PV Disaggregation generates consumption data for PV homes using machine learning
More and more homes are being equipped with Photovoltaic (PV) Solar Panels, transforming them from pure energy-consuming locations into both electricity-producing and consuming locations. When a location installs a PV system, it introduces new data streams that contribute to the overall energy usage. Unlike a pure energy-consuming location that only consumes imported electricity, a PV-equipped location has a more complex energy profile, where the produced energy can either be self-consumed or exported.
Accessing all the necessary data to provide PV insights to end users can be challenging, as it requires data from both the PV inverter and the electricity meter. Data from PV inverters typically require continuous integrations to stream and provide insights. Homeowners often have a wide array of inverters, each requiring unique integrations to gather the data for all end users.
PV Disaggregation is a feature designed to address this challenge. It generates consumption and production data using the electricity meter's data alone, unlocking insights for all PV-equipped homes through algorithms and machine learning, without needing access to PV inverter data.
Eliq can support the receiving and returning of PV-related Energy Data to and from the Eliq Platform that help consumers track and understand their Energy Usage over time.
PV equipped homes have 4 streams of data that for electricity; Production
, Consumption
, Import
& Export
that are interesting to show to consumers in as high resolution and detail as possible.
Eliq offer 2 high-level Use Cases for PV Owners:
Import
, Export
, Production
, Consumption
data. We call this experience PV Submetered Insights.Production
and Consumption
data. We call this experience PV Dissagregated Insights.There are 4 Data Streams that are typically available from a PV-equipped home as described below.
Import is the energy that is “bought” from the the grid (sometimes called offtake) and that a PV equipped home will pay for on their energy bill. This Data is usually available from a Smart Meter in the consumers home.
Export is the energy that is “sold” to the grid (sometimes called injected) and that a PV equipped home will get payed for in some way, usually with a discount on their import energy bills if energy is bought and sold from and to the same energy company. This Data is usually available from a Smart Meter in the consumers home.
Production data is the electricity that is generated by the PV system and that can exported to the grid or consumed by the home.
import
, export
and weather data, calculates the production data for the location.Consumption is the energy that the home consumes that is either provided through Import or through Production.
import
, export
and weather data, calculates the consumption data for the location.PV Disaggregation is a feature that fills a gap for consumers that have a PV system where only 2 of the 4 required data streams are available, namely Import
and Export
. If a home can only provide Import and Export data streams, it isn’t possible to show insights related to Consumption such as forecasting
, Similar Homes Consumption
and Energy Usage Categories
since they require Consumption Data for a location.
PV Disaggregation solves this, unlocking full insight into a homes 4 data streams and unlocks all Consumption Insights for he PV equipped home.
PV Disaggregation makes it easy to offer great insights to any PV equipped home without a need to gaining access to every homes PV system.
If all 4 data streams are available, Eliq can offer PV Submetered Insights for those end users. It is possible to offer a combination of both disaggregated and submetered insights in the same app by applying dynamics.
The Eliq PV Disaggregation system can be divided into two parts
PV Parameters – are attributes of the PV installation that we estimated based on imported
and exported
energy data in order to be able to estimate (disaggregate) produced solar energy. These parameters are – azimuth of the PV system, tilt of the PV system, size and efficiency of PV system, and temperature coefficient.
Weather ML model – is model that is trained on actual solar energy production data and weather data. The aim is to have an estimate for weather effects on solar production. Cloudiness, precipitation, temperature and other factors directly and indirectly affect how much solar irradiation is converted to electric energy.
import
and export
) in hourly or higher resolution.This is the only real requirement to have PV parameters estimated. However, with only 7 days of data, we won’t have good PV parameters. Therefore we will continuously update those in the coming months (quite frequently to start and less frequently as we have more data) up to the point where we have a PV parameters based on 1 year of data. After we have those parameters, we consider them stable
and we will update them once every 6 months at best.
PV Disaggregated Insights are only available for residential homes (Locations with residential
Profile Types). For more information about Location Profile Types
, please visit the Location Profile article
Eliq runs a daily process that detects new locations with export
energy data and starts by estimating PV parameters for each location for which PV Disaggregation has been enabled. Each individual location takes around 30 seconds (as of 2023-10-31) to estimate (the latencies will improve as we improve the process). Depending on the number of new locations needed to process, the entire daily process can take 1 hour or more to complete.
PV Disaggregation runs for every location with Estimated PV Parameters. Each time energy data is added, which completes the day, we trigger the disaggregation service and saves Production data to the platform. Once this is done, we can return all 4 data streams through our Insights API Locations endpoint.
PV Disaggregation Insights depend on our ability to estimate weather effects to solar energy production in a particular country. To make sure that all the relevant weather patterns are accounted for and we have the best possible accuracy we localize PV Disaggregation Insights to each country based on ground truth data. We also offer a fallback mechanism for the clients that need time to gather this data.
To get the best outcome (lowest error) it’s best to tailor Weather ML model for each country (distinct geographical conditions). For this we need ground truth data – solar energy production, as well as, import and export energy data. We then train model which estimates weather effects comparing real energy production with the ideal conditions estimate provided via the PV parameters.
Collecting actual PV production data is not straightforward unless a client does has direct access to the systems. Contacting and incentivizing clients takes time. With this in mind we have created an Europe ML Weather Effects model. This model is trained on all data we have in the different European markets to reflect the average weather conditions across the continent. With this model we can start providing PV Disaggregation Insights for our clients while they are gathering ground truth data. This convenience comes with a slight price – the generic model has about 5% higher MdAPE error compared to it’s country specific counterpart. We believe it’s acceptable price to pay for the convenience and speed that this unlocks, however to offer the best possible experience we should localize countries as soon as possible to collect ground truth data.
Offering Insights to consumers who have a PV equipped home is something that needs to be set up by Eliq, please contact us for more details on how to do this.
Data can be added through the Eliq Data Management API. All locations must fulfill the Data Requirements (see above).
Apps can determine whether or not to present PV Disaggregation Data Streams in an app dynamically, by utilizing the “Location Metadata” in the Insights API Endpoint GET Location . Information to help determine whether or not a Location has PV data and if the source of the data is based on the Eliq PV Disaggregation feature can be determined in the response. Example below.
"production":
{
"resolution": "hour",
"data_from": "2020-01-01T00:00:00",
"data_to": "2021-01-01T00:00:00",
"source": "pv_disagg"
},
"consumption":
{
"resolution": "hour",
"data_from": "2020-01-01T00:00:00",
"data_to": "2021-01-01T00:00:00",
"source": "pv_disagg"
}
Visit the Getting Started documentation for more details on getting started using the Eliq Platform
API Endpoints
Use Get Location to determine which data is available and to build app dynamics. The source
parameter determines if PV disaggregated data exists or not as well as whether or not Import
and Export
Data exists.
Use Get Location Production to fetch Disaggregated Production Data for a location with pv_disagg
.
Use Get Location Consumption to fetch Disaggregated Consumption Data for a location with pv_disagg
.
Use Get Location Import to fetch Import Data for a location with a PV system.
Use Get Location Export to fetch Export Data for a location with a PV system.
With the import/export summary card in the home screen users will get a quick summary of the last day and ongoing months data streams. If a location has Import/Export price formulas, you could show price as well.
Example of how to visualize import/export data in graphs with different resolutions
Example how you can toggle between import/export and consumption/production within graphs