16:20 | 17:00
Keywords defining the session:
- RE management
- Machine learning
Takeaway points of the session:
- New data technologies provide an opportunity for business in renewable energies sector
- Opera Digital Platform: towards to the best performances & profit of PV technology
We are developing a business to provide services regarding with the O&M management of PV installation; such as monitoring, performance check, predictive maintenance, asset management, etc.
The Opera Digital Platform, in combination with an specific monitoring hardware, both entirely developed by us, allow us to do this work with avery low investment cost and manpower requirements, thanks to use of open sources and free licenses. In addition, this platform offer a very simple process to manage the data in thru real time, it is easily scalable and also easily customizable. So, we can substantially reduce de O&M cost and upscale the performance of the system, growing thus the profit of the installation.
In the first stage we are looking for early-adopters and stakeholders to put the Opera platform at trade level.
With almost 5 GW of PV generator installed up to now in Spain, O&M_PV represents a market over 60 M€, only in our country.
Opera architecture: the platform have been implemented on a distributed big data platform based on a lambda architecture. The main advantages of this architecture are the scalability of the system in terms of storage and processing power and the ability to process data in both streaming and batch data flows.
The streams of data from the PV systems and external sources are sent to a publish/subscribe streams storage system. From there, the data are processed through the real time layer of the platform by a stream processing framework in order to implement all the calculations for the performance indicators. The stream of enriched data are then indexed in a search engine to allow real time monitoring and control of the system. A visualization plugin contain the monitoring dashboards where the indicators and performance charts are published, also allowing data discovery for end users through detailed queries.
A distributed file system optimized for storage of large amounts of data is the core component of the batch layer of the platform. The stream is sinked into this file system to persist the data and store historical information of the systems. This historical data are the input to train predictive analysis models with the objective of identifying and anticipating performance issues in the PV systems. The models runs with the current data in scheduled batches (frequency to be determined, e.g. weekly or twice a day) to generate the necessary alerts or reports for proactive O&M, and the results are also indexed in the search engine for visualization of the metrics in the management dashboards.
Besides, in order to be able to collect real time data of the PV system performance by the platform, we have developed a new data collection mechanism to get the real time data of the system’s operating variables, by means of new sensor that allow the use IoT (Internet of Thing) connection concept.
Opera data science: The monitorization of the performances of the PV generator is made following the European Standard IEC 61724.
On the other hand, in order to be able the implementation of an effective diagnostic method and an early damage and fault detection, and a predictive maintenance, it is necessary to compare the monitoring variables and ratios with the theoretical ones. To do it, the platform implements a behavioural model for PV systems.
Furthermore, the platform includes heuristic algorithms and machine learning to improve the forecast of the system behaviour along time.