基于时序数据-BP神经网络的电力能耗分析研究

Power Consumption Analysis Based on Time Series Data-BP Neural Network

为了更好地对电厂中机组的能耗问题进行分析研究,根据电厂不同工况的参数指标,建立电力企业能耗仿真BP神经网络模型。归一化处理电厂运行过程中各类传感器采集的数据,采用负荷、环境温度、排烟温度、背压、含氧量等指标,并加入时序历史能耗作为输入参数;利用电力企业短时能耗作为输出参数;通过采用不同时间窗口的连续时序能耗参数指标和热力学相关参数作为输入,在神经网络中不同中间层的隐层节点数下进行仿真实验。结果表明,基于时序历史能耗数据的电厂指标参数在包含21个隐层节点数的BP神经网络模型上能够在线仿真出精度较高的短时供电能耗数据。所建的电力能耗预测模型将为后续电厂的节能减排、负荷优化,提供理论参数支撑。

In order to further study and analyze energy consumption of units in power plant,a BP neural network model for energy consumption simulation of power enterprises is established according to the parameter indexes of different working conditions of power plant.Data collected by various sensors was normalized during the running of the power plant,indicators such as load,ambient temperature,exhaust gas temperature,back pressure and oxygen content was adopted.With time series historical energy consumption being input parameters,the short-time energy consumption of electric power enterprises was taken as the output parameter.By using the continuous time series energy consumption parameters and thermodynamic parameters of different time windows as input,the simulation experiments were carried out in the hidden layer nodes of different intermediate layers in the neural network.The results show that the power plant index parameters based on time series historical energy consumption data can simulate online the short-term energy consumption data with high accuracy on the BP neural network model with 21 hidden layer nodes.The power energy consumption prediction model will provide useful theoretical parameters for energy conservation,emission reduction and load optimization of power plants.