An Interview with Giuseppe Pirrelli
November 22, 2022
Giuseppe Pirrelli is a Data Scientist at MiPU, a leading software development company whose mission is to bring prediction to factories and cities to reduce waste, costs and difficulties and to enhance sustainability and social inclusion.
Can you give us an overview of the SOLar forecaST In dynamiC Environments solution that your company is testing with PLATOON?
SOLSTICE, the acronym for SOLar forecaST In dynamiC Environments is a data analytics application whose purpose is quite simple: its goal is to provide greater visibility into future power production from photovoltaic modules. In particular, solar power is a non-programmable source, like other renewables, meaning that we cannot turn on and off the production at will, but we need favourable weather conditions (the absence of clouds, solar radiation, etc.). As a result, power provision from renewables entails significant challenges in terms of planning and organisation of operations. In order to overcome such limitations, we employ advanced analytics and artificial intelligence methodologies to forecast future power production and allow better operational planning. The algorithm takes weather forecasts as input and estimates future power production from photovoltaic arrays, with hourly granularity, and up to seven days ahead. The training phase entailed a comparison of past weather forecasts and observed power production, so that the algorithm learns to approximate the relationship between the two. The model is a physical-hybrid machine learning model, meaning that it takes into account both statistical relations between the variables, as well as features coming from the physical laws governing the process.
The training has been challenging for the following reason: the PV modules are installed at MultiGood MicroGrid LAB (MG2LAB), Politecnico di Milano, which is an experimental facility to test software and analytics tools for automatic smart grid management. As a consequence, different experiments are carried out at MG2Lab, many of which foresee artificial limitations to the power produced. We had to recognise and filter out power production observations corresponding to periods when experiments were conducted.
After training and deployment of the algorithm, its output enters in a flow where it becomes an input for the software taking care of the automatic management of the smart grid. The forecasting performance is in line with the academic literature, still there is room for improvement. In particular, it is a "stacked forecasting" problem, meaning that our predictions are produced on the basis of other predictions, namely, weather forecasts. Accordingly, if weather forecasts show high error margins, it is very difficult to develop a forecasting algorithm whose error margin is lower on average.
Further developments may entail the use of fisheye cameras pointed at the sky to estimate clouds’ movement. This would provide better predictions on future weather conditions, but at the same time, the algorithm scalability would be reduced, as not many users may have the camera needed to collect cloud images.
What impact do you think PLATOON will have on the experience of energy customers—the end users?
I see the highest potential of PLATOON platform in a business-to-business context, but of course, this does not mean end users will have no benefit. At the macro level, the long-term goal of innovation and digitisation initiatives such as PLATOON is to reach an overall increase in efficiency in power production and distribution. This entails greater production, less waste, better programmability, demand-response, and distribution efficiency, among others. Energy providers may produce more with the same cost of production as before, or experience lower costs.
Distribution operators will have the possibility of leveraging advanced analytics and AI algorithm to better plan and automatise the power distribution. As a result, end-users will likely experience better access to energy, increased availability with less service disruptions, and maybe lower costs. In my opinion, the full potential of energy data spaces and PLATOON will be reached with a greater diffusion of energy communities and the "prosumer" model, where households and neighbourhoods are energy consumers and producers at the same time, injecting into the grid their excess power. When data integration and communication into the platform will reach this level of granularity, and prosumers and energy communities will exchange information and use data applications on data spaces, the positive network effect will be incredible. The room for optimisation is huge, and it is proportional to the amount of information we are providing to the PLATOON platform. In addition to that, privacy and trustworthiness of data exchange on PLATOON will ensure ownership and confidentiality of citizens’ information, which today is a significant hurdle for the development of data economy and data-intensive applications.
What do you see as the particular strengths of PLATOON?
PLATOON aims at the development of a software platform providing a trusted, secure data and applications exchange environment, benefitting the whole energy production and distribution value chain. It is something beyond state-of-the-art on so many levels: technical development of the software platform, type of data-intensive applications available on it, human and machine-readable contracts for service provision, semantic data models and standardisation of the information flow, and so on. The consortium is a perfect mix between big players of the energy industry, academia, research centres and institutions, having the experience and know-how to really make a step forward in the direction of interoperability and digitisation of the energy industry.
PLATOON allows data providers to exchange potentially sensitive information in a secure and trusted way, providing authentication, data brokerage and clearing house microservices; at the same time analytics providers, who are data consumers, can rely on a software architecture where the deployment of AI models is easier and way more efficient thanks to the implementation of standardised semantic data models. While not yet fully exploited, PLATOON provides the technical means and the business and contractual frameworks needed to leverage data as a production input, just like physical goods, capital or labour, elements we can confidently assign an economic value to.
The platform can seamlessly connect with third-party systems and any data sources, making it easy to build an ecosystem where a number of actors leverage to different degrees PLATOON’s applications and data, eventually integrating the information flow with their own IT architecture.
