Understanding short-term PV power variability

Read full document

Overview

As the penetration of solar energy into global power grids increases, managing the variability of photovoltaic (PV) power generation becomes critical for grid stability and project revenue. While the variability of Global Horizontal Irradiance (GHI) is often used as a proxy for the variability of PV output, including our previous publication, the relationship between solar resource variability and the actual power injected into the grid is complex.

This study investigates the short-term variability of both GHI and PV power output (PVOUT), focusing on "ramps" - sudden increases or decreases in power. By comparing Solargis 15-minute satellite-based data with 1-minute synthetic data, the paper highlights the risks of relying solely on 15-minute time series for high-criticality projects and proposes improved metrics for assessment.

Methodology

The research analyzed data from eight locations representing diverse geographies and climate zones, including tropical, arid, and continental climates (e.g., Brazil, USA, India, and Spain).

The study compared two types of input data:

• 15-minute resolution: Standard Solargis satellite-based time series.

• 1-minute resolution: Synthetic time series generated using Solargis’ proprietary multi-scale hierarchical approach to mimic high-frequency ground observations.

Using the Solargis PV Simulator, we converted this irradiance data into PVOUT, accounting for real-world factors such as inverter clipping, thermal losses, and shading. Variability was quantified by counting "ramps" exceeding specific thresholds (Low, Medium, and High severity) defined by grid impact limits.

Key Findings

The study revealed significant discrepancies between 15-minute data resolution and the reality of high-frequency fluctuations:

• Underestimation of Severe Events: Data with 15-minute resolution consistently underestimated the occurrence of "High" severity ramps compared to 1-minute data, particularly in tropical and temperate climates where cloud cover changes rapidly.

• GHI vs. PVOUT Differences: PV power variability does not perfectly mirror solar irradiance variability. Factors like inverter clipping can reduce the magnitude of ramps, while thermal inertia (the time it takes for modules to heat up or cool down) acts as a dampener on the power signal.

• Seasonal and Diurnal Patterns: In mid-latitude regions, the relationship between GHI and PVOUT ramps shifts seasonally. Notably, during winter mornings and evenings (low sun angles), PV output ramps can be disproportionately large compared to irradiance ramps, likely due to temperature effects and inter-row shading.

Conclusions and Implications for the PV Industry

The findings suggest that single-value variability metrics often used in pre-feasibility studies are insufficient for the detailed design of complex projects.

• Need for High-Resolution Data: For high-criticality projects (large capacity or weak grids), developers should utilize 1-minute time series data and perform full PVOUT simulations rather than relying on GHI proxies.

• Design Optimization: Choices made during plant design, such as the DC:AC ratio, directly influence the variability the plant will experience. Using high-resolution data helps optimize these parameters to mitigate grid impact.

• New Metric Proposed (RMSR): The study proposes the Root Mean Square Ramp (RMSR) as a superior metric. Unlike a simple count of ramp events, RMSR accounts for the severity (size) of the ramps, providing a clearer picture of variability risk.