So why is bulk solar PV different? Should it be?
I think many of us who have been attending UVIG meetings over the past ten or fifteen years were surprised by the explosion of solar PV in the US and around the world. We shouldn’t have been, in my opinion, as we had front row seats to what happened with bulk wind generation over the decade prior. I must confess further that I did not foresee the rapid growth of bulk solar facilities – i.e. those connected directly to the transmission network. I guess that is why I do what I do.
Charlie Smith was a better prognosticator, and saw the need for UWIG (at that time) to shed the “W” and provide the same technical guidance for this new and fast growing renewable generation technology. A major concern to many was that solar PV might somehow miss out in taking advantage of the grid code learning curve that bulk wind generation had climbed over the previous several years. If you recall, UWIG was a key player in the technical discussions that lead to NERC’s clarification of low voltage ride through and other performance requirements for wind generation, as encapsulated in Order 661a. With this hurdle cleared, the vision of bulk wind plants behaving as well as (and in some aspects much better than) conventional plants with synchronous generators became reality.
It appears, however, the learning curve for bulk solar appears to be of a different shape and slope. Last fall, I and many others learned details of the infamous “Blue Cut” fire in California. In a report from NERC released last summer, the conditions leading to the sudden loss of solar PV generation were documented. Fires in an important SCE transmission corridor led to over a dozen 500 kV line faults, one of which resulted in nearly instantaneous tripping of 1200 MW of solar PV generation. A variety of factors were identified and documented, of which the two most important were:
- Perceived off-nominal system frequency, as transduced from a local bus voltage phase angle and misinterpretation of NERC Standard PRC-O24-2 (which allows for instantaneous generator trip at frequencies less than or equal to 57 Hz);
- Inverter “momentary cessation” due to system voltage reaching the low voltage ride-through setting of the inverters.
The first factor turned out to be relatively correctable, once understood – system frequency and frequency ascertained through cycle-by-cycle measurements of a local voltage phase angle are not the same thing. The measurement is essentially a mathematical derivative, which can be very unstable for very sudden changes in the input. Filtering or delays in acting on the quantity can be remedies. (An aside: I’m reminded of the emerging issue in the bulk renewable industry of ROCOF – rate of change of frequency. Locales with high renewable penetrations, like Australia and Ireland, are exploring standards related to ROCOF. This makes me nervous because the ROCOF measurement is effectively the derivative of a derivative!)
The second factor in the Blue Cut Fire event is actually of much more importance in my opinion. Cessation of current injection to the grid was never a consideration in the past discussions about wind plant grid code requirements. In fact, it was thought to be desirable, especially for weak interconnections, to inject maximum reactive current during the fault, so that maximum support would be available for the voltage at the point of interconnection when the fault was cleared.
As the wind industry in the US began to grow significantly, turbine vendors quickly realized that not sustaining operation through grid faults would be an eventual damper on their markets, and undertook design activities to meet the requirements, once clarified. In my opinion, their task was more difficult relative to PV, since the rotating inertia insured that mechanical energy continued to pour into the conversion system and had to be dealt with to prevent electrical failures.
Now it seems to me (disclaimer: I am not an inverter design expert) that ride through for PV is not as heavy of a lift. There is no mechanical inertia to deal with, and the maximum voltage on the dc side of the converter is capped by the array open circuit voltage, which the power electronics must be designed to accommodate. As the ability to deliver real power to the grid is hampered by low ac-side voltages, the dc-side voltage will rise and array current will fall. The limit here is where array current is zero and the array voltage is at the open circuit value. Effectively, the i-v characteristics of the array can lead to cessation of the current without any external control.
What I don’t understand is why the current cannot be restored to the pre-fault value immediately upon return of grid voltage, or why maximum reactive current cannot be injected during the fault. For very deep faults, there may be some very short delay until synchronization (via PLL or other means) can be reestablished, but this should be on the order of milliseconds or a few cycles. Unlike synchronous generators, the dynamics during this period can be well controlled and made benign to the system (unlike oscillatory behavior of synchronous machines following the restoration of voltage).
There was some discussion of this at the last Reliability Working Group meeting in Nashville last fall. Mahesh Morjaria of First Solar gave an excellent presentation on the Blue Cut event and what they have done in its wake. NERC also conducted some follow-up studies that seemed to indicate inverter cessation for up to 5 seconds was “probably ok”.
For what it’s worth, I’m not convinced that it is “probably ok”. There are situations (think Hawaii with current and future renewable penetrations where 5 seconds might as well be an eternity). We’ve likely not heard the last word on this issue.
I think we need to invite the solar inverter manufacturers to UVIG.
Robert Zavadil, Executive Vice President
EnerNex
Leave a Reply