Estimating the impacts of wind on power systems1 May 2008
What do existing studies tell us?
Existing targets anticipate quite high penetration of wind power in many countries. It is technically possible to integrate very large amounts of wind capacity in power systems, the limits arising from how much can be integrated at socially and economically acceptable costs.
Wind power production introduces more uncertainty in operating a power system: it is continuously variable and difficult to predict. To enable proper management of the uncertainty, there will be a need for more flexibility in the power system, in terms of generation, demand and/or transmission between areas. How much extra flexibility is needed depends on how much wind power there is and on how much flexibility already exists in the power system.
In recent years, reports have been published in many countries investigating the power system impacts of wind generation. However, estimates of the costs of integration differ and comparisons are difficult to make due to the different methodologies, data and tools employed, as well as the terminology and metrics used to describe the results. To address these issues an R&D project – Design and Operation of Power Systems with Large Amounts of Wind Power Production – was initiated in 2006 under the auspices of the IEA (IEA WIND Implementing Agreement "Task 25").
The first phase of the project has reviewed existing studies and resulted in a report* – summarised here – on the current state of the art in estimating the system impacts and costs of wind power integration. The end point of the project will be guidelines on methodologies and best practice.
Adding wind power to the grid has beneficial impacts such as reduced emissions from electricity production and reduced fuel consumption in conventional power plants. Wind power will also have a capacity value for a power system. However, possible negative impacts also have to be assessed to make sure that they do not outweigh the benefits and also to ensure security of power system operation.
The possible impacts of wind power on power system reliability and efficiency are depicted in Figure 1. Here we are dealing with system wide issues as opposed to local issues such as power quality. The impact studies considered fall into three focus areas: balancing; adequacy of power; and grid.
Balancing covers the impacts on allocation and use of short term reserves (timescale, minute to half an hour) and efficiency and unit commitment of existing power capacity (timescale, hours to days). In the timescale hours to days the positive impacts of wind power can be seen, reducing the use of fossil fuels thus cutting operating costs for the power system as well as decreasing emissions.
The unpredictable variations in large area wind power should be combined with any other unpredictable variations the power system sees, such as unpredicted variations in load and outages at conventional power plants.
The impacts of wind power on transmission depend on the location of wind power plants relative to the load, and the correlation between wind power production and demand. Wind power affects the power flow in the network. It may change the power flow direction, reduce or increase power losses and exacerbate bottlenecks.
There are a variety of means to get the most out of existing transmission lines, eg use of on-line temperature and load data, FACTS and wind power plant output control. However, grid reinforcement may be necessary to maintain transmission adequacy and security. Grid extensions are commonly needed if new generation is installed in weak grids far from load centres. The issue is generally the same be it modern wind power plants or any other power plant. The cost of grid reinforcements required to accommodate wind power is therefore very dependent on where the wind turbines are located relative to load centres and grid infrastructure, and one must expect numbers to vary from country to country. With current technology, wind power plants can be designed to meet industry expectations such as riding through voltage dips, supplying reactive power, controlling terminal voltage, and participating in SCADA system operation with output and ramp rate control.
Power adequacy is associated with static conditions of the system. It is about total supply available during peak load situations (time scale, several years). Estimation of the required generation capacity takes into account system load demand and the maintenance needs of production units (reliability data). The criteria that are used for the adequacy evaluation include the loss of load expectation (LOLE), for instance. The issue is the proper assessment of wind power's aggregate capacity credit in the relevant peak load situations – taking into account the effect of geographical dispersion and interconnection. Wind generation will provide some additional load carrying capability to meet projected increases in system demand. This contribution can be close to the average power produced by wind power at times of peak load, when the penetration of wind power is not high, and the capacity value of wind will decrease as wind power penetration increases. Aggregating large areas has a positive impact on the capacity credit of wind power.
