The relationship between weather variables and electricity demand to improve short-term load forecasting.

摘要

The power utility industry has become highly volatile with a deregulated market on the horizon and with enormous profit and loss swings in the energy trading market. Electricity, in particular, has become a commodity that is bought and sold at market prices, where load forecasting plays a crucial role in the composition of those prices. Public and private utilities must contend with the fact that a small error in an electric load forecast can create a large financial loss for the company. Hence, improving the accuracy of electricity load forecasts has become necessary for the long-term viability of all power utilities.; Weather has a significant impact on load demand and load forecasting. However, the weather-load relationship is unknown at the substation-level—mostly because substation-level load data have rarely been available to those outside the corporate infrastructure. Equally as important, most utilities have made inconsistent and antiquated use of weather data.; This study used electric load data from four substations in Oklahoma and concurrent weather observations from co-located Oklahoma Mesonet sites to: (1) determine the interrelationships between weather variables and electric load demand; (2) determine the impact of weather on the consumption of electricity by different customer classes (e.g., residential, commercial, industrial); (3) establish thresholds of temperature associated with changes in the patterns of the use of electricity; and (4) produce load model simulations to quantify the improvements in the accuracy of a load forecast. This study also links a much improved, high-resolution numerical weather prediction model to a neural network load model to quantify the economic value of improved accuracy in load forecasts. In the end, this dissertation determined that a comprehensive understanding of the relationship between weather variables and electricity demand will improve the accuracy of load forecasting. The results of this study can save a small utility in excess of {dollar}0.5 million annually. If the results are applied to the larger power companies around the United States, a decrease in operating costs could exceed millions of dollars.