The tool aims to evaluate the energy savings achieved following the implementation of an Energy Conservation Measure (ECM). An ECM may involve actions related to building energy retrofits, such as:
- Window replacement
- Installation of thermal insulation
- Modification of opaque/transparent surfaces and internal spaces
- Improvement of the roof's thermal properties
- Change in lighting type and/or systems
It can also include actions aimed at improving indoor comfort conditions and energy consumption through the installation of automated control systems.
The energy savings assessments are based on the IPMVP (International Performance Measurement and Verification Protocol), which is the international standard for the measurement and verification of energy savings.
The tool relies on the development of analytical models to estimate what energy consumption would have been in the absence of efficiency interventions, i.e., to build the so-called reference scenario or baseline.
An analytical model is a mathematical or statistical representation that describes the relationship between energy consumption and independent variables (such as external temperature, occupancy, operating hours, etc.). This model is calibrated using historical data collected before the intervention.
In the context of IPMVP, analytical models are commonly used in Option C – Whole Facility Analysis. The approach consists of:
- Building a statistical model (e.g., linear regression, polynomial regression, machine learning models, etc.) of energy consumption during the baseline period (pre-intervention).
- After the intervention, using input data (e.g., external or internal temperature, etc.) from the reporting period to simulate what energy consumption would have been according to the model.
- The difference between the simulated consumption (baseline) and the actual consumption represents the energy savings.
- They allow for continuous evaluation of savings, even in the absence of conditions identical to the pre-intervention period.
- They are ideal for complex systems where many variables influence energy use.
- They handle environmental fluctuations (e.g., seasons, weather) more precisely.
Once the dataset (cleaned) is defined, along with the input variables, output variable, and time periods (baseline — divided into test and validation — and reporting period), the tool attempts to generate analytical models using algorithms such as:
- Polynomial Regression
- Random Forest
- LightGBM
- Neural Networks (LSTM)
Starting from the simplest model, statistical parameters are evaluated. A model is considered valid only if it exceeds certain performance thresholds.
Among the input variables, past effects (lagged inputs) are also considered in relation to the current output (also called the target). This is based on the assumption that the effect of certain variables does not manifest instantly, but after a delay (lag). For example, the impact of outdoor temperature on energy consumption may be delayed due to thermal inertia.
For the evaluation and selection of the analytical model, refer to the details described within the following paper: Building performance evaluation through a novel feature selection algorithm for automated arx model identification procedures
Within the main() file, the pipeline to be executed is defined.
In the example shown, the goal is to assess how the use of a home automation system with an adaptive algorithm can reduce energy consumption in a building by managing indoor temperatures during working hours, nighttime, and distinguishing between weekdays and holidays.
The following steps are implemented:
- Generation of an analytical model for indoor temperatures during the baseline period.
- Simulation of the indoor temperature model during the reporting period.
- Creation of an analytical model for the thermal energy consumption of the building, which also takes indoor temperatures into account.
- Simulation of the thermal energy consumption model during the reporting period, using the simulated indoor temperature data as input. These temperatures represent the values that would have occurred in the building in the absence of automation.
- Evaluation of energy savings by calculating the difference between the actual data and the model-simulated data.
Inside the report folder, it is possible to generate an HTML report with KPIs.
Modify all inputs in the generate_report_model.py file to generate the report for your specific case.
Bug reports/Questions If you encounter a bug, kindly create a GitLab issue detailing the bug. Please provide steps to reproduce the issue and ideally, include a snippet of code that triggers the bug. If the bug results in an error, include the traceback. If it leads to unexpected behavior, specify the expected behavior.
Code contributions We welcome and deeply appreciate contributions! Every contribution, no matter how small, makes a difference. Click here to find out more about contributing to the project.
Free software: MIT license
For any question or support, please contact:
- Daniele Antonucci [email protected]
- Olga Somova [email protected]
This work was carried out within European projects:
Moderate - Horizon Europe research and innovation programme under grant agreement No 101069834, with the aim of contributing to the development of open products useful for defining plausible scenarios for the decarbonization of the built environment