A distribution planner is reviewing a neighborhood where transformer load has remained relatively flat for years. Over the same period, customers there have bought EVs and swapped gas furnaces for heat pumps. Where did that new load go?
Most of it is still there. Rooftop solar now serves a growing share of it behind the meter, so AMI data never records it. The historical trend looks flat until a hot, overcast afternoon, when the transformer carries load the planner never saw in the data.
For utilities making infrastructure investments that will last decades, that gap matters. Net load shows what the grid carried in the past. Forecasting what it will carry next requires knowing how much customers consume and how much their panels generate—separately.
The Challenge: Separating Consumption from Generation
Advanced metering infrastructure (AMI) measures electricity flowing between customers and the grid. For customers with rooftop solar, the meter typically records net load rather than total electricity consumption.
Net Load = Electricity Consumption − Solar Generation
Consider two neighborhoods drawing 2 MW from the grid on a sunny afternoon.
In the first, customers consume 2.2 MW and generate 0.2 MW from rooftop solar. In the second, customers consume 4 MW and generate 2 MW.
Their net demand is identical, but their underlying consumption and generation profiles are dramatically different.
Historical net load remains essential for understanding actual grid loading. But forecasting future demand requires understanding the independent factors driving it.
The challenge isn’t measuring today’s net demand. It’s understanding how that demand will change as electrification and distributed generation accelerate.
The Science: Recovering the Signals Behind Net Load
Separating electricity consumption from rooftop solar generation presents a fundamental data science challenge: the utility observes a single net load signal but needs to estimate two underlying components.
With only net load, multiple combinations of consumption and generation could produce the same meter readings. Recovering those components requires additional information about how electricity consumption and solar production behave.
Recurve’s FLEX platform addresses this challenge through a physically informed modeling approach that combines interval meter data with geographically relevant solar irradiance information.
Using Physics to Constrain the Problem
Solar generation follows predictable physical relationships with available sunlight, although actual production varies with system orientation, shading, weather, and other site characteristics.
FLEX uses solar irradiance data to establish physically grounded constraints on the generation embedded in net meter readings.
The methodology checks observed load patterns against available sunlight, so FLEX doesn’t attribute every dip in daytime demand to solar..
This allows FLEX to estimate solar production without requiring dedicated generation meters or detailed specifications of each customer’s solar installation.
Separating Consumption from Solar Production
Electricity consumption and solar generation exhibit different temporal behaviors.
Solar production rises and falls with sunlight. Consumption follows customer routines, which repeat by day, week, and season.
FLEX uses these differences to distinguish the two signals, including information from periods when solar generation is limited to better characterize underlying consumption.
The methodology evaluates consumption and generation together, refining estimates.
Modeling Individual Customers, Not Population Averages
Solar installations vary significantly in orientation, shading, and production characteristics. Similarly, electricity consumption patterns differ across customers.
FLEX estimates solar generation separately for each meter, without assuming neighboring customers behave alike.
The methodology accounts for seasonal changes in each installation’s solar production. These individual estimates can then be aggregated across customer segments, geographic areas, and larger populations.
Understanding Model Uncertainty
Solar disaggregation produces estimates rather than relying on solar telemetry, making uncertainty an important consideration for analytical applications.
Recurve evaluates its methodology against known measured solar production and characterizes estimation uncertainty based on observed consumption and generation patterns.
While individual meter estimates can exhibit variability, aggregating results across larger populations can reduce the relative impact of individual estimation errors.
Analysts can then judge the reliability of disaggregated estimates at different levels of aggregation, particularly when using the results for planning and forecasting.
Finding the Customers Net Load Hides
FLEX makes estimated solar generation and underlying consumption available for analysis across individual customers and populations, enabling utilities to explore demand patterns that net meter data alone cannot reveal.
Utilities can identify customers based on estimated solar generation, compare gross and net consumption profiles, and analyze how rooftop solar changes daily and seasonal load shapes.
They can also segment customers by gross electricity consumption during specific time of use periods, including summer and winter peaks. This reveals customers with substantial underlying demand that may be obscured by solar generation.
For example, utilities can identify populations with high summer afternoon consumption despite low net demand, find customers with elevated winter peak usage, or isolate those with both substantial solar production and high underlying electricity needs.
These insights support more precise customer targeting, DER analysis, and demand flexibility planning, and provide a clearer picture of how consumption and distributed generation vary across a service territory.
Better Inputs for Distribution Planning
Distribution planning increasingly requires understanding how customer demand will evolve rather than simply extrapolating historical net load.
By separating consumption and solar generation, utilities can evaluate the independent drivers of demand and develop more informed scenarios for future electrification and DER adoption.
What happens to feeder demand if EV adoption accelerates while solar penetration remains unchanged? How will additional rooftop solar affect daytime loading and evening peaks? Where is underlying consumption growing despite relatively stable net demand?
Combined with distribution asset information and FLEX’s scenario modeling, disaggregated consumption and generation provide additional inputs for evaluating future capacity requirements and infrastructure investments.
Contact us to see how solar disaggregation in FLEX can help your utility understand changing customer demand and plan for the grid ahead.