Informetis
Contact: yuichi.abe@informetis.com
2021 Discussion
José Alcalá, Head of Data Science Team,
Boruo Xu (Bono), Head of Data Solutions Team,
A keystone in the transformation to Smart Grids is the synergy between solar panels and batteries. This is a complex task that heavily depends on the weather, energy tariff and human stochastic behaviour. Informetis is an AI energy start up company that aims to empower the user in the Digital Transformation of electricity data. We propose the implementation of an Artificial Intelligent Agent capable of deploying the best strategy for charging and discharging of domestic batteries. This will aid the user to select the correct battery size to maximise household energy bill savings. Thus a communication channel, such as a native or web based app, needs to be opened with the user. The project is suitable for a team of 6 people and it might be divided into three modules: Machine Learning strategy for battery control, 2 people; backend to support communication with the cloud, 2 people; and front-end interface to show results to the users and, potentially, receive feedback, 2 people. Informetis will provide historical household power consumption and historical household solar generation for a set of houses. We are hoping to use the tool that you develop to showcase to stakeholders as a demonstration of smart energy technology.
Feedback:
I can see that implementing an algorithm to optimise battery selection would be an important problem, but unfortunately our students are not taught any topics relevant to battery technology, so would have to treat the battery as a black box. My own estimate is that, if you had a suitable training dataset, then implementing a machine learning algorithm to solve this task on a black box basis might only take an hour or two using Keras and TensorFlow (perhaps followed by a few more hours playing with hyperparameters).
A p2p energy trading system sounds more appropriate. Here are a few examples of previous projects that involved the implementation of market or trading systems.
Revised suggestion:
Home batteries are rapidly evolving and disrupting the market along with PV panels. However, do you actually know whether you can benefit from it? How much energy can you save? When do you reach the ROI? Finally, if the number works for you, which one to buy? Your task is to create an application that helps consumers to answer these questions. The user will input through the interface different parameters for batteries such as capacity, price, cycles, voltage, and so on. Your application will use these user inputs and the historical data (i.e. energy consumption and solar panel generation) to run a simulation of the best battery control strategy and will yield how much energy will be saved. The application may also recommend changes to user habits to improve efficiency. This task entails very challenging Machine Learning problems such as pattern usage modelling and forecasting, reinforcement learning (for battery control policies) and optimization. Regarding the interface, it has to be friendly and intuitive so the user can interact often and give continual input to the system. In the future, p2p energy trading can be also implemented.
Feedback:
- You say “in the future p2p energy trading can be also implemented”. Does this mean that the team should implement an API for future integration with another system after the project ends, or is energy trading part of the system requirement? (If it is part of the system requirement, then how should the team interpret “in the future”?)
- If the main purpose of the application is battery selection, why would users be required to give "continual input"? Would users need to replace new batteries very often, or does the continual input relate to some other area of functionality?
- If reinforcement learning is to be used, what system would be used to test the control strategies? Would this be a battery simulation, or actual batteries?
- Where machine learning is to be applied for pattern usage modelling and forecasting, are there existing data sets for training and test, or would the students need to construct or acquire these?
- If I understand correctly, the main output of the system is "which one to buy”. Presumably this is dependent on retail price, availability, and installer / supplier bundling. Are there existing suppliers whose sites could be scraped for this data, or would the purchase decision be based purely on technical performance and OEM pricing?
2020 Project
Client: José Alcalá <jose.alcala@informetis.com>
Proposal awaiting finalisation:
Activity Analysis based on Smart Meter data
A rapidly aging population means assisted living is fast becoming a major societal challenge. To support “Carers” assisting “Carees”, local technology company Informetis provides smart sensors installed in the fuse-box that determine if household appliances are ON/OFF. As routines are typically closely related to appliance use, this proxies for inhabitants' wellness. Your task is to create an app for carers to support householders when they struggle to perform daily routines. You may choose focus on detection of activities from the raw data, or the interaction design challenges of choosing what should be communicated to the carer and when.
Past Years
2020 project: Activity Analysis based on Smart Meter data
2019 suggestion: Activity Recognition and Analysis based on Smart Meter Data
2018 project: Energy Budget
2017 offer: Energy with Social Conscience
2017 suggestions:
Smart Homes and IoT
The Internet of Things (IoT) is now almost mainstream. There are a lot of internet-enable devices such as smart sensors, connected appliances, connected medical appliances and of course smart phones. These devices are all constantly generating big data dependent on individual lifestyles. In particular, smart homes are increasingly a growing topic of interest amongst IoT technologies. Informetis, provides a cloud based service which, through the use of a ‘single clamp’ sensor, provides consumers with an ‘itemised view’ of their electricity bill. In other words, the Informetis solution can transform ordinary and ‘non-connected’ appliances such as fridges, washing machines and microwaves into ‘virtual’ smart appliances. Your task is to design and implement an application which complements the itemised power consumption data (that we will provide) with other 3rd party data such as temperature, personal information and social demographics data to make the combined user experience more useful and fun. Your ultimate goal is to make the user far more engaged with their energy consumption than they are today!
Alternative
When electrical appliances "waste" electricity, they convert it into heat. The irony is that if you "save" energy by switching off a TV or lightbulb, you then have to switch on a heater to stay warm! Since heating accounts for most energy use in the UK, The only really effective way to save energy at home is to turn the thermostat down. But nobody likes to be cold. You have a business opportunity to compete with an incredibly successful Internet of Things product - the "Nest" intelligent thermostat. Instead of wasting energy on toasty houses, your product will help customers stay warm for free by regulating your own comfort. Monitor humidity, differential airspeed etc to maintain comfort by putting on suitable clothes, eating hot food or even going to bed.
