Use of Digital Twin Weather extremes for fine-tuning energy prediction

Gerard Mor Martinez

Míriam Méndez Serrano

Presentation of the DestinE Use Case

 

 

 

April 12th, 2024 10:00

 

Statistical learning methods for energy assessment in buildings with applications at different geographic levels

- Background -

Electricity load characterization of districts

Supported by:

ELISE (ISA2 project) and Joint Research Center (Ispra)

Journal: Energy Reports

JIF (2020): 6.870 - Q1

Goals

To design and implement a visual tool that characterises energy consumption at a local scale (postal code) based on forecasted weather data

  • Accounting for different conditions:
    • Building characteristics
    • Activity sectors
    • Weather
    • Socio-demographic
  • Using interoperable datasets:
    • INSPIRE-harmonised datasets for cadastral data
    • Weather datasets coming from DEDL
  • Geographical scope: Catalonia (a region of Spain)

Data inputs (2018-2023)

  • Aggregated hourly electricity consumption
    • By tariff
    • By similar-to-district geographical level (Spain: postal code)
    • By type of usages (Industrial, residential, tertiary)
    • Indexed prices for PVPC energy tariff
  • Building information (INSPIRE harmonised)
    • Typology
    • Built area
    • Year of construction
  • Weather data
    • ERA5Land (historical data)
    • Digital Twin extreme weather forecast

Optional data inputs

 

  • Aggregated socio-economical information
    • Annual net incomes per household
    • Incomes sources (salary, pension, benefits,...)
    • Gini index (inequality)
    • Population age
    • People per household
    • Population

Consumption data      Socio-economic data       Buildings information

Postal code         ~          Census tract          >        Building level

Aggregation

Energy characterisation method

Once datasets are harmonised to postal code level: 

  1. Data cleaning
  2. Inference of usage patterns
  3. Modelling the electricity consumption using ML techniques
  4. Outcomes of disaggregate consumption
    • Baseload
    • Space heating
    • Space cooling
  5. Calculate KPIs
  6. Use the model to predict multiple weather scenarios

 

Conclusions

This methodology will be:

  • Capable to infer the drivers of consumption in the building sector and their occupants.
  • Predict electricity consumption of large regions using extreme weather forecast.
  • Demonstrates the potential of using DT, open data and INSPIRE harmonised datasets

Text

Thanks for your attention

 

Use of Digital Twin Weather extremes for fine-tuning energy prediction