It is time to phase out fossil fuels and invest our efforts in green energy production through a major restructuring of the energy system. At the same time, more people are acquiring electric vehicles (EVs), thus creating a higher demand of electricity, and solar panels, allowing the consumer to also be a micro-producer. In order to systematically perform these changes, it is important to gain a better knowledge of the current customers as well as be able to make more accurate predictions about their future consumption. Vattenfall Eldistribution (VE) is one of several operators of the electric grid and, as of this day, still produces effect forecasts based on static estimations using the Velander formula. This has been a successful method in the past, however, with the current rate of change and the complexity in the consumption behaviour, it has become more difficult to estimate the aggregated load on the grid. It is also unattainable to cover the future demands by only expanding the grid. This creates the need for optimising the current grid, making more dynamic effect forecast and creating a smart grid. Our purpose is to help VE develop typical load profiles (TLPs), a more dynamic way to estimate peak loads, for private customers in the Uppsala region. VE provided us with time series data regarding the customers' consumption, as well as, attribute data describing the fuse size, heating system, contract type, etc., of these customers. A third dataset was also acquired through the Swedish Transport Agency regarding EV owners. These datasets allowed us to implement the three different parts of this project. The first part involved the creation of Attribute based TLPs with the help of the different attributes found in the VE's database. The goal for this part was to investigate the impact of specific attributes on the TLPs. The second part concerned the development of Behaviour based TLPs by implementing clustering algorithms that groups the customers based on behaviour alone. Thereafter, the distribution of attributes in the different groups was examined, in order to evaluate if there is a connection between the attributes and the consumption patterns identified. The third part studied the effect of EVs on the consumption behaviour. For this part, we implemented both attribute and behaviour based TLPs. The results of the Attribute based TLPs part concluded that fuse size has minimal impact on the TLPs whereas heating system entails a larger variation. In the second part of the project, Behaviour based TLPs, TLPs were successfully created with the help of clustering algorithms. However, no clear linkage between the consumption patterns and the attributes could be determined due to an evident overlap in the attributes between the created clusters. The final part of this project, EV owner based TLPs, verified the hypothesis that EV owners most likely charge their vehicles during the evening and night and established a clear visual increase in the consumption pattern in relation to non EV owners. An overall uncertainty that affects the results of all parts of this project is the accuracy of VE's data attributes and in order to confirm the conclusions of this thesis the degree of accuracy of the attributes should be determined.
| Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:uu-477326 |
| Date | January 2022 |
| Creators | Manousidou, Aikaterini, Lundberg, Martina |
| Publisher | Uppsala universitet, Industriell teknik |
| Source Sets | DiVA Archive at Upsalla University |
| Language | English |
| Detected Language | English |
| Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
| Format | application/pdf |
| Rights | info:eu-repo/semantics/openAccess |
| Relation | UPTEC F, 1401-5757 ; 22 023 |
Page generated in 0.0025 seconds