How I flooded Ambridge

What's "Ambridge"? I can hear my foreign readers ask. Well, Ambridge is the setting for the BBC's long-running radio soap opera The Archers.

One day (2002-05-29 to be precise), Ian Jelf - who is a professional "blue badge" guide - wrote on the newsgroup uk.railway:

Ambridge was in fact originally based on Hanbury, near Bromsgrove. This was the home of the family farm of Godfrey Baseley, the head of Agricultural programming for the Midland Region who devised the Archers in 1950.

and I followed up with:


This referred to an event from my youth which was first published in, I think, 1997. So here, for your reading pleasure, is the tale of how I flooded Ambridge.

[Warning: some of the exact numbers in this story are taken from memory and may be wrong. However, all are in the right range and relative values, when important, are correct.]

When I was young, I had a summer job at East Worcestershire Waterworks Company. They had sponsored some research in optimization and control theory, and I had been employed to convert the research systems to practicality (at £40 per week plus bed and breakfast).

The relevant parts of the water system are:

  1. Sugarbrook boreholes. These are located close to the Queen's Head pub, on the canal just south of Bromsgrove. There are four deep holes drilled down to a layer of permeable rock, and water is pumped up several hundred metres, through purifiers, and into:
  2. Sugarbrook reservoir, a sealed underground reservoir that can hold about 10 million litres of water. As well as being gravity fed to the Droitwich area, water then flows to the:
  3. Sugarbrook High Lift pumps. This is a bank of variable speed electric pumps with a total capacity of 13.5 Ml/day. They draw water from the reservoir and pump it down the:
  4. Sugarbrook-Evesham main. This is a pre-stressed concrete pipe. It is about 1 metre in internal diameter, with walls 50 cm thick, and is buried a couple of metres underground. At the far end is:
  5. Evesham reservoir, another sealed underground reservoir with a capacity of 6.5 million litres. From here water is gravity fed to the Evesham area. This is the only water supply for that area, and an average day's consumption is 6 Ml.

Historically, Sugarbrook High Lift pumps were left running at little under half speed all day, so that they would deliver the day's water to Evesham in an even flow - because demand varies throughout the day, the reservoir level would go up and down within fairly narrow bounds. This had been the situation for many years.

The software I was working on had three major components.

  1. Analyse real-time logs of water flows and reservoir levels to determine the actual demand over the last 24 hours on a half-hourly basis.
  2. Use a complex Fast Fourier Transform combined with a 48-dimensional quadratic predictor with alpha smoothing to predict demand over the next 24 hours.
  3. Schedule pumping of water so as to minimize the total cost of electricity to the company in supplying that water, noting:

Note that the software didn't control anything directly. It simply generated a list of instructions for the TeleControl Room staff to carry out (one man could control every pump and monitor everything in the network via a UHF radio system).

[For those interested, the software was written in Fortran 4 on a PDP-11/40. This had 64 kb of RAM and 10 Mb of hard disc. It was booted from paper tape.]

After fine-tuning the software, it was decided that we would go live at 0001 on 1980-08-01. So the machine was booted and the daily analysis run, and the instructions were printed out. They contained a feature that we'd noticed before but thought nothing of - after all, it made sense.

At the time, the Midland Electricity Board's summer high voltage tariff was dominated by one thing: there were different rates for energy consumed in the day and at night. Just like domestic "white meters", except that the day rate was about 5 times the night rate. So the software had decided that the best way to provide Evesham with its daily 6 million litres was to wait until 1 am, when the tariff changed, and then run the High Lift pumps at full power until Evesham reservoir was at the maximum permitted level; the pumps would then be turned down to keep the reservoir at that level until the full 6 Ml had been transferred, then turned off. Even with the reduced efficiency of the pumps, it should be possible to reduce the cost of that water by 75%. [This was an extreme case; the overall savings from computer prediction and control turned out to be about 12%. Still not to be sniffed at.]

So at 1 am the duty controller hit the "start" buttons for all 6 High Lift pumps and, seconds later, selected full speed. As near as we could judge afterwards, a pressure wave ran down the 26 km of pipe, bounced off the far end, and hit a following wave. The resulting surge literally blasted a piece of concrete pipe out of the ground, followed by a fountain of water. As the Water Research Council later said: "we didn't know you could *do* that to concrete mains". And for the next two hours, until the controller became alarmed at the discrepencies in the level of Evesham reservoir, the HL pumps proceeded to dump about a million litres of water over the Worcestershire countryside.

[After that eventful start and a change in starting procedures for Sugarbrook High Lift pumps, the system settled down and ran happily for many years.]

BackBack to the miscellanea index. CDWFBack to Clive's home page.