Periodic Reporting for period 1 - AI-METHOD (INNOVATION ASSOCIATE FOR RESEARCH, DEVELOPMENT, AND APPLICATION OF INTELLIGENT MODELLING FROM EARTH OBSERVATION DATA)
Período documentado: 2019-10-15 hasta 2020-10-14
The main objective of the project was to significantly strengthen geopredict’s resources, knowledge and skills in order to amplify the competitive advantage of our geo-forecasting solutions based on Intelligent Inductive Self-Organizing Modeling and Forecasting technologies. The project played a strong role in the long-term business strategy of geopredict towards a major European player in R&D driven forecasting and self-learning predictive modeling technologies, leveraging the power of earth observation data, for highly-relevant industrial and societal problems.
The key objectives of the project have been achieved successfully. Some results have been implemented, already, and they have shown a positive impact on our CLIMFOR solution applied to one of our pilot users in the energy sector.
Main R&D results from this project are: (i) concept development of a storage and query framework for satellite earth observation data in a geospatial big data platform, (ii) design, development and implementation of a sophisticated algorithm for probabilistic forecasting, (iii) design, development and implementation of an algorithm for adaptive forecasting of time processes based on probabilistic forecasts, and (iv) design and development of an algorithm for structural and parametric identification of poly-harmonic functions of optimal complexity.
The obtained results were published in two papers and presented and discussed at two conferences. An extended version of one paper was also accepted for publication in the book series "Advances in Intelligent Systems and Computing" by Springer to be published in 2021.
We succeeded with our ESA AI-Kickstart project SEED (https://business.esa.int/projects/seed) on technical and economic feasibility of our system for daily renewables energy production forecasting for the Indian energy market. A development version of our system has been applied for a solar farm company in India. It has been shown that the CLIMFOR system, also based on the work of the Innovation Associate on probabilistic forecasting within this AI-METHOD project, is able to reduce penalty payments of renewable energy producers according to India’s Deviation Settlement Management regulatory framework up to 50% in comparison to existing solutions.