While I am usually performing research in Software Engineering and HCI, I am happy to present a first milestone that we’ve reached in our Smart Energy project at MIT Coaching. In our Lab32, we are building photovoltaic (PV) power output predictions in the pre-alpine terrain of Switzerland.
To that purpose, we used actual PV power output data as ground truth, and use meteorological forecast data (such as temperature, global irradiance, clouds) as well as PV system data (such as location, tilt, type of modules and inverters) to build predictive models based on artificial neural networks (ANN), that are able to forecast the future PV power output on a intra-day horizon of 12-24 hours.
This already works fairly well, compared to related work, given the challenging mountainous terrain we are working in. The following screenshot shows a selected (good) prediction (and reality) of three locations in Fall 2020:
In a recently published executive summary, we describe our goals, approach, including detailed data cleaning and feature engineering descriptions, we present our results, and compare them to related work.
As next steps, we formulate the following goals:
- Improvement of our intra-day models (12-24h)
- Building new intra-hour models (15-60min) for utility companies and power grid operators
- Develop applications for consuming our intra-day and intra-hour PV power data prediction models
To that purpose we are looking to establish industry and research partnerships in Switzerland, Germany or Austria:
- Energy management or monitoring platform providers, power grid operators and energy provider companies; to gain access to additional measured PV power output data, integrate power output predictions into power/grid management platforms, and co-fund projects
- Extending our existing research partnerships for knowledge and expertise exchange
- API access to consumers’ power usage to develop models for smart grid and load power management
Please contact us in case you have any questions, inputs, and suggestions to our analysis, or want to discuss potential collaborations. Thank you!
Download the Executive Summary here.
Comments are closed.