Wind Energy: Production Prediction Predicament and Future of Pakistan’s Energy Sector

Pakistan relies heavily on traditional energy production methods to meet its needs.  Most, if not all, of the country’s industry currently relies on imported oil, which by mid-2018 is expected to increase to $80 per barrel from its current price of $68.5 per barrel. The other energy option the country is exploring are the coal deposits from the Thar Desert. Although work on the Thar project has started, costs involved in the extraction along with the environmental costs could make this coal a less viable option for the country economically.

Compared to these hydrocarbon options, Pakistan’s renewable energy options of wind and solar seem not only better for the environment but could be more cost-effective for the country.

However, this does not mean that renewable energy production is without a set of its own unique challenges. Clean energy sources are unreliable. Sun can stop shining at any time of the day and the wind can change direction at the drop of a hat. But energy demands of the county need to be met on a regular and consistent basis. Grid operators, power producers and energy traders have a critical need to accurately forecast energy generation from renewable sources  to offer bids in the day-ahead market. For wind power generation, the forecasts need to be made for two discrete lead-time intervals.

Number one is the day-ahead forecast which is used to set electrical markets that auction off the right to sell power into the grid for bidding on delivery of electrical generation for each hour of the following day. Number two is the short-term forecast that is typically issued an hour or few in advance and is of particular interest to electrical grid operators, who must balance electrical supply and demand perfectly  as they can save money through reduction in costly last-minute power purchases or curtailments, when power producers are asked and at times paid to cut back generation.

However providing these accurate wind power forecasts is no mean feat. Renewable generation is a nonlinear and bounded process. Factoring in the chaotic nature of the atmosphere and the fact that wind generation is directly related to wind speed and direction, the characteristics of this process are constantly, albeit slowly, changing with time. Our weather comprises the lower part of the Earth’s atmosphere and its turbulence is a result several natural factors like the planet’s surface, natural obstacles like mountain’s and forests, temperature gradients etc.

The engineering side needs to take into account all these dynamic factors to model the   power output of a wind turbine generator. The WTGs power curves are computed based on ideal wind conditions of that of a wind tunnel i.e. at a given air density, no altering obstacles, zero to minimum turbulence, and the turbine always incident to the wind. In practice, however,  turbines operating on windfarms will never completely replicate the computed power curves. This is due to the fact that some turbines within a wind farm could mask the output of the other turbines  due to the so-called shadowing effect. This effect can manifest  due to several factors including wind direction and turbulence due to the surrounding topographic and orographic effects.  These factors can have a negative impact on a turbine’s output. It is to be noted that various WTGs tend to experience wind conditions different from the free-stream wind within the farm. Consequently, the resulting power curve has features far more complex than the theoretical power curves provided by the manufacturers for individual wind turbines.

Another important element to consider is wind shear as it along with turbulence relates to fluctuations in wind speed and  direction over short distances. The potential power output is computed while assuming that the wind will consistently interact with the turbine in parallel to the rotor axis. Misalignment of the rotor axis with respect to the wind direction results in a yaw angle error and in turn substantially reduces the power output. Also, wind direction shear is accompanied by shear in wind speed across the rotor swept area and the combined weighted effects of both these phenomenon needs to be considered for accurate wind power output assessment. Practical demonstrations have shown that the influence of turbulence, wind direction and wind speed shear can produce deficits in available power that could be up to 10-14%.

Forecasting of wind power generation is a complex task, whatever the time-scale and it involves the following fundamental steps:

1. Numerical Weather Prediction (NWP) Data: The wind power forecasting begins with the Numerical Weather Prediction (NWP) models which are based on equations that  relate to the motion and forces affecting the movements of fluids. However each computation starts with initial conditions originating from recent measurements of weather variables such as temperature, humidity, wind speed & direction etc.  The records are collected from satellites, weather balloons, thousands of weather stations, connected smart phones & cars, flight data and various other avenues globally. This results in terabytes of data daily which is managed and then intelligently blended through machine learning into supercomputers running global weather models multiple times a  day to generate each forecast. It is to be noted that  meteorological forecasts are not for very specific as in these forecasts are generated at specific nodes of a grid of an area. Since generally wind farms are not situated on these nodes, it is then needed to extrapolate the forecasts to the location and at the required turbine hub height.

2. Site-specific Refinement and Model Adaptation In this step, the global weather models are refined to adapt to local climate, topography and geography, with hyperlocal forecasts at precise latitudes, longitudes of turbines. Other static variables are incorporated relating to the site generator hub height etc., and in turn the individual NWP models are synthesized using intelligent model combination algorithms to produce the ensemble best average forecast. Subsequently the consolidated model is tuned using real time wind mast data (which captures weather parameters such as wind speed, direction, pressure etc.) to attach dynamic weights as well as static weights based on historical data of a few years. This ensures that any bias and forward error is corrected in the prediction models.

3. Power Conversion Once we have hyperlocal wind forecast with a reasonable degree of accuracy, the key bit thereafter is the translation of these metrological figures into effective energy production metrics. Complex power generation conversion computations are made by taking into account the maximum capacity, turbine essentials, such as make, model, hub height, and turbine power curve data. The results are eventually tailored using machine learning algorithms for power model adaptation using the real time and archived data from the furnished SCADA system. SCADA provides information into the operational status of the wind turbine generators, power generated as measured by the energy meters, capacity factor, curtailment information etc.

Wind power generation has a stochastic nature and exhibits fairly unique features that require deeper knowledge. Even better understanding and modeling both the meteorological and power conversion processes cannot reduce the inherent and irreducible uncertainty in every prediction. Today, prediction uncertainty is either expressed in the form of probabilistic forecasts for better decision making or with risk indices provided along with the traditional point predictions.

More research is required to  improve weather predictions model and power conversions  to assist stakeholders in making better decisions.

LMKT is working towards developing and promoting robust forecasting systems for renewable energy power producers and regulators. We are working closely with leading wind energy companies in Pakistan to pilot an advanced forecasting solution. As a company, LMKT is committed to investing in clean technologies and collaborating with the Government on energy policy advocacy to meet the looming energy crisis.

About the Author An energy sector enthusiast with a keen interest in smart grids and clean energy technology, who is presently working as an engineer at LMKT. He is also an avid Manchester United fan.