Forecasting electricity is an important process for utilities and system operators and has been so since the beginning of the 20th century. This is because electricity, unlike many other commodities like gas or fuel for heating, cannot be cost effectively stored at a bulk level. On top of this, the transmission of electricity requires that the infrastructure in place operates within its prescribed limits, otherwise there is an elevated potential for faults and outages. Therefore, companies which own, operate or regulate infrastructure need accurate forecasts of future electric loads which can be used to inform their operations, ensuring reliability and cost effective operation. Nowadays, with the proliferation of smart grids and intermittent renewable energy resources for supply, load forecasting is of even greater importance due to its applications in planning of demand side management and integration of distributed energy resources like PV panels. This looks set to increase in future as other electric loads appear on the grid, like those from EVs or low-carbon heat systems.
There are very many factors which influence electricity loads, including meteorological variables, such as temperature, wind-speed, humidity, cloud, precipitation, etc, calendar variables, such as time of day, day of week, holidays, and economic information, such as land use or building type. These variables are all considered when building a forecast method. Electricity load forecasting is also classified according to the time horizon, since forecasts for different future time windows use different methods and the relative importance of the input variables is likely to be different. Typically, electric load forecasts can be classified into the following categories:
- Very Short Term Load Forecasting (VSTLF) - from second to hours ahead of real-time.
- Short Term Load Forecasting (STLF) - a day to a couple of weeks ahead of real-time.
- Medium Term Load Forecasting (MTLF) - weeks to months ahead of real-time.
- Long Term Load Forecasting (LTLF) - months to years ahead of real-time.
STLF for distributed energy systems and the effects of aggregation level
In the ESES lab, our interest with forecasting is predominantly in the STLF domain due to its link with system operation and renewable energy technologies such as wind and PV. Moreover, the increase in popularity of Distributed Energy Resources (DERs) is resulting in dramatic changes in the demands at the ends of the distribution network. Consumers are increasingly supplying power into the network and modifying their behaviour according to price signals in the electricity market. These changes offer a new set of economic opportunities which can be exploited if accurate STLFs at a highly disaggreagted level are available. From a utility perspective, a failure to adapt to these changes also has the potential to cause both significant operational and economic disruption. Therefore, accurate STLF forecasts on a distributed scale are crucial for ensuring reliable system and market operation, as well as for longer term planning activities.
We are testing a number of methods of improving disaggregated STLFs and studying the effects of different aggregation levels of forecasts.