Semi-scheduled Generation Availability Forecasts–how to improve?


The most recent summer raised questions on the performance at high temperatures of intermittent (wind and solar) generators and how accurate the forecasts were. Of broader interest than summer is how intermittent generators and AEMO work together to forecast the availability of the generator and how this can be improved over time, a topic that features strongly at AEMO’s occasional Intermittent Generator Forum meeting. AEMO had to cancel the meeting scheduled for late March 2020, but has circulated the slides for industry to comment. This article discusses the forecast calculation mechanisms and outlines key factors in improving the quality of the forecasts.

This is my first article for WattClarity. I’ve recently joined Global-Roam as a software engineer and market analyst, primarily to further develop the ez2view market visualisation tool and support ez2view customers, using my knowledge of how generators operate in the market. My early career was in electronics and software development, moving into energy through the Master of Energy Systems at the University of Melbourne. I spent two years at AEMO in operational forecasting (hence my strong interest in today’s topic) and the past three years at renewables developer and operator Tilt Renewables. I’m looking forward to working with new and existing ez2view customers to help you understand the market better.

Why forecast?

Accurate forecasts of semi-scheduled generation availability are essential in maintaining power system security, including in dispatch to reduce system frequency variation and in pre-dispatch and ST-PASA (out to a week) timeframes to enable accurate reserve calculations to be made. For intermittent generators there are two parts to the equation – the generator’s obligations to provide information to AEMO, and AEMO’s role in calculating the availability in the generation offer using the AWEFS and ASEFS systems.

Terminology check

  • AWEFS/ASEFS: the Australian Wind Energy Forecasting System and Australian Solar Energy Forecasting System – AEMO-managed central forecasting tools that calculate availability using detailed weather forecasts, statistical models, generator availability submissions (via AEMO EMMS portal) and SCADA (real-time measurements), along with generator information provided at commissioning through the Energy Conversion Model (ECM).
  •  “Availability”: in the generator’s energy offer – the maximum capacity the generator has to offer at that time, incorporating plant and energy limitations. Note that a scheduled generator provides this directly in its MAXAVAIL entry in its offer. For semi-scheduled the MAXAVAIL input in the offer is ignored as AWEFS or ASEFS provides the availability. In the dispatch (5-minutes ahead) timeframe only, some semi-scheduled generators provide their own availability forecast in place of the AWEFS/ASEFS, known as “Participant Forecasting”. This article isn’t going to discuss Participant Forecasting any further but more details are available on AEMO’s website.
  • “energy” and “plant” availability: “availability” as a term is qualified by: “energy” – what’s available right now given how windy or sunny it is (the energy input), and “plant” – assuming no energy input limitations, the maximum power the plant could produce. plant availability is an NER defined term. NER 3.7B(b) obliges participants to notify AEMO of plant availability.

Improving the process

There are two mechanisms that a semi-scheduled generator uses to communicate information on its availability to AEMO for input to AWEFS or ASEFS. In real-time there are SCADA readings, including the Turbines/Inverters Available signal, plus the Local Limit signal which indicates a plant-wide limitation on output. Generators also provide a range of real-time meteorological measurements including wind speed, wind direction and temperature. For the pre-dispatch and ST-PASA timeframes (out to a week) there is the AEMO EMMS portal, where generators enter the number of Turbines or Inverters that are unavailable for each half-hour, plus an “Upper MW Limit” which is a plant-wide limitation on output.

There are at least four factors that determine how effective this forecasting process is:

  • technically how accurately the forecasts can be calculated;
  • how well determined and understood the forecasting roles and information flows are between the different parts of the system;
  • how easy it is for participants to provide the right information; and
  • the capacity for feedback and continuous improvement.

Technical difficulty

In normal conditions the technical forecasting is a straightforward enough modelling exercise, particularly for the participant in forecasting the timing of planned outages of turbines or inverters (affecting their turbines/inverters available entry in AEMO’s EMMS portal), and of balance-of-plant such as transformers or reactive plant or other local limitations on the whole plant (affecting their Upper MW Limit entry). During unplanned outages the participant can estimate and update a return time for the components as the repair progresses. In more difficult conditions, most notably in high temperatures, forecasting the reduction in plant availability is more complex than knowing the maintenance schedule or likely return to service. The response of the plant is complex and unique to each turbine or panel or inverter type, and may include de-rating (where the maximum output of an element is limited by temperature), individual elements cutting out (stopping production) for some period, and a reduction in efficiency as temperature increases. It’s a much more difficult task than entering component outages to make a good guess at what might happen during high temperatures across an afternoon, and some of these factors (such as de-rating and a reduction in efficiency) don’t immediately correspond to the “Turbines/Inverters Available” or “Upper MW Limit” columns in the AEMO EMMS portal.

Share of responsibilities

Which brings us to the next question – who’s responsible for what part of the forecasting, and how does the information flow between the participant and AEMO?  A specific question is what, if anything, the AWEFS and ASEFS systems take into account in the forecasts during high temperatures*. The Energy Conversion Models (ECMs) submitted by participants at commissioning include standing data describing the response of turbines, inverters and panels to high temperatures, including the temperature for cut-out and an efficiency loss for PV, which may suggest to participants that AEMO is already taking all of these into account and the participant does not need to. However, recent communication from AEMO and also in the slides for the March 2020 Intermittent Generator Forum, indicates that AEMO does expect participants to take an active role during high temperatures. AEMO’s documentation on AWEFS and ASEFS is across a number of documents in here and here. These are somewhat contradictory particularly on the purpose of the Upper MW Limit – with the “Guide to Data Requirements for AWEFS and ASEFS” stating it is related to planned and unplanned maintenance that puts a MW limit in the control system, but the “Guide to Intermittent Generation” refers to the more general plant availability. The “Guide to Intermittent Generation” specifies what the Elements (i.e. turbines / inverters) Unavailable field should be, describing outages but not forecast effects of high temperature as reasons to mark an element as unavailable. While the participant is obliged to make their own interpretation of NER 3.7B(b) and plant availability, further clarity from AEMO would be valuable here.

