Leveraging data-intensive analytics for optimal integration of renewable Distributed Generation (DG) and storage in distribution networks for optimized Energy Community (EC) planning, sizing and operation.
Technology
Machine Learning
Region
EMEA
Team lead
Christian Melchiorre
Team members
algoWatt SpA: Stefano Bianchi Carlo Alberto Bassi Christian Melchiorre Diego Piserà Lorenzo Bianco
University of Genoa: Prof. Stefano Massucco Prof. Federico Silvestro Gabriele Mosaico Matteo Saviozzi
Nice work, algoWatt! This is a very impressive team, working on an interesting topic. One outstanding question I have for the group is: how I should think about your forecasting and optimization findings, given that the data is mostly from 2019? As we all know, COVID drastically changed how we work and live. How much do you think your results/recommendations would change if you used data from, say, 2022?
Thank you for your interest in our Hackathon submission and we greatly appreciate your insightful comment @LGroves, I am writing back on behalf of the Team that prepared this answer.
Your observation is quite correct that the pandemic has changed people’s behavior and habits. During and post pandemic, the way we live and work has changed considerably. So, it is conceivable that the energy usage patterns may have changed during this period and consequently will have an impact on the recommendations that we arrived at using older data.
We were working with a couple of constraints: a) getting access to recent data – the energy distribution company that we were working with had some difficulty providing access to current data for a variety of reasons including the proprietary nature of the data, b) limited time available for the hackathon.
Consequently, we had a very limited dataset to work with and we did the best we could with the given data. However, our main focus was to come up with an overall approach for addressing the energy community issue and developing a generic process that could be applied to different context. The approach basically boils down to gaining a good understanding of the energy profiles of the substations and the PV plants through forecasting and use this as the input for the optimization part.
This framework could be applied to different contexts/datasets representing different periods of time, different regions, etc. So, the major value add from the hackathon is coming up with an overall process for figuring out how best to form energy communities and demonstrate it through some sample data. The work done during the hackathon provides a good starting point for us and we will continue to work with algoWatt in refining the model and applying it to larger datasets.
Due to the difficulty in getting the current data, a call was made that we will work with the pre-pandemic data and build our models and come up with some preliminary findings. Then, we would use the same models with the post-pandemic data and see if we observe any considerable shifts in energy usage patterns and how that would affect the energy community formation. In a sense, this would help us understand the pre and post-pandemic scenarios, which would help us make better long term decisions.
To add to a further insight to your comments, we would also expect more changes influenced by the recent inflation/expectation of gas shortages due to the geopolitical facts we all know. These facts have caused energy prices to skyrocket for a few months, with effects on consumption due to both the increase in electricity prices themselves, but which are probably also affecting electricity consumption patterns as a side effect.
The fit of the general model to different conditions (electricity market and its externalities) will come from the data itself and the optimization model, which is "data-driven" through the forecasting model.
In conclusion, the point was precisely to develop a versatile model, rather than to focus on the actual results that are specific to the data set and thus to the specific market conditions (which include "externalities") that, in fact, affect the electric consumption model and, consequently, the consumption data collected on the field.
Thanks again for your interest and feedback on our project.
Thank you for your interest in our Hackathon submission and we greatly appreciate your insightful comment @LGroves<>, I am writing back on behalf of the Team that prepared this answer.
Your observation is quite correct that the pandemic has changed people's behavior and habits. During and post pandemic, the way we live and work has changed considerably. So, it is conceivable that the energy usage patterns may have changed during this period and consequently will have an impact on the recommendations that we arrived at using older data.
We were working with a couple of constraints: a) getting access to recent data - the energy distribution company that we were working with had some difficulty providing access to current data for a variety of reasons including the proprietary nature of the data, b) limited time available for the hackathon.
Consequently, we had a very limited dataset to work with and we did the best we could with the given data. However, our main focus was to come up with an overall approach for addressing the energy community issue and developing a generic process that could be applied to different context. The approach basically boils down to gaining a good understanding of the energy profiles of the substations and the PV plants through forecasting and use this as the input for the optimization part.
This framework could be applied to different contexts/datasets representing different periods of time, different regions, etc. So, the major value add from the hackathon is coming up with an overall process for figuring out how best to form energy communities and demonstrate it through some sample data. The work done during the hackathon provides a good starting point for us and we will continue to work with algoWatt in refining the model and applying it to larger datasets.
Due to the difficulty in getting the current data, a call was made that we will work with the pre-pandemic data and build our models and come up with some preliminary findings. Then, we would use the same models with the post-pandemic data and see if we observe any considerable shifts in energy usage patterns and how that would affect the energy community formation. In a sense, this would help us understand the pre and post-pandemic scenarios, which would help us make better long term decisions.
To add to a further insight to your comments, we would also expect more changes influenced by the recent inflation/expectation of gas shortages due to the geopolitical facts we all know. These facts have caused energy prices to skyrocket for a few months, with effects on consumption due to both the increase in electricity prices themselves, but which are probably also affecting electricity consumption patterns as a side effect.
The fit of the general model to different conditions (electricity market and its externalities) will come from the data itself and the optimization model, which is "data-driven" through the forecasting model.
In conclusion, the point was precisely to develop a versatile model, rather than to focus on the actual results that are specific to the data set and thus to the specific market conditions (which include "externalities") that, in fact, affect the electric consumption model and, consequently, the consumption data collected on the field.
Thanks again for your interest and feedback on our project.
Well said, @marsac! I appreciate the added context, particularly the challenges associated with optimization during a global pandemic and wartime supply-chain disruptions. So, thank you for your thoughtful response. I also agree that a Hackathon should focus on the models, methodology, and data that you have today, with the goal of creating a versatile model that can be rolled out as more recent data becomes available. And I think that's a nice reminder to all of us: we have the ideal... and we have what is. Modelers focus on the latter 🙂
So, thanks for that input - and, again, great job in this project, @algowatt!
