Load disaggregation for heat pumps and their possibilities

Heat pumps have developed enormously in recent years. Due to environmental issues and the high prices for fossil fuels, the number of heat pumps is increasing from year to year. According to forecasts, two million new heat pumps will be connected to the power grid each year by 2025 in Europe alone. Thanks to their ability to transfer heat from a hot to a cold source, they inherently consume less energy than classic heating solutions such as underfloor heating.

According to a study by EnergieSchweiz [1] from 2021, the average efficiency of heat pumps in Switzerland is between 3 and 4.5. This means that a heat pump heats the household 3 to 4.5 times more than a classic electric boiler for the same electricity consumption. Although they have high efficiencies, heat pumps are still large consumers of energy and thus a major factor in reducing electricity costs for end users and in terms of grid stability for utilities.

Heat pump monitoring for better grid management

Active monitoring of heat pump usage is therefore crucial for both end users and utilities. Based on the monitoring data, several improvements can be made. Utilities can use smart meters to more accurately predict and forecast load, providing a better estimate of the amount of electricity to purchase or produce and a better understanding of the grid. End users can track their heating habits and make adjustments based on this information. In case of anomalies in operation, notifications can help take timely action. In the future, monitorable loads will also be the first step towards smart grid initiatives and flexibility solutions (e.g. SmartGridready), which will enable greater use of renewable energy and energy saving measures.

How can heat pumps be monitored non-invasively and effectively?

Monitoring can be done with an electricity meter such as the CLEMAP Energy Monitor connected directly to the heat pump or via a smart meter. As part of the Social Power Plus research project(SSP), we had asked ourselves whether it is possible to monitor heat pumps with existing hardware such as a smart meter. The normal case is for a smart meter to record the electricity consumption of the entire household, not just the heat pump. Non-Intrusive Load Monitoring (NILM) is a machine learning technique that allows us to break down the loads running in a household based on a single meter point. With NILM, we can therefore use existing smart meters and only need to add one software component, which greatly reduces setup costs.

CLEMAP machine learning algorithm for heat pump monitoring

Within SSP, CLEMAP developed a NILM algorithm for heat pump disaggregation that aims to raise people's awareness of their energy consumption. The algorithm works by processing electrical time-series data of active (P) and reactive (Q) power - summed over the three phases - with a temporal resolution of 15 minutes. Heat pumps exhibit specific electrical profiles and can thus be detected by statistical methods. The machine learning algorithm developed by CLEMAP was evaluated using a reference data set. For this purpose, different electrical data were collected during several months: those from smart electricity meters measuring the whole household and from specific smart electricity meters for the heat pumps in 20 households in eastern Switzerland equipped with different types of heat pumps.

For a machine learning model, the following metrics can be evaluated:

The accuracy of heat pump detection: how well can we predict whether the heat pump is on or off at a given time? (Classification problem)

Energy breakdown error: How well can the proportion of total energy consumed by the heat pump be broken down when looking at energy consumption over a period of time? (Regression problem)

These metrics were evaluated for the entire data set with the following results:     

Heat pump detection accuracy

Current research reports ([2], [3]) state an accuracy of 90-95% for the detection of large loads like heat pumps - under laboratory conditions and sometimes with a higher sampling rate. Our accuracy of 86% in field tests can therefore be considered a success.

To better understand the result, we can look at a boxplot showing the metrics for individual households and the statistical indicators for the entire dataset. Here we look at the accuracy of heat pump detection (balanced_accuracy_state_ts[%]).

The scattered blue dots on the left indicate the accuracy of each household. One point can be seen that has a poor accuracy of about 40%. This outlier was investigated and it was found that its electrical profile was significantly different from the other heat pumps, leading to this result. It was decided to leave the outlier in this benchmark to show that while the model performs well, fine tuning may be needed for specific cases. In the future, further analysis based on other data sets and over larger time periods will further improve the performance of the algorithm.

Results and next steps

Heat pumps are developing rapidly and offer both environmental and cost benefits. Monitoring them is critical for end users and utilities as well as for heat pump manufacturers (e.g., the field performance of their installed base). This monitoring can be done in a non-invasive way using machine learning algorithms and using electricity data from existing smart meters to keep costs down. CLEMAP has developed a powerful algorithm that detects heat pumps with 86% accuracy. In the future, this will enable the implementation of energy reduction, grid efficiency and flexibility management solutions. The current implementation of the solution is done through the existing customer interfaces, where customers have the possibility to explicitly use their data. As the outcome of improved efficiency is evident, regulators must adapt and allow such solutions to be effectively implemented and deployed on the power grid as the volume of smart meter data grows.

Sources .

[1] SwissEnergy, report "Field measurements of heat pump systems heating season 2020/21".

[2]"Results from Non-intrusive Load Monitoring", A. Hutter et al., CSEM Scientificand Technical Report 2016.

[3] Salani, M., Derboni, M., Rivola, D. etal. Non intrusive load monitoring for demand side management. EnergyInform 3, 25 (2020). https://doi.org/10.1186/s42162-020-00128-2

Relevant solutions

Discover matching solutions to this blog post.


Sign up for our newsletter to never miss anything else from CLEMAP.
Thank you very much! We have saved your email.
Your request could not be sent. Please try again.