Due to a number of factors, there is a growing interest in time-varying energy calculations.
Traditionally, most wind farm energy production calculations use some sort of annually averaged wind distribution. Typically, the measured wind distribution will be fitted to a Weibull curve to reduce the computational requirements of handling large amounts of data. This means that the wind distribution represents an average year. It also means that time-varying changes in parameters such as air density, turbulence, shear and veer during the year – and within each day – are not taken into account. They are simply represented by annually averaged values.
Improving efficiency and reliability
The time-varying calculation method is very helpful in identifying operating issues that are responsible for production losses so that efficiency and reliability can be improved. Making these time-based adjustments to the power curve and the wake decay constants allows for a much more precise matchup and correlation of concurrent calculated and measured energy production. This leads to a more precise determination of energy losses that can be evaluated by mean wind speed, temperature, turbulence, etc. Thus, the problem areas can be identified and quantified and actions can be taken to improve the operation.
Because of the ever-increasing number of wind farms being built, the likelihood of having a nearby operating wind farm is increasing. This presents an opportunity to fine-tune and calibrate the wind model for pre-construction wind and energy assessments.
In addition to improving efficiency and reliability, using the time-varying method in a performance check also helps to isolate and identify any remaining modeling errors (e.g., elevation or roughness modeling) that may be responsible for differences between predicted and actual energy production. The end result is that the refined wind model will be more accurate for predicting pre-construction energy production for a new wind farm that will be located nearby, as well as have a better understanding of how the local terrain influences the wind flow. This provides valuable feedback on how to fine-tune the model to reproduce the actual production as accurately as possible.
Once the performance check and the associated model refinements are completed using a time series of wind data that is concurrent with the operating wind farm’s production data, the wind flow model can be updated to use a time series that represents the long-term average year, or simply calculate based on recent 20-year hourly meso-scale model data that are calibrated to reproduce the nearby wind farm for the period with concurrent data. Using the long-term representative time series in conjunction with the time-varying calculation method has been shown to reduce the uncertainty of the yield estimate from a range of 10% to 15% to a range of 5% to 10%. This is primarily due to the increased accuracy that is achieved by fine-tuning the wind flow model in the performance checking exercise when testing the model calculations against real operational data.
In some cases, the Weibull-fitted distribution curve will be poorly fitted to the actual measured wind distribution. In this case, because the Weibull fitting is typically energy-weighted, it will fit better at the higher wind speeds and not so well at the lower wind speeds. This can be a real problem for lower-wind-speed sites. Studies indicate that there can easily be a 20% difference between energy yield predictions using a poorly fitted Weibull distribution versus using the time-varying energy calculation method.
Valuing the wind energy in the financial model (pro forma) is becoming more complicated as the value of the wind energy becomes tied to the time of generation. As the electrical grid becomes more congested with renewable energy generation, it becomes increasingly important to understand the relationship between the time that the energy is produced and the time that the energy is needed. The value of the energy is closely tied to this relationship. Wind energy that is generated in the middle of the night will have little value if the utility’s load profile indicates that customer demand for power at that time is at a minimum. Conversely, wind energy that is generated in the middle of the afternoon in July will likely have a much higher value due to air conditioning requirements. Time-varying energy calculations are essential to predict the value of the wind power generation in a market where prices are often no longer fixed but vary by time of day and time of year.
Turbine manufacturers are constantly working to improve the efficiency and reliability of their turbines. Many of these improvements involve control strategies – for example, using variable or multiple power curves to vary the power output for different climatic conditions or for different curtailment reasons. These variations are difficult to capture and model in the energy calculations using the annual-average wind distribution method but are easily handled with the new time-varying calculation structure.
The Weibull-fitted wind distribution has been around for a long time and has served the wind industry well. One can argue that many of these time-varying elements will cancel each other out over the course of the year and that the annual average will be correct with regard to the annual energy yield estimate. Although this may still be true in most cases, the changing energy market and renewable energy’s value within the overall energy mix have made it increasingly important to understand the time-of-generation value of the energy and impact on the grid with regard to congestion and curtailment.
In addition, precisely matching time-specific wind speeds with time-specific generation becomes more necessary as additional megawatts of generation come online, as this information is critical to making sure that the wind farms are being operated in the most cost-efficient and reliable manner possible.
This article was published in the November 2016 issue of North American WindPower