An erroneous trading order created manually without verification caused “strangely robust” Belgian day-ahead electricity prices for delivery on 1 May, a study published by Belgian energy regulator CREG has revealed.
Several traders in centralwest Europe told ICIS that human error being behind the bid was no surprise, but questioned what they saw as a lack of safety checks which allowed the order to go through.
The error resulted in the market participant being long after the day-ahead results were calculated, and several actions were brought into place to partially recover the position after gate closure on Sunday 30 April.
CREG, which analysed the results in the Belgian bidding zone for the specific day, contacted the market participant to better understand the reasons behind the move.
The company confirmed it had issued a limit order, bidding for each hour on 1 May a volume of 3,000MWh/h at a maximum price of €36.00/MWh instead of 36MWh/h at a maximum €3,000/MWh.
Because of the type of order used, the large volume and the competitive price at which the order was bid, the flow-based algorithm had sufficient commercial flexibility to maintain the balance between commercial supply and demand. Accordingly, the order was accepted by the Paris-based power exchange EPEX SPOT during hours 1 and 19, according to CREG.
The size of the purchase meant it became the order during the day, hence setting the market clearing price. During hours 1 and 19, the clearing price was €36.00/MWh except for hour 16.
“I am not surprised that these things can happen, as there will always be room for human error, especially in a stressful environment like this,” one day-ahead trader at a trading house in northern Europe said.
“I do however find it a bit odd that they were able to upload the bid due to threshold rules. The company must have a pretty massive credit line at EPEX SPOT.
“I think for the future it would be great if EPEX incorporated some sort of ‘sanity check’ when companies upload their bids, in case something is out of the ordinary,” the trader added.
Another power trader at a large utility in Europe agreed: “The exchange should have checks and should contact the counterparty in cases where such unusual bids are sent.
“So I assume there were a number of things that went wrong.”
In normal circumstances, the actual placement of a company’s bids and offers is fully automated based on technical and economic input parameters collected through a portfolio optimisation process. But on Sunday 30 April, when the order was introduced, third party data required for the automated process to run was lacking and the order had to be processed manually.
CREG said the market participant, in the aftermath of the blunder, had since implemented validity rules which control whether the volume size or the order exceeds a certain threshold value, in a move to prevent a repeat occurrence. The CREG report is here in full.
The head of trading at a French energy company said that application programming interface (API) was imperative for trading in short-term markets because it limited the chances for mistakes.
APIs integrate internal trading systems, back office and operating tools to customise automated reporting.
But he added that third party data in Belgium – which appeared to be the source of the error on 30 April – was less transparent than in France, particularly related to real-time data.
“It’s quite tricky to have all the information you need,” the trading head said.
A spokeswoman at EPEX SPOT described such an event as “extremely rare” thanks to processes and procedures in place that accompany members throughout the trading process.
“Some checks can be conducted by the members themselves through parameter tools, and EPEX SPOT disposes of tools to monitor activity and trigger alerts according to configured parameters,” she said.
But with the power bourse’s number of orders constantly rising, processes becoming more and more automatised and more and more data needing to be processed, this did not eliminate the potential for mistakes, EPEX SPOT acknowledged.
“Errors – related to actions by a person or a machine – can theoretically occur,” she said. email@example.com