Solar photovoltaic (PV) power output variability caused by intermittent clouds is a major barrier to the expansion of solar power. Regulations have been introduced to restrict the maximum power ramp rates for photovoltaic (PV) plants. These restrictions typically invite one of two approaches: (1) Compensate the power variability through energy storage systems. (2) Curtail the PV output to smooth up-ramps reactively and provide a buffer for smoothing down-ramps proactively. Knowing the worst-case expected PV production ramp rate (WCS-RR) would make it possible to determine the battery reserve required to offset power fluctuations.
For a cloud passage, the amplitude of power ramp depends on the cloud optical depth, while the ramp rate is also a function of how fast the moving cloud passes over the plant. We propose a novel approach to estimate WCS-RR by a model consisting of point irradiance measurement, the geometrical layout of the PV plant, and cloud velocity.
The database covers 10 months of 2 sec resolution power output measurement from a roof-top PV system located on a building at the campus of the University of California, San Diego with a total capacity of 30 kW. Cloud speed is obtained from our in-house cloud speed sensor. Results are calculated for 86 cloudy days.
The WCS-RR estimate is validated against the observed 2 s ramp rate for June 21, 2018 [Figure upload timed out; so we were not apply to provide a figure.]. The WCS-RR generally widens with time due to the gradually increasing clear sky power and cloud speed. Clouds are observed to move with a varying speed between 2 to 5 m s-1. On this day, kt does not have a strong impact on the WCS-RR, because the cloud optical depth is constant over the 5 hour period. The largest observed ramp rate of the day occurred on 21:28 UTC, when a short clear period was followed by homogenous cumulus cloud layer. Because this change in cloud condition exactly matches the assumption in the ramp estimate model, the WCS-RR estimate captures the magnitude with minimal overestimation. Overall, the observed 2 s ramp rates are perfectly enveloped by the WCS-RR at all times, demonstrating that the proposed model of the maximum expected ramp rate works as designed.
The proposed method is also evaluated by the daily compliance rate which was 96% for a 1 min evaluation window over the 86 day dataset. While performance varies with the evaluation window length, our sensitivity analysis demonstrates that the proposed method successfully envelopes the real ramp rate most of the time.