Balancing the power system
Figures 2 and 3 summarise the results of impact studies that address balancing issues. Generally an increase in balancing requirement is a function of wind penetration level. However, countries and power systems differ in the way the variability and unpredictability of wind power impact the allocation and use of reserves as well as the costs incurred. Important factors are the region size relevant for balancing, time scale of system operation (how late updated forecasts will be taken into account), initial load variations and how concentrated/distributed wind power is sited. The added costs of balancing due to wind power will depend on the generation mix, marginal costs for providing regulation and mitigation methods used in the power system for dealing with increased variability (Figure 3).
In many cases the extra requirements for reserves may already be available which means that no extra investments are needed. Market rules that determine imbalance pricing can have an impact on costs if penalties are applied, so technical costs can be different from market costs. In addition to short term reserve allocation and use, variability of wind power also affects the way conventional capacity is run and the variations and prediction errors for wind power change the unit commitment.
The increase in reserve requirement has mostly been estimated by statistical methods, combining the variability of wind power with that of load. In some studies sudden outages in production are combined with reserve requirements (disturbance or contingency reserve).
To determine the impact on the operation of power systems, simulation models are run and most results are based on comparing costs of system operation without wind to those with varying amount of wind. The costs of variability are also addressed by comparing simulations with flat wind energy to varying wind energy (for example in US Minnesota and Greennet Nordic+Germany). The studies referred to in the figures are the following:
• Nordic/Finland 2004 (H Holttinen, Wind Energy, Vol 8, 2, 2005): Operating reserve requirements due to wind power in the Nordic countries were estimated in 2004, with up to 4 GW of wind in the 14 GW peak system of Finland and up to 18 GW wind in the 67 GW peak system of Nordic countries (10-20% penetration). Methodology is statistical, combining the standard deviations of wind and load variation time series. When looking at the increase in hourly variations from load to net load (load minus wind production) four times standard deviation of the variations time series is used as confidence level (4s). The estimate is made from 3 years of synchronous hourly time series for load and (up-scaled) wind power. The better predictability of load was taken into account applying load forecast errors instead of load time series (as part of the load variability can be forecast). The cost estimates take into account both new reserve capacity and the increased use of reserves. Estimates were made for the Nordic area assuming similar amounts of wind power in all four countries and separately for Finland. The Finland 2004 results do not take into account the interconnection capacity available.
• Sweden: The Swedish additional reserve requirements for 4-8 GW wind (7-13% penetration) for the 26 GW peak system were estimated based on a similar approach to the Nordic 2004 study, combining the standard deviation of load and wind variation time series (Elforsk Report, 2005). Time scales used were for one and four hour forecast errors separately. Several years of wind data were acquired based on meteorological data and synchronous load forecast error data was available. No cost estimates were made. Estimates do not take into account the interconnection capacity available.
• Germany: In the Dena study the regulating and reserve power capacity requirement to accommodate up to 36 GW wind in a 78 GW peak system (14% penetration) was estimated. Time scale was day-ahead: requirement for the following day was determined in relation to the forecast wind in-feed level using statistical methods. Probability density of wind forecast errors was combined with the probability of load forecast errors and outages of conventional power plants. One year of data was used. The additional reserve capacity required could be provided by the existing conventional power stations so no cost estimate was made. Estimates do not take into account the interconnection capacity available.
• Nordic countries and Germany: The GreenNet-EU27 study estimated increases in system operation costs as a result of increased shares of wind power for a 2010 power system case covering Denmark, Finland, Germany, Norway and Sweden. The integration costs of wind were calculated as the difference between the system operation costs in a simulation model (WILMAR) run with stochastic wind power forecasts and the system operation costs in a model run where the wind power production is converted into an equivalent predictable, constant wind power production during the week. Wind impact on reserve allocation was made in the model combining the probability of wind forecast error with the probability of load and generation variations (outages).
