in output and hidden layer respectively, and BP arithmetic is used to train it, and self-adaptive learning rate and momentum coefficient are also used to accelerate the learning speed. The simulation results show that difference between the output value of WNN and the one of real model is in permissible range. WNN can approach the model of a thermal power unit very well.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2003年第10期220-224,共5页
Proceedings of the CSEE
