Energy Help

Inspiration

We decided to participate in the hackathon because we had a great idea in mind to implement and test for the benefit of the whole society. Russian invasion of Ukraine caused a rapid and unpredictable increase in the prices of energy resources, that caused a significant increase in household expenses. Part of the society may try to save in the short term, but for many who were already forced to live frugally before this energy crisis, paying the huge bill is an unbearable burden. In such a crisis situation, the government must provide necessary support to society. But how to make it as targeted as possible to those groups of society who need it the most, and without spending to those whose lives it does not significantly affect? The answer to this question lies in the collection and analysis of as many different types of data as possible in their mutual interaction, which is made difficult or even excluded by the fact that the data sets required for analysis are not always collected in one place for effective use, but are also not even conscious. So - let's get started!

What it does

Tool target users are government decision makers and policy planners. The tool would provide four main functionalities (described below).

Collection of all information in one place, as information obtained from various State Information Systems, municipal systems and private company systems in a single data warehouse, including the following data groups:

• data on energy resources consumed by the consumer;
• indicators of consumer purchasing power.

Offering the applicable discount amount to the consumer depending on the various "characteristics" of the consumer (consumer "profile"):

• revenues;
• owned properties;
• household composition;
• turnover/profit of the company;
• consumption of different energy resources in mutual connection;
• possible statistical expenses for other payments (food, housing, medicine, etc.);
• amount of real estate tax;
• already received discounts;
• use of environmentally friendly resources;
• etc.

Modeling the preliminary impact of the calculated discounts on the state budget items using BI means. Analyzing modeling results and making data-based decisions.

This tool will create a household profile that corresponds to the affluence ratio Kp=EXPENDITURE:INCOME (aimed tu be reduced by reducing expenses for consumed energy resources by applying discounts). For example, selection includes households where Kp per person is 0.8-0.9 (quite bad ), then a discount of 5% is applied to this household profile. Then it is checked whether Kp (after the discount) reaches the desired level, for example 0.7, and how much such a discount for households meeting this profile requirements reduces revenues in the state budget, etc. model the desired confidence Kp (maybe 0.75 is enough) and discounts (maybe 4% is enough) and the Kp of the goal to be evaluated and the admissibility limits of the impact on the budget.

How we built it

First of all, we identify the data that could affect particular household's need of support and how much support would be useful. Then we identify which types of data are maintained by which data holders and what is the quality of this data in each data source. When the data and the data holders are aware, we start negotiations with the data holders and conclude appropriate data use agreements, while being aware that for an effective result it will certainly be necessary to obtain and process the data of natural persons, therefore regulations should be achieved in such a way that would allow the data of a natural person (non-personalized ) use for state purposes also without the consent of the person specified in the FPD protection regulations. When legal data is available, we develop specific system-level interfaces (APIs) for receiving data from data source systems and collecting them in a single database, where we update the data with the necessary regularity. Next, based on the data collected in the data warehouse, we allow the user to enter various indicators (the amount of discounts for different groups of households and energy resource consumption volumes), which will provisionally affect both the amount of discounts granted to the specific person or household, and undoubtedly also the state budget, which will have to cover the burden of the granted discounts . Only by collecting and analyzing calculated impact data using BI tools, it is possible to make balanced and correct decisions. To ensure all the mentioned steps, our team performed the following works:

• designed and created database structure (normal form data structure);
• designed and created data warehouse structure (denormalized data structure for data analysis);
• a set of test data was prepared and imported into the database;
• forms for inputting data to be modeled and viewing results have been developed.

Challenges we ran into

• Identifying identificators that characterize household livelihood ratio/level
• It was not yet possible to integrate product baskets with maintained real time price of products' basket
• Converting business requirements into system requirements

Accomplishments that we're proud of

A real operational prototype of the system has been created, and it is completely flexible and can be supplemented with arbitrarily many and various other decision-influencing data, not only in the field of energy cost support, but also in other fields, such as medicine, education etc.

What we learned

• The different usages of machine learning to accelerate development
• What counts into relative expenditure basket positions
• How exciting it is to participate in hackathon, and to work together in a team! :)

What's next for Energy Help

• Offer to adjust the system for real specialists involved in state budget planning
• Identify data sources, their holders, and the quality of that data
• Conclude agreements with data holders to access their data sources
• Adjust legislation to justify the usage of personal data without explicit consent

Built With

• apis
• database