Global warming is a serious threat to the well-being of future generations and their ability to interact with this planet. While there are many industries that need to change protocols to help combat Climate Change, one of the most critical is the energy sector. Electricity has allowed us to develop incredible technology and connect us across the globe. However, it is also a large contributor to greenhouse gas emissions.
"According to a study conducted by the EPA in 2018, electricity production produced 26.9 of total US greenhouse gases." [1]
Renewable energy has a large part to play in reducing this number.
There are many types of renewable energy, one being wind power. Wind power is generated by using large turbines to harness the kinetic energy of wind. Wind causes large turbines hooked up to generators to rotate and produce electricity. These turbines can be very large, standing over 80 meters high. Generally, these turbines are built in wind farms, which contain a large number of turbines close together.
In this post, we will look at wind power generation at the High Winds Project in Northern California which was conducted as part of my project for the Weather and Storms class at Stanford University. The goal of this post is to give a general sense of the relevant Data involved and how it can be used to estimate wind power. I also want to show the general magnitude of numbers for things such as power generation of a typical wind farm. To give a sense of the current state of wind power in California:
In 2018, California wind projects generated 15,078 gigawatt-hours of electricity which accounted for 7.2% of all power generated . [2]
What is Wind? Did you know there are actually different types of wind? In class we learned about the four main types: geostrophic, gradient, surface along straight isobars, and surface along curved isobars. Each wind differs in the combination of forces that drive it. For example, surface winds acting along circular isobars have the coriolis, pressure gradient, friction, and apparent centrifugal force acting.
![Forces for Surface Wind Along Circular Isobar [3]](https://towardsdatascience.com/wp-content/uploads/2021/01/129VXRyHct9eAvwBQ5i2p-A.png)
Topography of an area also plays a role in the wind generated for that region. In northern California, specifically the San Francisco bay area, the terrain is mountainous and located close to the sea. Near the sea, when the land is warmer during the day you can get a land breeze, where the wind blows from the land to the ocean. Conversely, at night when the land is colder you can get a bay breeze, where the wind blows from the bay onto land. Pressure and temperature play a big role in wind formation in the atmosphere.
There are two major types of wind turbines: vertical axis and horizontal axis. Vertical axis wind turbines have their axis perpendicular to the wind stream, while horizontal axis turbines axis’s are parallel to the wind stream. In general, the majority of wind turbines in production are horizontal axis.
Taking a closer look at the mechanics of horizontal axis turbine power generation, these structures harness the energy from wind via large propeller-like blades. Often there are three steel blades mounted to a large tower. Blades are mounted high up to take advantage of the faster wind speeds at higher altitudes. Wind blowing across the turbine causes low-pressure on one side of the blade, causing lift which rotates the blades. Due to the large inertial mass of the propeller blades, the shaft typically has a rotational speed of 5–20 rpms. To produce AC electricity, the shaft of the propeller blades is connected to a gearbox which increases the rotational speed up to 1800 rpm.
A study conducted by El-Ahmar et. al (2017) investigated the most critical factors for wind power generation and devised the following list: wind speed, air density, rotor swept area, and tower height [4]. Each of these components influences the power output differently. As you would expect, power generation is directly related to wind speed (i.e. larger wind speeds produce more power). In fact, the equation below is often used to measure the potential power output of a turbine. We will be using this moving forward to calculate power.
Diving further into the data, this project used climate information and historical wind power generation statistics to understand the operating capacity of the farm for a selected time period. As mentioned earlier, the High Winds Project was chosen from the USGS wind turbine database for analysis and data was collected for 2018, ranging from January 1, 2018 to December 31, 2018 [5]. The project, started in 2003 and is located northeast of the bay, just southeast of the city of Fairfield. The project contains a total of 90 1.8 MW rated Vestas V80–1.8 turbines, for a total capacity of 162 MW. Each turbine stands 60 m tall with a rotor diameter of 80 m. Below is a map highlighting the location of the 90 turbines.
