# :computer: Electricity Forecasting Model (EFM) for Swedish Households

###### tags: `Tag([e-mail])`

> **Author**: Sotirios-Ilias Tsivras ([e-mail])
> **Project outline**: The [EFM](https://solar.cs.kau.se/) application has been developed with the aim to forecast the electricity consumption of Swedish households. It bases on machine-learning (ML) algorithms trained from electricity consumption data collected from 400 Swedish households by the [Swedish Energy Agency](https://www.energimyndigheten.se/globalassets/statistik/festis/elmatning-i-bostader/final_report.pdf). Datasets of Photovoltaic (PV) generation are also included using PVGIS in order to forecast the self-consumption-rate (SCR) of roof-mounted PV installations.
> **Acknowledgements**: The author would like to thank the Swedish Energy Agency for giving him access to the datasets during his MSc studies and PVGIS for developing simulations of solar power generation.

## :high_brightness: A few words about EFM

During the years 2005-2008 the **Swedish Energy Agency** (SEA) run an end-use metering campaing in 400 Swedish households with the aim to promote electricity savings in the residential sector. After completing this research, SEA developed the **e-nyckeln** :key: database, including the electricity consumption of each appliance running inside these households.

During his studies in Dalarna University (MSc in Solar Energy Engineering) the author had the opportunity to use this information in order to develop a **machine learning** (ML) model in order to forecast the **Self-Consumption-Rate** (SCR) of potential PV installations in Swedish households. The current version of this model is clustering households and users based on a set of parameters, such as the living area, the number of people living inside and the appliances they mainly use.

This model is now available online as a web application written in **Python** (Django framework) with the aim to promote de-centralized and small-scale investments in renewables and energy-efficiency (encouraging people to be **prosumers** rather than consumers). The author's intention is to continue transforming all this complex information into simple graphs and to extract as much knowledge as possible.

## :keyboard: Why should you use the EFM app?

Residential electricity consumption depends on a number of parameters, such as the logement type (detached house or apartment), number of occupants, living area, type of heating appliances etc. It also depends on the habits the occupants have and the time they spend at home. Based on e-nyckeln database an **artificial neural network** (ANN) has been developed that encapsulates all this information and tries to forecast the load demand of a Swedish household.

On the other hand, using PVGIS database one can forecast the electricity generation of a roof-mounted PV system. Combining all this information one can estimate the self-consumption (SCR) and self-sufficiency rate (SSR) of a PV system. These are key-parameters in order to calculate the pay-back period of a PV investment. The SCR and SSR can be estimated using the EFM application.

## :bar_chart: Load demand analysis (LDA)

The LDA app complements EFM by providing useful information of the electricity consumption of a single consumer. The user is requested to fill in the location of the residency (in order to generate the ambient temperature tables) and to upload his/her electricity consumption dataset (can be downloaded from the corresponding electricity provider). These temperature-electricity datasets are independently analyzed in order to calculate the average electricity consumption over a single day or a specific season. The **distribution** of the electricity consumption is also generated in order to compare the skewness and kurtosis among different users.

Moreover, the temperature and load demand tables are being **merged** in order to generate an anomaly score for the user. Anomaly score is used to calculate the daily load demand during a specific season that seems to deviate from the most common pattern. By identifying a deviating curve one can decide whether to take energy-efficiency measures in order to maximize the SCR of a PV installation.

:information_source: For more information please visit [EFM](https://solar.cs.kau.se/).
:information_source: For asking questions and/or giving feedback, feel free to contact the author of this website: [e-mail]