• Ireland: In a 2004 study by SEI the operating costs of the 6-7 GW peak system of republic of Ireland (2010 scenario) were estimated with up to 2 GW of wind (14% penetration). The wind input was a time series generated from statistical manipulation of historic wind power plant data (half-hourly). The load data was not synchronous. The methodology was generating system simulation for selected days (winter peak, summer valley, shoulder business day). Wind and load forecast errors were combined for different time horizons (1-8 hours ahead). Thermal power was simulated with outages, ramp rates and start/stop costs. Hydro power flexibility was not used in the simulations.
• UK: In a 2002 study by G Strbac/Ilex and a 2007 study by G Strbac et al (Electric Power Systems Research, 77, 9, 2007) the amounts of extra plant for reserve requirements were estimated at around 5% of the wind plant capacity, at the 20% penetration level (% of gross demand). Taking the original values and dividing by produced wind energy resulted in £2.38 per MWh of wind produced for 10% wind, rising to £2.65/MWh at 15% and £2.85/MWh at 20%. Estimates of extra reserve costs used market costs, which may be expected implicitly to include a capital recovery element. Statistical methods of combining wind and load variations were used, and also simulations for selected days were performed to calibrate and to make sensitivity and cost assessments.
• US/Minnesota 2004 and 2006: Three year data sets of 10-minute wind power profiles from atmospheric modelling were used to capture geographic diversity. Wind plant output forecasting was incorporated into the next day schedule for unit commitment. Time-synchronised historic utility load and generator data was available. The first Minnesota Dept. of Commerce/Enernex Study (2004), estimated the impact of wind in a 2010 scenario of 1500 MW of wind in a 10 GW peak load system. A monopoly market structure, with no operating practice modification or change in conventional generation expansion plan, was assumed. The second Minnesota Dept. of Commerce/Enernerx study (2006) took as a subject power system a consolidation of four main balancing areas into a single balancing area for control performance purposes. Simulations investigating 15%, 20%, and 25% wind energy penetration of the Minnesota balancing area retail load in 2020 were conducted. Incremental regulation and intra-hour load following burden due to wind was estimated at 3s confidence level. Hourly to daily wind variation and forecasting error impacts were found to be the largest cost items. For the 2004 study, a total integration cost of $4.60/MWh was found, where $0.23/MWh was due to increased regulation. For the 2006 study, the cost of wind integration ranged from a low of $2.11/MWh of wind generation for 15% wind penetration in one year to a high of $4.41/MWh of wind generation for 25% wind penetration in another year, compared to the same energy delivered in firm, flat blocks on a daily basis. The cost of the additional reserves attributable to wind generation is included in the wind integration cost. It was about $0.11/MWh of wind energy at the 20% penetration level. The remainder of the cost is related to how the variability and uncertainty of the wind generation affects the unit commitment and market operation.
• US/Colorado: The Xcel Colorado/Enernex Study (2006) examined 10% and 15% penetration cases (wind nameplate to peak load) in detail for a ~7 GW peak load system. Regulation impact was $0.20/MWh and hourly analysis gave a cost range of $2.20-$3.30/MWh.
• US/California: The CA RPS Integration Cost Project examined impacts of existing installed renewables (wind 4% on a capacity basis). Regulation cost for wind was $0.46/MWh. Load following had minimal impact.
The following differences have been noted:
• Time scales for prediction errors of wind power. For Nordic 2004 and Ireland only the increased variability during the operating hour has been estimated. For UK, the increased variability to 4 hours ahead has been taken into account. For US studies also the unit commitment impact for day-ahead scheduling is incorporated. For the Greennet study, the unit commitment and reserve allocation are done according to wind forecasts but the system makes use of updated forecasts 3 hours before delivery for adjusting the production levels.
• Costs for new reserve capacity investment. For the Greennet and SEI Ireland studies only incremental increase in operating costs has been estimated whereas investments in new reserves are included in some results (Nordic 2004).
