Solar PV Performance Monitoring

Performance Monitoring

Irradiation Performance Monitoring

We have been working with a leading data science consultancy whose products and support are used by the National grid. The objective to integrate observed irradiation data into our monitoring platform in a way that is meaningful and reflects a sites installed components and attributes.

Domestic rooftop and C&I sites rarely have pyranometers installed. As such it is difficult to know what sites in any given location should have produced daily with any degree of accuracy.

Many portfolios are managed against Industry datasets based on long term average irradiation. This approach makes it impossible to manage O&M against accurate irradiation for performance guarantees. Or if adjustment factors are applied for month or quarter end reporting any underperforming issues will be in the past meaning the generation and associated revenue in the period are lost.

Approach

• Site historic generation, availability and irradiation data creates a site specific digital twin that reflects installed components, capacity, orientation, pitch etc Also serves to check the accuracy of asset registers
• Measured irradiation from official world meteorological sites is integrated into our platform and translated by the digital twin for each site providing accurate daily generation numbers that reference actual irradiation
• As this is calculated daily as a minimum (can be up to half hourly depending on system size and preference) specific underperformance alerts are triggered, reasons identified meaning generation and associated revenues can be maximised quicker and not lost

Data Sources

• Weather observation data from official World Meteorological Organisation (WMO) is used – approximately 250 sites across the UK with good spatial coverage.
• Simulated observations from an extended set of approximately 1000 virtual sites across the UK used to fill in any spatial gaps in WMO network.
• Data produced by blending outputs from the 2 main UK weather models – UK Met office and European Centre for Medium Range Weather Forecasting ECMWF.

Output

• Using complex algorithms expected site generation is split into statistically probable categories including P50, P90 and a blended model that draws data from sites in close proximity if there is insufficient site specific data for the model to work
• 12 month site specific rolling forecast based on updated long term average information from measured irradiation and site performance
• The table below outlines all model outputs that are generated daily for each site.

Daily Generated Model Outputs

Field	Description
metered_generation	Actual metered generation from the site
model_estimate	Modelled estimate of what is expected from the site based on historic behaviour and actual weather conditions
p95_low	95% confidence lower limit, low level of the confidence band that with 95% probability should be the generation figure
p50_low	50% confidence lower limit, low level of the confidence band that with 50% probability should be the generation figure
p50_high	50% confidence upper limit, upper level of the confidence band that with 50% probability should be the generation figure
p95_high 95% confidence upper limit, upper level of the confidence band that with 95% probability should be the generation figure	95% confidence upper limit, upper level of the confidence band that with 95% probability should be the generation figure
clear_sky_limit	Upper limit of generation based on clear sky weather conditions; note this is a modelled value so can be exceeded in extreme conditions
normal_weather_estimate	Estimate of generation that would be expected based on seasonal normal weather conditions. Rolling long term (10 yr) seasonal average, mapped to weather locations.

Peer Review Methodology