Unfortunately, the forecasts proved pretty inaccurate. As the sun rose and demand climbed to its daily peak, it was clear that the day-ahead wind forecasts were significantly off. Bizarrely, the same-day forecasts were even worse. Wind output was high, but not as high as forecast.

The level of discrepancy between what was forecast and what was generated was extremely difficult for the control room to manage that day, and would typically trigger substantial re-dispatching (sending new orders to power stations to change their running plans).

As a result of the discrepancy, there were multiple changes of direction and contradictory flows in the undersea electricity cables that link the UK with its European neighbours. There were times when Britain was importing from France over the IFA2 undersea cable, while exporting over its sister link, IFA1. And times when Britain imported from Scandinavia over the North Sea Link (from Norway) while exporting over the Viking Link (to Denmark).

• Robert Colvile: Ed Miliband’s wind plan has left him hostage to suppliers

The National Energy System Operator (Neso) made multiple interconnector trades up to 2pm, all of which were exports, and all at negative prices, meaning Britain was paying other countries to take away its excess power.

The control room staff were left scrambling to deal with the impact of all these changes. Gas power stations, such as Marchwood near Southampton and Grain in Kent, received multiple up and down instructions that coincided with shifts in wind output and interconnector flows.

In the Midlands and the North of England, gas stations such as Staythorpe, West Burton and Keadby were largely asked to power down — a move that is consistent with high local generation from wind and solar.

In Wales and the North West, the Deeside and Rocksavage gas stations were bid up more consistently.

The overall pattern was a day of complex system balancing. Poor wind forecasts against export commitments left the system short at times, while the uneven geographic distribution of windfarms, together with grid constraints, made it hard to manage grid stability. High wind output, despite being below expectations, created local surpluses.

By 2pm on May 29, the control room had undertaken almost 6,500 balancing actions (15 per minute over the whole shift). The rest of the day was no less hectic, with the actions running at more than 17 per minute).

How the grid is supposed to work

The basis of our power grid is alternating current, generated from large turbines in gas and nuclear power stations to give current and voltage at 50 hertz (50 cycles per second).

The energy transition has ushered in a major deployment of intermittent renewable generation. In general, wind and solar energy is converted into direct current — meaning that Britain has more direct current generation trying to work on an alternating current grid.

Wind and solar also lack an important property known as “inertia”. Conventional generators are big, heavy machines that resist changes to their speed of rotation. They act as a brake, slowing changes in grid frequency.

That wind and solar do not possess inertia means that as we replace conventional generation with renewables, we reduce the levels of inertia on the grid, so making it less resilient to the effects of faults (which is basically why the as-yet-unidentified grid fault in Spain led to a system-wide blackout in April).

Grid operators also have to “balance” the grid –— make sure that supply and demand are equal at all locations on the grid, at all times, to keep the frequency at 50 hertz everywhere. They do this by re-dispatching new orders to power stations. It was this “balancing mechanism” that was tested almost to destruction on May 29.

What can we learn from this day of havoc?

May 29 is an example of what happens when you connect weather-based direct current renewables, in a haphazard way, to an alternating current grid. Under the direction of industry regulator Ofgem, the priority has been to connect windfarms and worry about grid bottlenecks later.

This means we often have to pay windfarms not to generate, because their electricity cannot be transmitted to consumers. But it also makes the grid difficult to manage because weather is notoriously hard to forecast at the best of times, and there has been under-investment not just in grid infrastructure, but the tools needed to manage it.

The software that underpins the balancing mechanism was written in the 1980s and is subject to frequent outages; had there been such an outage on May 29, the control room staff would have had to revert to manually instructing plants by phone, which is hard to do at a rate of 17 per minute. The grid could easily have collapsed, resulting in blackouts.

The operator’s likely response would be to revert to a more traditional grid — turning off the wind power and turning up the gas, and hoping this can be done quickly enough to maintain stability.

The control room staff do a heroic job with inadequate tools, but the heroism of individuals is no way to run a secure modern power system.

Kathryn Porter is an energy adviser and founder of the Watt-Logic consultancy