It is not likely that the digitisation and a paradigm shift in the energy industry information flow management will be carried out in a completely centralised way. Instead, it seems more reasonable that multiple elements will have to be put into communication and effectively coordinated. In my opinion, PLATOON has perfectly identified such an innovation path, and aims to be one of the first building blocks of the new data economy in the energy industry.
Part of your award from PLATOON was an invitation to join the Technology Transfer Programme. How do you feel this has benefitted your company?
The opportunity to participate in PLATOON’s Technology Transfer Programme (TTP) has been invaluable for Mipu Energy Data, for a number of reasons. The Programme has been designed to develop data-intensive solutions, or software building blocks of the platform, starting from a low technology readiness level (TRL). This means that beneficiaries have been able to carry out research and development operations funded by the European Commission, increasing the TRL of strategic applications that will feature in the companies offering portfolio and, at the same time, are in line with European Union strategic industrial policies. In addition to that, the R&D phase has been supported by the Mentors, giving an important contribution on the basis of their experience and know-how. In our case, Politecnico di Milano provided academic knowledge in the development of AI-based PV power production forecasting models.
Apart from technical innovations, participation in PLATOON TTP allowed us to gain strategic insights into the future of the energy industry, and how different actors are cooperating to envision the next generation of technologies and business models for energy provision, distribution and complementary services all along the value chain. For example, we participated at the "Conference on data sharing and governance for energy applications". Held in Bilbao in September 2021, and organised by the PLATOON consortium and Basque Energy Cluster, the conference allowed universities, start-ups, research centres, SMEs and big players to share views and illustrate current and forthcoming initiatives for digitisation and interoperability in the energy sector. Mipu Energy Data is willing and determined to keep the pace of such innovation, and these kinds of initiatives are of paramount importance to consolidate our position as a trailblazer in applying advanced analytics in industrial settings.
One of your company’s stated purposes is: "We renew industry and cities with respect for sustainability and social inclusion." Can you talk a little about how you think Big Data promotes social inclusion, and its role in meeting the SDGs?
If I think about inequalities, insufficient or absent education, poverty, and the other challenges tackled by the Sustainable Development Goals, I see many problems that are generated by a lack of access to information, or at least situations where an easier and better access to information would reasonably bring significant improvements.
Big Data and analytical methodologies such as AI simply increase the amount and the quality of information at our disposal. The potential is endless. Forecasting energy production from non-programmable sources such as renewables would increase their use, reducing carbon footprint; AI-based optimisation of irrigation and use of pesticides would increase the agricultural production per square meter; automatic analysis of satellite imagery would help in the prevention of natural disasters, just to name a few SDGs-related Big Data applications that come to mind. Consider the social impact of those applications: the link between renewables and energy communities, where access to energy is extended and more autonomous; the effect of increased agricultural production on malnutrition and hunger; earth observation for reduction of geohazards, and so on.
All of this would have been sci-fi just 20 years ago… and not because we did not have the mathematical tools: AI algorithms have been around since the ’60s, but we did not have the infrastructure to collect sufficient data volumes nor the computational power to analyse them.
On behalf of Mipu, we are more than proud that Sustainable Development Goals are a point of reference and guiding principles for the way we do business. For example, Mipu has a great gender balance in technical and management roles, a lot of our projects utilise energy production from renewables, and last but not least, we are working more and more with water management companies, especially in northern Italy, developing specific know-how on how to apply AI to reduce water waste.
What challenges could lie ahead for optimising energy networks with smart technologies and Big Data?
I believe the energy industry is the perfect field for an extensive application of data-intensive technologies powered by artificial intelligence and machine learning. My optimism comes from two elements: first, the industry collects lots of data (data acquisition systems are already in place in the large majority of cases), and then, energy production and distribution are processes abiding by underlying physical functional relations, which AI algorithms are very good at approximating. On the other hand, energy production and distribution is a complex value chain, where a large number of actors play their roles, and have different goals, needs, technical capabilities and methodologies… even more so with the spread of energy communities and the "prosumer" model.
In order to best leverage the power of artificial intelligence, it is pivotal to take into account the relationships between actors of the system, entailing the integration of multiple data sources, both open and proprietary. This process requires a good degree of standardisation of information, data formats, trustworthy security and privacy measures to ensure secure data exchange, interoperability between different software systems and IT architectures, let alone regulatory and contractual frameworks for services provision, etc.
In addition to that, we do not have to underestimate the complexity arising from the geographic distribution and interconnectedness of energy networks. They are a strategic asset from the political and social perspectives, as we have been witnessing over the first months of 2022 in Europe, unfortunately. As a result, technical hurdles are intertwined with economic, political and social challenges.
Platoon and the International Data Spaces initiatives are a first step in the direction of tackling technical complexity, and a very important one. Moreover, the European nature of the projects requires concerted actions from national and supranational institutions, paving the way for a common strategy for the energy provision value chain that takes into account its political and economic centrality.
Back to H2020 PLATOON main page
© Sustain Europe 2022. All rights reserved. Powered by 100% Green Energy. Our pledge to the Environment.
Follow Sustain Europe