Electricity Load Prediction
LoadFor is a software solution for forecasting of electricity load (demand). The solution is a self-learning and self-calibrating system. It is based on machine learning and uses weather forecasts and historical demand data to automatically produce very accurate electricity load forecasts.
How does LoadFor work?
LoadFor is a self-learning system based on machine learning. On the basis of historical electricity load, historical meteorological data and meteorological forecasts, the system is not only able to predict the electricity load, but can also automatically and continuously calibrate and improve its predictions as it is fed with more data. In addition, online power measurements can be used as input (if available) to increase forecast accuracy.
Based on input data, LoadFor automatically identifies and takes the systematic behavior of electricity consumers into account. This means that LoadFor continuously adapts to the actual situation by continuously monitoring the consumption and adapts to changes, such as:
- Changes in consumer behavior
- Changes in the number of consumers
- Changes in the meteorological models
- Changes in the physical characteristics of the power grid
The self-learning mechanism has the benefit, that LoadFor will identify the impact of any changes by itself and quickly adapt.
Electricity load forecasting can be complicated by the fact that the dynamics of buildings in some geographic regions affect the cooling or heating demand on an hourly basis. LoadFor system automatically applies an optimal smoothing effect which solves this issue, such that the physical properties of the underlying energy system are modeled correctly and the forecast shows the appropriate response to changes in temperature or sun irradiation.
Optionally, LoadFor can be deployed in combination with MetFor (ENFOR service for locally optimized weather forecast) to obtain a more accurate local weather forecast, which will result in superior electricity load forecast accuracy.
LoadFor is provided as an integrated service from the ENFOR platform which contains a data collection and validation module. The data collection and validation module collects the necessary data, ensures that the necessary data is available and contains a toolbox for automatic detection and correction of missing and/or erroneous measurements. The module then feeds the validated data into the core LoadFor™ module which then provides the forecasts.
The ENFOR platform also provides LoadFor with data integration modules through either FTP, SFTP or web-services such that LoadFor can be seamlessly integrated with various data sources.
LoadFor can be installed locally on customer systems or hosted by ENFOR as a service. It is also possible to get a customized support and maintenance agreement from ENFOR or one of the partner companies.
The following key features are provided by LoadFor:
- Self-learning and self-calibrating algorithms for accurate forecasting of electricity load
- Scenario generation
- Uncertainty bands based on quantile forecasts
- Integrates seamlessly with locally optimized weather forecasts (ideally provided by MetFor™)
- Web-interface available for configuration and monitoring of system
- Data integration interfaces based on FTP, SFTP or web-services supporting numerous formats and file types (CSV, XML, SOAP, JSON etc.)
- Runs on common server platforms (Windows, Linux)
Wind power prediction for wind farms
Solar power prediction for PV plants
Load forecasting, temperature control, and optimization of district heating networks