Projected Assessment of System Adequacy

Introduction

YouDo is a Trans Tasman software engineering firm with over 13 years’ experience in creating powerful solutions for ambitious players in the energy industry. With our work encompassing projects across generation, distribution, wholesale trading, DER and retail, we’re plugged into the energy industry, enabling us to deliver solutions backed by a deep understanding of the market.

Applying this knowledge, YouDo undertook the challenge of modelling somewhat opaque system availability data in Australia, in an attempt to pinpoint exactly which energy generation plant has a planned outage within an Australian state. The purpose of which, is to assist energy traders as they bid on future products in the energy market.

Problem Description

As a retailer participating in the National Electricity Market (NEM), there are a number of risks that you are exposed to which may directly affect the price you pay for your customers electricity.  

To protect against this, retailers take up a hedged position, purchasing futures contracts (e.g. CFDs) to protect against price spikes, and Settlements Residue Auctions (SRAs) to protect against locational risk, should their customer portfolio be spread across multiple jurisdictions.

When trading SRAs, there are a number of variables which affect the potential return from purchasing an SRA at auction.  One such variable is the availability of a plant located near an interconnector.

Such information is available in the Medium Term Projected Assessment of System Adequacy (MT PASA) schedule, which forecasts potential outages, two years in advance of the published date.  

This data, while giving an indication of the potential outages, is anonymised and aggregated up to a State level, meaning it is difficult to pinpoint exactly which plant has the planned outage.

Using the publicly available date from the AEMO data hub, YouDo attempted to develop a model which predicted with a high degree of certainty, the planned outages listed in the MT PASA Schedule.  

YouDo Study

YouDo developed an analytical tool that is able to retrieve and pull PASA datasets, alongside bid trading and dispatch data from NEMWEB, and use this data to build a predictive model that is able to predict scheduled generator outages and return to service times. 

Our initial approach to this project was to encode the data into a box packing model, and use mixed integer programming to come up with a solution. Unfortunately the PASA data proved to be variable and erratic enough to cause the solution space to grow so large that the chance of the optimal solution being selected by the model being the correct one became unreasonably small.

To be able to visualise the problems behind the obfuscated data better we built a user interface (UI) that allows the a user to visualise the evolution of PASA data over time, and to select and analyze a specific window of PASA data by displaying the actual generator availability breakdown VS the availability breakdown generated by a rudimentary probabilistic model.

Integrations

As part of this project, YouDo integrated directly with AEMO’s NEMWEB, pulling and analysing data per 5 minute trading-period.

In this project, YouDo pulled the following datasets;

• Bids
• PASA
• Trading
• Actual Generation
• Dispatch data

Conclusion

YouDo found that the data published in the PASA datasets was highly variable and more noisy than anticipated, making it difficult to conclude with certainty the availability of the underlying stations, or attribute significant and non-obvious changes in PASA to a single responsible station.

Taking a quote from an article published on WattClarity;

“There are those who believe they have some proficiency in ‘reading the tea-leaves in MT PASA data’ to the point where it is a perceived competitive advantage.”

We think a more specialized approach which combines profiling of individual stations combined with enhancements to the model to incorporate additional inputs (e.g. weather data) would yield better results.

Project Update

On the 20th of February, 2020 a new AEMC rule dictated a range of changes to the MT PASA to improve the transparency and accuracy of information about potential electricity shortfalls. Key features include requiring AEMO to:

• Publish information about planned maintenance for individual generation units (currently only aggregated information about generator availability is published)
• Provide greater clarity on how future generation is included in the forecast, as well as how generation availability impacts forecasts under a range of scenarios
• Use the same format when publishing forecast and actual demand so the data can be easily compared
• Require the data that participants provide into the MT PASA to be based on participants’ ‘current intentions and best estimates’.

This ruling of course strikes at the heart of the problem described in this case study and alleviates any value from progressing further.  

Even so the study proved a useful exercise in interrogating and analysing publicly available AEMO data to draw useful insights.

Demo?

Interested in seeing what we built?  Contact us for a demo and to discuss how YouDo can assist your trading teams outperform the competition.