Neural Network-Based Approaches to Nonlinear Short Term LoadForcasting for Electric Utilities.

电力负荷需求以一种高度复杂的方式受到几个变量的影响，例如一天中的时间、一周中的哪一天，以及温度、湿度和风速等天气变量。本研究拟探讨三种不同的神经网络方法在短期负荷预测问题中的应用。第一种方法是使用反向传播的前馈神经网络。第二种方法将尝试使用使用线性阈值单元的神经网络架构进行预测。第三种方法将使用神经网络来模拟非线性系统和天气动力学，利用确定性混沌理论的思想。这些方法将在一个合作的实用程序提供的实际负载和天气数据上进行测试。

Electric load demand is influenced in a highly complex manner by several variables such as time of day, day of the week, and weather variables such as temperature, humidity and wind speed. The research proposed is to investigate the application of three different methods for applying neural network methodology to the problem of short term load forecasting. The first approach will be a feedforward neural network using backpropagation. A second approach will attempt to forecast using a neural network architecture using linear threshold units. The third approach will use a neural network to model nonlinear system and weather dynamics using ideas drawn from deterministic chaos theory. These approaches will be tested on actual load and weather data provided by a cooperating utility.