Working with wind

The increased availability of wind power is beginning to cause utilities problems as engineers attempt to synchronise the power grid.

Nobody can question that wind power generation has expanded rapidly in the past two decades. In Denmark 20 per cent of electricity is generated by wind power, and in Germany, Spain and Ireland that figure

is approaching 7 per cent. It is likely that Ireland will be the first of these countries to reach 10 per cent, probably by 2010. The problem there is that the Irish grid does not have the benefit of strong interconnection with neighbouring grids, as is the case with Denmark. If wind power is to push past the 10 per cent level in Ireland, new system operational strategies will be needed. Other, larger systems will need similar approaches as alternative energy generation expands beyond the 10 per cent level, including the UK grid. The problem is largely economic. When wind energy becomes available, the system operator will be tempted to de-commit large, fossil-fuel fired (thermal) generators - mainly gas. The potential difficulty is that an unexpected wind power lull would require generation to be synchronised quickly. This may not be possible with thermal generators, which have minimum down times of several hours. The only option then is to run up quick-start generators, such as open-cycle gas turbines or diesel generators. These units have low thermal efficiencies and are therefore expensive to run. However, it has been found that the long-run cost is less than keeping large units online.

Effect of wind power

Utilities have tended to ignore the operational aspects of wind power until there are commercial pressures to take positive action. The nature of these pressures may be understood by considering the classic problem of unit commitment. Suppose we have a system with ten similar 200MW generating units supplying a demand of 1000MW. Each unit is assumed to have a running cost which is related to unit output (see graph below). The slope of the graph is the incremental cost a1, taken here to be £50/MWh. Note that there is a 'no-load' cost, which is assumed to be £1000/h, or just under 10 per cent of the full-load cost, as is typical in practice. The no-load cost ao covers the cost of the fuel needed to supply the heat losses from the boiler, turbine, alternator and transformer. The system operator must decide which generators to run, ranging in this example from five units running at 100 per cent of output to ten running at 50 per cent of output. The corresponding operating cost will range from £55/MWh for five units operating at 200MW each, to £60/MWh for ten units operating at 100MW each. The example highlights a conflict in system operation between economy, achieved with as few units as possible, and security, achieved with many units running at part load. A practical solution in this example might be to employ six units, each operating at 167MW. The energy cost would then be £56/MWh. Bringing an extra unit on line adds £1/MWh to the operating cost, but also provides 200MW of operating reserve. This margin is needed to cater for unexpected changes in the system demand, and to provide emergency reserve should one of the six units fail without warning. Suppose that the system contains a number of wind farms, with a total capacity of 300MW. If the wind generation is 200MW at the time of interest, it would be possible to de-commit one of the six thermal generating units and run the remaining five at 160MW each. The energy cost would then be £45/MWh. However, wind power is variable and may decrease, but is very unlikely to decrease quickly, say within an hour. Should this happen, it may take several hours to bring a thermal unit back online, and the only solution would be to run up quick-start generation or disconnect consumers. System operators are averse to such action, and will tend to keep extra thermal generation connected in case of such a wind power lull. The cost of so doing, with six thermal units running at 133.33MW each, is the extra no-load cost, £1,000/h. The resulting energy cost will be £46/MWh, an extra £1/MWh.

Suppose the demand remains unchanged at 1,000MW for the next 12 hours and that the wind power is expected to remain at 200MW for the duration of this period. In the event, it decreases to 150MW in the first hour, 100MW in the second hour, and recovers to 150MW in the third hour, before reverting to 200MW for the remaining nine hours. We will examine the economics of replacing part-loaded thermal generation with quick-start generation for this scenario. The quick-start generation is assumed to consist of a portfolio of diesel generators with a running cost of £100/MWh. This compares with an incremental cost of £50/MWh for the thermal generation. The 2:1 ratio is fairly typical, and varies little with fuel price variations. Applying a conservative loading strategy, without use of diesel generation, six thermal units are used at a running cost of £562,000 over the 12 hours. Applying the alternative strategy, with diesel generation compensating for wind power shortfalls, we need only five thermal units. The cost of running the five thermal units at a constant 160MW is £540,000. To this must be added the cost of 200MWh of diesel generation to supply the wind power shortfall, at £100/MWh, giving a total cost of £560,000. The saving is £2,000. In effect, the £10,000 spent on diesel generation has enabled the system operator to dispense with a thermal generator, whose no-load cost over 12 hours would have been £12,000. Clearly, the success of the 'quick-start' approach depends on the quality of the wind power forecast. We will, however, demonstrate the feasibility of this approach with the use of a wind power forecast produced from publicly available low resolution wind speed forecast data.

