Turning Dust into Data: A Technical and Operational Look at Soiling Losses in Solar Operations

In solar power plants, even a thin layer of dust, sand, or salt can make a big difference.

Soiling, the gradual accumulation of dirt on PV modules, is one of the most common and costly sources of performance loss. Yet it's also one of the hardest to measure accurately.

For operators managing large or geographically dispersed portfolios, the challenge is clear: installing soiling stations everywhere is rarely practical. Decisions about when to clean, how to estimate losses, or where to prioritize resources often rely on assumptions or siloed data.

The result? Uncertainty, unnecessary costs, and missed production.

TGS Prediktor's Soiling Analysis feature in PowerView changes that, by turning "dust" into reliable, actionable data.

From Uncertainty to Quantifiable Impact

PowerView's Soiling Analysis estimates soiling losses over time and across geography without requiring specialized sensors.

Instead, it combines:

• Production measurement data on string or inverter level
• Weather sensor measurement data
• Physics-based performance models
• Advanced signal processing and analytics

This enables PowerView to isolate the soiling effects from other performance drivers such as asset technical issues, degradation, shading, temperature, clipping, or seasonal irradiance variations.

The methodology, developed by TGS Prediktor's data engineering and product team, uses a transparent and explainable approach, validated against field measurements and peer-reviewed academic models - no black boxes, no hidden corrections.

Soiling shown in a solar park

The Technical Side: How Dust Becomes Data

Behind every bit of "dust" lies a complex set of variables influencing PV performance:

• Particulate composition (organic, inorganic, agricultural, industrial)
• Particle size distribution
• Local humidity and dew cycles
• Wind speed and direction
• Panel tilt and geometry
• Proximity to sea, farmland, or desert
• Rainfall patterns and soiling recovery after rain

Historically, soiling assessment has been difficult and cumbersome, as it relies on visual inspection or specialized equipment that requires daily maintenance.

Today, modern plants generate vast amounts of data; the challenge is extracting the right signals from noise.

PowerView's Soiling Analysis does this through a four-stage technical pipeline:

1. Data Capture

Performance traces, irradiance data, weather data, cleaning logs, and site metadata feed into a unified data layer.

2. Data Standardization & Quality control

Raw inputs are contextualized, normalized, timestamp-aligned and validated in accordance with governance protocols. This step guarantees consistency and comparability across different assets and regions.

3. Multi-Source Data Fusion

Information from diverse systems, including inverters, string combiners, weather stations, and operational logs, is aggregated and fused to create enriched datasets. Adding context enables us to derive deeper insights into the PV plant performance beyond what isolated measurements can do.

3. Advanced analytics and modelling

Advanced analytic algorithms run on rich and large datasets containing both historical and real-time data to quantify soiling at the inverter level.
At the core of this process is the Combined Degradation and Soiling (CODS) algorithm, illustrated in the figure below. In essence, CODS applies an iterative decomposition technique to separate overlapping performance signals into three distinct components:

• Sawtooth pattern representing soiling accumulation and cleaning events
• Linear trend capturing long-term system degradation
• Sinusoidal variation reflecting seasonal effects

This approach enables precise differentiation between soiling losses and other performance factors, ensuring more reliable asset monitoring and optimization.

CODS iterative optimization process

Å. Skomedal and M. Deceglie, “Combined estimation of degradation and soiling losses in photovoltaic systems,” IEEE J. Photovolt., vol. 10, no. 6, pp. 1788–1796, Nov. 2020

4. Decision Support

Operators receive:

• Alerts when soiling passes economic thresholds
• Cleaning recommendations
• Post-cleaning validation to quantify the actual gain

This transforms soiling from a reactive maintenance task into a predictive, portfolio-level strategy.

What the Data Reveals

Once transformed into actionable metrics, soiling data offers insights that were previously invisible:

• Optimize cleaning cycles: Instead of calendar-based cleaning, teams can clean exactly when the performance loss justifies the cost.

• Quantify energy losses: Operators can finally answer: How much energy did we lose to soiling last week? Last month?
• Compare regions and climates: Some sites foul slowly and predictably; others fluctuate sharply with humidity, agriculture or industrial activity.
• Compare soiling in different locations around the plant: What is the impact of the factory located close to some of the panels of the plant?
• Validate cleaning ROI: PowerView shows the performance rebound after each cleaning event, enabling data-driven contracting.
• Identify high-impact soiling periods: Daily or weekly curves clearly show when deposition accelerates.

From Data to Decisions

Cleaning cycles account for a significant share of O&M costs. Cleaning too early wastes money; cleaning too late wastes production.

With PowerView's Soiling Analysis, operators gain a factual basis for decisions such as:

1. When is the performance loss significant enough to justify a cleaning?
2. Which sites are most affected by dust, humidity, or sea salt?
3. How do soiling patterns evolve seasonally and year over year?
4. What is the optimal cleaning frequency for each asset?
5. What used to be manual and reactive becomes repeatable, verifiable, and data-driven.
6. What is the most appropriate location for my next PV plant to minimize soiling?

Engineering Validation & Methodology

The development of this module was a collaboration between Prediktor's data engineering and product teams, IFE (Institute for Energy Technology) and Scatec, integrating:

• Field experience from operational PV assets
• Close partnership with the PV assets owner to understand the product requirements
• Rigorous testing against the CODS iterative optimization method
• Validation against reference models from literature
• Comparisons with sensor-based soiling data
• Statistical performance baseline correction

As Pierre Turquais, Data Scientist, explains:

"Our goal was to make soiling analysis accessible without new hardware. The data already exists in the plant; we just needed to extract the right signals and transform it into soiling insight for O&M teams."

A New Layer of Operational Insight

Soiling Analysis is available in PowerView, giving owners and service providers the ability to:

• Monitor soiling behavior at portfolio scale
• Benchmark assets across regions
• Plan cleaning interventions based on measurable gain
• Combine soiling insights with fault detection, weather, and maintenance history
• Improve overall yield and reduce unnecessary costs

It's another step toward helping operators run their portfolios based on facts, not assumptions, making decisions that directly impact production and profitability.

Learn more

Explore how PowerView enables continuous performance analysis, loss detection, and smart operations.