![High Winds Project in Northern CA [5]](https://towardsdatascience.com/wp-content/uploads/2021/01/0qTMrKgX3Nd9tpZYl.png)
In addition to the individual turbine information, summary power output information was gathered from the California Energy Commission [6].
Climate information was gathered from the National Oceanic and Atmospheric Administration (NOAA) global forecast system (GFS) atmospheric model. The GFS system covers the entire globe with a resolution of 28 km. For this analysis, the following variables were considered: time, surface temp (K), temp at 2m (K), eastward wind velocity (m/s), northward wind velocity (m/s), relative humidity at 2m (%), and sea level pressure (Pa). Again data was collected for 2018 at a central coordinate of approximately (38, -122).
What did I find? According to this analysis, the farm would be operating below capacity in 2018. There were assumptions made in this analysis, such as using a simplified wind model, which could explain the slightly lower operating capacity than typically found for wind turbines. A comparison was conducted between the expected output based on the wind speeds and the total capacity. The output and capacity were used for a single turbine. In the figure below, the red line represents capacity and the blue line represents power output for the wind speed on that day. Looking at the graph, most days are well below capacity.
Creating a distribution of the actual capacity percentage helped put the previous figure in perspective. From the figure below, it is clear to see that the turbines in 2018 were operating at well below capacity. Most of the days were 10% or below.
In addition to operating capacity, the trends in wind direction were calculated. The figure below summarizes the direction of wind over 2018. The 360 compass degrees were divided into equal sections of 45 degrees and any data point inside that range was assigned the corresponding compass direction. From the results we see that the winds mostly come from the north/east direction. A possible explanation could be the topography, which explains the Diablo Winds phenomenon. To the northeast of the bay there is a gap between the mountain ranges, which could allow for mountain breezes to develop and head towards San Francisco.
In addition to the wind direction, the speed (green), temperature (yellow), pressure (purple), relative humidity (blue), and power capacity (red) were all plotted together. The next figure shows these variables plotted over the entirety of 2018.
For a full dive into the methodology, check out the final report and jupyter notebook on github at the bottom of the article
Renewables are growing every year, and need to continue if we are to reach emission goals to curb global warming. Wind energy has been, and will continue to be, an important part of the solution.
I hope this provided a good look into the complexities of predicting power generation from wind farms. While there are liberties taken with the simplicity of this analysis, it goes to show how much insight can be gained from using weather variables for predicting wind energy production.
In order to achieve better results, more complex wind models can be used to get the speeds and directions at the locations of each turbine. Data can be collected over a longer period of time to better understand historical trends and forecast when production will be high or low. And with the advancement of machine learning, forecasting these variables has become more precise which can play a role in demand planning moving forward.
Github with full report and code:
References:
[2] – California Energy Commission, Energy Almanac, Total System Electric Generation (2018). (See https://www.energy.ca.gov/data-reports/energy-almanac/california-electricity-data/2018-total-system-electric-generation.))
[3] – DANIEL A. VALLERO, 5 – The Physics of the Atmosphere, Editor(s): DANIEL A. VALLERO, Fundamentals of Air Pollution (Fourth Edition), Academic Press, 2008, Pages 123–153, ISBN 9780123736154, https://doi.org/10.1016/B978-012373615-4/50006-6.(http://www.sciencedirect.com/science/article/pii/B9780123736154500066)
[4] – M. H. El-Ahmar, A. M. El-Sayed and A. M. Hemeida, "Evaluation of factors affecting wind turbine output power," 2017 Nineteenth International Middle East Power Systems Conference (MEPCON), Cairo, 2017, pp. 1471–1476, doi: 10.1109/MEPCON.2017.8301377.
[5] – https://eerscmap.usgs.gov/uswtdb/
[6] – https://ww2.energy.ca.gov/almanac/renewables_data/wind/index_cms.php