*A related question is what AWEFS and ASEFS forecasts for the specific situation of high-wind cut-out, where turbines shut down in extreme conditions. There’s a separate (optional) SCADA signal for real-time notification of this. Further clarity from AEMO on what’s expected from participants particularly for the PD / ST-PASA forecasts would also be useful here.

Possible confusion

Given the apparently contradictory information on AEMO’s website, and the significant difference in effort (and therefore cost) for participants between the tasks of informing AEMO of maintenance outages compared to updating forecasts of the effects of high temperatures on their plant, it appears there is an information gap that AEMO would do well to close, with the goal of achieving improvements to semi-scheduled availability forecasting and hence system security. As a first step, updating the documents on the AEMO website to be clear on what AWEFS and ASEFS do and don’t take into account in their forecasts, with AEMO’s corresponding expectations of participants, would be of value to new and existing participants. AEMO could also work with industry to consider whether putting more capability into AWEFS and ASEFS would give a better outcome, as these systems have existing high-quality weather forecasts (including wind, solar and temperature) and modelling capabilities. Including a case study (real or fictitious) in the documentation of a wind and a solar participant’s processes for providing the availability information to AEMO would provide significant clarity, would assist in providing direction and justification for participants to update or implement fit-for-purpose systems and procedures*, as well as giving a starting point for a more detailed discussion between participants and AEMO on the process and what expectations are reasonable.

*Some participants may find support from a third-party control room provider such as Overwatch Energy to be helpful here.

Making it easier

The quality and timeliness of information that a participant can provide depends on how easy it is for them to provide it.

In the Intermittent Generation Forum slide pack AEMO asks if the EMMS portal where participants enter their availability information should have its useability improved. While there are many participants who use third-party tools to update the availability information via FTP, input via the EMMS Portal needs to be easier than it currently is to enable other participants to update their availability information in a timely manner. The current interface (in contrast to the bids and offers interface) is tedious to use especially given the slow response of the portal when a new date is selected, slow navigation (requiring mouse clicks) around the interface, and the lack of a fill-down feature. The bids and offers interface is much faster to use and can be successfully used to rapidly enter bids, and would be a good starting point for a re-design of the availability interface.

A further improvement that would give flexibility to the participant in providing availability information would be to enable optional submission of a limit on availability through the MAXAVAIL column of the energy offer, to be used as a cap on the AWEFS/ASEFS forecast. This would be particularly useful in the dispatch timeframe where the Local Limit SCADA signal is the only option currently to cap the dispatch availability. As the SCADA signals are generally calculated in vendor-specific industrial plant controllers, there may be limitations on the complexity of calculation that can be implemented and how readily it can be changed, and overriding in unusual circumstances may require tedious manual intervention in the SCADA system interface.

Transparency

To get the best results across the industry there needs to be transparency on the quality of information provided by participants. AEMO made a first improvement in the November 2018* MMS release 4.28 that added private tables with the availability submissions, so participants no longer had to download these from the portal to audit.  AEMO asks in the Intermittent Generation Forum slide pack if these availability entries should be made public next-day. Publishing the availability entries next-day would greatly improve the transparency of the availability information and the industry’s ability to evaluate and improve the process, and would lend itself to visualisation and comparison across generators in tools such as ez2view, and to be studied in reports such as the GRC2018 and GSD2019. To this end, the relevant SCADA data (Turbines/Inverters Available, Local Limit) would be essential to complete the set along with the MW generation as the “actual” of the availability forecasts.

*A note for AEMO – please list more than one of the recent data model upgrade reports on the website as participants are often one or more behind – this report was found with Google.

About the Author

Marcelle Gannon
Marcelle is a software engineer and market analyst keen on the transition to renewable energy. She has energy sector experience at AEMO and a renewables developer, joining Global-Roam in 2020.

3 Comments on "Semi-scheduled Generation Availability Forecasts–how to improve?"

  1. Welcome aboard Marcelle. Fantastic article. Easy to read, supported with many relevant links and recommendations. Learnt a lot. Many thanks.

  2. Terrific article Marcelle, contains a wealth of information about a set of processes that are pretty opaque from the outside – and it turns out not all that well engineered on the inside. The recommendation that bid max availability for S/S plant be treated as a cap and not totally ignored seems like a very sensible and easily implemented first step in addressing some of the problems we saw over summer with the current processes.

    I wondered if you had any insights into the disconnect we saw in AWEFS forecasts feeding 5 minute and 30 minute predispatch / PDPASA processes on Friday 20 December – see http://www.wattclarity.com.au/articles/2019/12/out-of-the-blue-an-lor2/
    the dispatch-level (5 minute) forecasts generally tracked actual output levels (but with some strange exceptions), but the 30 minute forecasts were off in la-la land for much of the critical period. Any thoughts on how this happened?

    Allan

Leave a comment

Your email address will not be published.


*