• Size of balancing areas. The Greennet, Minnesota 2006 and Nordic 2004 studies incorporate the possibility of reducing operation costs through power exchange to neighbouring countries, whereas Colorado, California, German Dena study, Sweden, Finland, UK and Ireland studies analyse the country in question without taking transmission possibilities into account. The two studies for Minnesota show the benefit of larger markets in providing balancing. The same can be seen from the Nordic 2004 results compared with results calculated for Finland alone. Larger power systems make it possible to smooth the wind variability.
Summary of results on balancing
Summarising the results on balancing we can say that at wind penetrations of up to 20% of gross demand (energy penetration), system operating cost increases arising from wind variability and uncertainty amounted to about 1-4 r/MWh. This is 10% or less of the wholesale value of the wind energy. It can be seen that there is considerable scatter in results for different countries and regions.
The increase in short term reserve requirement has been estimated to be less than 4% of installed wind capacity with low penetration (<10% of gross demand) and for hourly variability of wind, to about 5% for forecast errors for 4 hours ahead and to nearly 10% if day-ahead forecast errors are left to be balanced with the short term reserves.
Some studies take into account the impact on short term reserves only, whereas some studies estimate also the impacts of wind variability on how the conventional power plants are scheduled and operated (day-ahead, unit commitment). Some studies take into account costs of building new reserve capacity whereas most studies calculate the increased use of (existing) regulating and reserve power.
A general conclusion is that if interconnection capacity is allowed to be used also for balancing purposes, then the balancing costs are lower compared with the case where cannot be used in this way. Larger balancing areas allow access to larger balancing resources and the aggregating benefits for wind power (smoothing of variability and reducing forecast errors). Also operating the power system closer to the delivery hour will reduce the imbalance due to wind forecast errors and lower the negative impacts of wind power.
Wind impacts on the transmission grid
Grid reinforcements may be needed for handling larger power flows and maintaining a stable voltage, and is commonly needed if new generation is installed in weak grids far from load centres. The grid reinforcement costs, according to the studies investigated, vary from 50 r/kW to 160 r/kW. The grid reinforcement costs are not continuous; there can be single very high cost reinforcements. Also there can be differences in the way costs are allocated to wind power.
Adequacy of power
Wind generation will provide some additional load carrying capability, capacity value, to meet expected increases in system demand. The results of national and regional studies on capacity value are depicted in Figure 4. The capacity value of wind power can be up to 40% of installed capacity if wind power production at times of high load is high, and down to 5% with higher penetrations and if local wind characteristics correlate negatively with the system load profile. Aggregating larger areas benefits the capacity value of wind power.
Issues to be addressed
Challenges for wind impact studies include developing representative wind power production time series across the area of study, taking into account the (smoothed out) variability and uncertainty (prediction errors) and then modeling the resultant power system operation.
For high penetration levels of wind power, the optimisation of the integrated system should be explored. Modifications to system configuration and operation practices to accommodate high wind penetration may be required. Not all current system operation techniques are designed to correctly incorporate the characteristics of wind generation and certainly were not developed with that objective in mind. For high penetrations also the surplus wind power needs to be dealt with, eg, by increasing flexibility in the generation mix, transmission to neighbouring areas, storage (eg pumped hydro) or even demand side management (avoiding wind power curtailment).
There is a need to assess wind power integration at the international level, for example to identify the needs and benefits of interconnection of national power systems.
Integration costs of wind power need to be compared to something, such as the production costs or market value of wind power, or the integration cost of other power generation systems. There is also benefit when adding wind power to power systems: it reduces the total operating costs and emissions as wind replaces fossil fuels. Indeed, the benefits are expected to be significantly higher than the costs. Taking fuel savings only, these will be roughly proportional with the wind generation, and a magnitude higher than the foreseen cost of balancing. In this summary only the cost component has been analysed. In future studies it is likely that the cost benefit approach will be seen. An example of this is the recently published Irish All Island Grid Study.