Energy Consumption is a topic of great interest, especially since a surge in prices in late 2021 has caused a considerable increase in discussion around the topic. Data from the Swedish Central Bureau of statistics (SCB) and the Swedish Meteorological Institute (SMHI) were provided for macroscopic regressors. These regressors are temperature, population, GDP, day length, electricity price, electricity production, production of variable renewable energy and average income in order to predict electricity consumption. Four models were created, a full multiple linear regression model using all regressors. A reduced multiple linear regression model using a subset of the regressors determined by cross validation. A ridge model and a LASSO model. These were then used to attempt to predict the power consumption of 20% of the data set that were left out when creating the models. The LASSO model was most successful in this as it had the smallest cumulative residual and the ridge model was the worst. Since the reduced and the full model both had very high multicollinearity the conclusion was that the LASSO model is the best model out of the four.
| Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:kth-314660 |
| Date | January 2022 |
| Creators | Moloisel, Victor, Lind, Carl-Fredrik |
| Publisher | KTH, Skolan för teknikvetenskap (SCI) |
| 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 | TRITA-SCI-GRU ; 2022:058 |
Page generated in 0.0017 seconds