Wind power forecasting

Public domain wind speed forecast data were recorded several times daily over a period of six months for Northern Ireland. These forecasts are originally produced by the Meteorological Office using a numerical weather prediction (NWP) model. The forecasts are available in three-hour segments with a maximum prediction horizon of 24 hours. Although the forecasts are updated during the day, this study uses only the 24-hour forecast available at midnight each day to help optimise the commitment of thermal plant for the forthcoming day. The recorded forecasts and subsequently the actual wind power generated over the six months were then used to produce a simple relationship between the forecast wind speed and actual wind power generation. The power output from ten wind farms with a total 120MW installed wind capacity were used throughout the study. The graph below shows the accuracy of the resultant wind power forecast based on the NWP data alongside the more basic persistence approach. The persistence method of wind power forecasting assumes a prediction equal to the last known measured value of wind generation. The prediction is updated only when fresh online data becomes available. The persistence forecast is more accurate for a time horizon of less than six hours, thereafter the NWP based forecast error has a standard deviation no greater than 0.2 per unit of wind power capacity. Commercial forecasting systems will, of course, surpass this accuracy due to finer time and spatial resolutions, but will never be error free. Therefore, to account for forecast errors, thermal plant must increase or decrease generation accordingly. Increasing thermal generation during an over-forecast will erode system reserve which, beyond a critical threshold, will compromise security margins. As described earlier, system operators are averse to let such a situation occur, and will opt to keep more thermal plant committed. This default approach is often referred to as the 'fuel-saver' mode. In fact, it leads to inefficient use of fuel, as we shall see.

System operation

The forecast was used to determine the commitment and dispatch of thermal plant in the Northern Ireland power system with the aid of fast-start diesel generation as described earlier. The use of a portfolio of diesel generators permits greater control over balancing capacity committed compared with the use of a few larger open-cycle gas turbines. Diesel generators can also effectively challenge open-cycle gas turbines on a purely economic basis. The table on the facing page gives the parameters of the thermal plant included in the simulation, where MUT, MDT, CSC & τ are minimum unit up-time, minimum unit down-time, unit cold-start cost and unit cooling time constant respectively. The scheduling and dispatch of the system is simulated over a nine-day period in October 2006 with the objective of minimising cost, subject to the various unit constraints. The technique of Lagrangian relaxation is an effective tool to solve this optimisation problem (see panel on facing page). The system load, wind generation and forecast for the period are shown in the diagram (right). This wind profile has a typical capacity factor of 0.38 and supplies 4.5 per cent of load in the period. As stated earlier, £100/MWh is assumed for the cost of diesel generation. The system reserve is shared with the southern utility, with the northern system supplying 100MW of reserve at all times. The simulations compare use of the forecast mode facilitated by the fast-start diesel generation with use of the default 'fuel-saver' mode. The cost of generation is £33.88/MWh in the default fuel-saver mode. However, with the use of the fast-start diesel generation and the forecast described the cost is cut to £33.17/MWh. The hypothetical case, where a perfect forecast is available, results in a cost of £32.42/MWh. Therefore, the use of the simple forecast and diesel generation reduces the margin by nearly 50 per cent. Use of more advanced forecasting techniques would permit closer convergence on this optimal value. The peak capacity of diesel generation needed to balance forecast errors in this case is 18MW. This capacity and beyond could easily be contracted from owners of diesel plant in the commercial and industrial sector of Northern Ireland, with modern communications permitting a start-up response time of only a few seconds. In the forecast mode the load factor of the large thermal plant is increased by an average of 4.4 per cent, as a smaller number of plant supplies the same load more efficiently. Therefore the pairing between diesel generation and wind forecasts permits large thermal plant to be de-committed with confidence if the opportunity arises. The results show the occasional expense of diesel generation is outweighed by the savings achieved through dispatching a smaller number of thermal plant in a more efficient manner. In the forecast mode the carbon emissions are reduced by 1.72 per cent compared with the fuel-saver mode and the cycling of large thermal plant is reduced. This will have a beneficial effect on other serious greenhouse gases such as NOx and also prolong the life of the plant. However, to curb carbon emissions an expensive carbon tax would need to be introduced to challenge cheap coal-fired generation. Diesel generation and OCGT peaking plant also possess a large carbon footprint compared with modern gas plant. In this case, the emissions from the proposed diesel generation can be reduced by blending conventional fossil diesel with biodiesel, an alternative carbon benign fuel. Biodiesel fuel, used in this application, is a convenient form of renewable energy storage. The small quantities of fuel required for fast-start balancing generation are well within the reach of the current UK supply chain. The added expense of biodiesel generation could be partially off-set through qualification for renewable obligation certificates. Fast-start diesel generation, with emissions on a par with larger conventional plant, would then be a suitable form of complementary generation for wind.


In the absence of a perfect wind generation forecast the flexibility of diesel generation appears an affordable method to allow wind importation at low cost while maintaining security standards. The infrastructure for such a service is already largely in place. Starting diesel plant at least once a month for reserve would also improve the availability of the units and ensure they work. Black start facilities could also be offered through this technology. A natural extension would be coordination of diesel generation and controllable thermal demand to provide two forms of reserve service, one in generation and the other in demand. In this scenario the markets would dictate which to use. Thermal loads, such as immersion heaters, could also be effective in storing excess wind energy in the event of an under-forecast. Therefore, both diesel generation and dynamic load would help mitigate the negative impacts of wind generation on large supercritical plant. Future improvements in wind generation forecasts will certainly help facilitate efficient wind power importation. However, without flexible plant like diesel generation in the mix, can economic and secure operation of the power system really be guaranteed?

Recent articles

Info Message

Our sites use cookies to support some functionality, and to collect anonymous user data.

Learn more about IET cookies and how to control them