About

IoT and AI focused to photovoltaic park supervision and maintenance In the next decade a boom is expected in the domain of Renewables and in particular PV plants. In parallel many IoT based and AI, ICT technologies and solutions are made available. We have seen currently a gap and need for the focused use of these technologies to bring supervision, production prediction and preventive maintenance of these plants to a new level that will be required in the new conditions of the liberal Energy market. The new model of these markets with high percentage of RES will require a new set of production prediction technologies to assure grid stability and reliability, especially for the day-ahead operation. We are aiming to a PV plant portfolio adapted IoT solution allowing data collection at representative string level in order to allow both preventive maintenance incl. error detection and building of accurate physical particular PV plants models that in combination of microclimate predictions and ML methods will allow also optimal day-ahead predictions.

Solution & Product

The envisaged solution to be offered to clients for the management of their PV plant portfolio and the prediction of its day-ahead production for energy market use, is based on the integration of - a PV adapted IoT solution allowing data collection at string level of chosen strings (based on self created string selection methodology). This low cost monitoring will allow in combination with existing data from inverters and local weather stations to build an physical model of the plant adaptable to real conditions. The physical model can be used both for production predictions (especially in clear sky conditions) and for preventive maintenance of the parks. - ML techniques (ANNs, Fuzzy Neural models and LSTM) to correct the prediction errors expected by numerical weather predictions and the physical model based on statistical historical data based methods. This will allow accurate day-ahead predictions also when the physical model cannot usually provide it (i.e. cloudy days)

Innovation & Scalability

- Our Innovation relies on the tested smart use of sample string level data collection that will allow with low cost to build an accurate physical model of the PV plant portfolio of clients that in combination with available data from inverters, numerical weather predictions at local level, and adapted Machine learning methods like ANNs, Fuzzy-Neural Networks, Long-short term memory (LSTM) that will be used for the correction of usual errors of the physical models, will allow very accurate day-ahead predictions at competitive cost. the solution is scalable to portfolios In case the product will be ready to market on time (est. one year) it will come at the right time for the many PV plants that started to be constructed from 2019-2020 but will skyrocket from 2021 and onwards for at least a decade. So there will be a booming and prosperous client base for such products taking also into acocunt the newly introduced liberal energy market which demands for accurate day-ahead predictions. So the potential is huge but also much competitions is expected after the first few years. The challenge will be to be able to early come up with our complete solution as envisaged in order to rapidly earn a good market share in the Greek PV market. So the plan is to spent first year on research with a small engineering team (maybe 1 addition) to construct the solution without significant revenues, limiting sales during that period to monitoring solutions that can already be const

Traction & achievements

We have from previous endeavours in developing interactive travel guides for active holidays and GPS path tracking already build a reliable CI/CD software deployment pipeline. Also a dedicated data collection and presentation platform based on Node-RED has been successfully build that can collaborate with multiple Databases like Mongo and timeseries DBs like Quest. Last but not least we have designed and tested a method for selecting certain strings to be monitored that allow with low cost to have an accurate representation of the physical model of the PV plant continuously adapted to real conditions. This model will be used both in predictions and preventive maintenance in combination with envisaged early defect diagnosis techniques like regular collectionspatio-temporal information from thermal images minimizing also standard maintenance costs reducing also human interventions. the model is currently deployed and tested in one small 100KW park in Goumenissa Kilkis and will be scaled now to 500KW and beyond in a new park currently created in Larissa.

Team & Why Us

- Chatzimladis Christos, 54, has studied Physics and has a long carrier as ICT consultant for Training Organizations and developer of mainly Multimedia and learning software and solutions. He is the founder of the company. - Vaggelis Kesistornou, 53, an Electrical and Computer Engineer from Aristotle University of Thessaloniki is the brain for the company and main software architect and developer. - Georgios Belogiannis, cooperating electrical engineer, external contractor, offers his deep knowledge on photovoltaic systems and plants to construct the accurate physical model needed for monitoring and predictions.

Funding Needs

For the first year dedicated in research and development and early marketing campaigns when nearing product completion, the founding needs should be limited to about 60000 Euro. During the next 2 years more resources will be necessary to both continue product development and evolution and enter the market succesfully with adequate campaigns.