Dear Colleagues

Windpower is shaping up as the main source of renewable power and the key competitor to fossil fuels; although apparently 85% of wind projects overestimated their production. As we all know with the uncertainty of wind, there are problems with getting the power at the right time. And storage of power generated when the wind is really pumping is still a tricky area. So this is where an entirely new field of statistics is developing – this is a vital part of generating wind power that you need to understand as an engineering professional – no matter whether you are  a tradesman or engineer – as wind power is going to be a key technology in the short term future. Wind is horribly variable - unlike the ‘good old’ coal powered power stations where you simply had to get the supply of fossil fuels right; shovel it into the furnaces, generate the power and you were home and hosed.

The European Union has committed that 20% of the energy generated by 2020 will be renewable and mainly through wind power. America has a similar target. Unhappily capacity in windpower does not actually mean the same thing as delivery of electricity. And this is where there is growing activity in wind forecasting. An entirely new engineering industry seeking to provide tight predictions.

Statistics forms a key part in doing the necessary forecasting. This is to achieve two objectives: long term statistics to get the necessary financing to build your wind farm and then once you are operating, you need to be able to forecast short term specifics of provision of power.

Bankers inevitably are not going to loan you money for your windpower project unless there are solid estimates of the wind farm’s capacity. One of the key techniques used is ‘measure, correlate and predict’ analyses. This involves measuring the wind at your proposed site for a year or so (two years would be great); correlating it with historical wind data from a nearby weather station; and then building a statistical model of the potential wind resources. Bear in mind that output from a windfarm can vary by 20% from year to year.

As you well know - wind is extraordinarily sensitive to the shape of the landscape; so there may no correlation between the wind at the reference site (the weather station which wasn’t designed to measure the vagaries of wind speed) and your proposed wind farm site. As wind turbines are much taller than measurement towers, you will often find significant differences in the strength of the wind at different altitudes. Once you have estimates of wind, you apply it to the power curve of the wind turbine you select. Power output is proportional to the cube of wind speed; so small fluctuations can result in huge changes to the energy output. You also need to estimate wake losses, which are the losses associated with an upwind turbine reducing the wind available to the down wind turbine.

Banks will lend money based on the so-called ‘P90’ wind value – this is the average wind speed in which they can be 90% confident. The closer the P90 reading is to the measured average speed (no greater than a 15% variation), the more attractive the site is to investors.

Once your wind farm is up and running, intermittency of wind is a serious issue. For example, in Denmark (20% of its electricity now sourced from wind); a change in the wind speed of 1m/s translates into a huge 450MW in national power output. The overall trick (as it were) is to make wind farms appear as close as possible to power stations in their provision of reliable power at particular times and this is where short term statistical forecasting is key.

Simple forecasting is called ‘persistence forecasting’; which assumes that the wind speed in an hour’s time is the same as now. This is the benchmark approach; but obviously not as accurate as some of the more sophisticated approaches using numerical weather prediction which model the atmosphere as a three dimensional grid, with cells of few kms on each side and grabbing physical data such as pressure, temperature and humidity from sensors. Accurate short term predictions are critical to wind farm operators – the difference between large profits and huge fines for non-compliance.

So next time; you look askance at statistics; realize that they are increasingly a key component in wind power engineering these days. But from an engineering professional point of view, it is certainly worth getting more familiar with how statistics can help you in your day-to-day work.

To quote from a former British prime minister, Benjamin Disraeli:
There are three kinds of lies: lies, damned lies, and statistics.

Hopefully, you use statistics in such a way, that this statement is not true.

Thanks to the Economist for an intriguing article on the subject and for some interesting stats.

Yours in engineering learning