Loading...

Forecasting Electricity Consumption in a Metropolis Using Machine Learning Algorithms

Roostaei, Mohammad | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57752 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Taghizadeh Manzari, Mehrdad; Aryanpour, Masoud
  7. Abstract:
  8. In smart grids, energy data such as electricity and gas consumption are recorded via smart sensors installed in buildings. Analyzing energy consumption patterns and classifying users based on their consumption behavior can assist policymakers in supply management and urban-scale energy control. This study proposes an unsupervised approach to classify energy consumers according to their monthly consumption patterns. The primary objective of the clustering is to group households with similar usage behaviors, facilitating a more effective analysis of electricity consumption at an urban scale. The K-means clustering algorithm is employed to classify households into K clusters based on their consumption data over various monthly time windows. The results reveal three distinct groups of buildings in the studied smart grid in London: low, medium, and high-consuming households. To enhance clustering performance and reduce computational complexity, Principal Component Analysis (PCA) is utilized to reduce data dimensionality. A key outcome of this approach is the creation of a new feature, Principal Component 1 (PC1), which can independently classify residential households, thereby reducing the number of attributes required for clustering. Finally, Long Short-Term Memory (LSTM) networks are applied to predict electricity consumption for each cluster. The findings from this analysis provide insights into underlying electricity consumption patterns and offer valuable guidance for city planners to develop more effective energy policies. The dataset used in this study is a real-world dataset containing half-hourly electricity consumption records for 5,567 households in London from November 2011 to February 2014. This dataset is publicly available through the London Datastore under the title "SmartMeter Energy Consumption Data in London Households."
  9. Keywords:
  10. Energy Assessment ; Machine Learning ; Electricity Consumption Efficiency ; Time Series Analysis ; Energy Consumption Time Series ; Urban Scale Electricity Consumption ; Energy Consumption Optimization

 Digital Object List

 Bookmark

No TOC