Performance Improvement of wind energy system
Interest in renewable energy systems is currently at the highest level for many reasons, such as unlimited availability, global warming and hazards associated with nuclear energy. Wind energy systems in particular have attained much more attention in comparison with other renewable energy sources. W
2025-06-28 16:34:26 - Adil Khan
Performance Improvement of wind energy system
Project Area of Specialization Artificial IntelligenceProject SummaryInterest in renewable energy systems is currently at the highest level for many reasons, such as unlimited availability, global warming and hazards associated with nuclear energy. Wind energy systems in particular have attained much more attention in comparison with other renewable energy sources. Wind is safe, green and clean source of energy. Cumulative capacity of wind energy systems has increased from 6.1GW to 432.4GW over the last two decades. Effective use of the latest technology, the ever-increasing size of wind turbines and the clustering of turbines in wind farms have made wind energy more attractive and affordable as compared to other renewable energy sources but due to wake effects and turbulence, wind turbines cannot produce power upto its maximum capacity. This results decrease in power production of wind turbines and causes upto 60% loss in energy.
In our Final Year Project (FYP), we are overcoming these problems through AI based model. The model would predict the pitching angles on runtime of each turbine and direct it to produce power according to wind characteristics. In this way the power production of whole wind farm would be increased.
Machine learning and data analytics techniques play vital role in our project that would give deep insights from the wind farm data including the factors that effect the power production and determining the wind characteristics.
After analytics, we would implement different optimisation techniques such as Particle Swarm Optimisation (PSO), Ant Colony Optimisation (ACO), Genetic Algorithms (GAs) and Brute Force (BF) in order to find out which one is performing the best so that we get the optimised power production from the wind energy system efficiently.
Objectives of our project are as follows:
- To increase the net productivity of the wind farm rather than just increasing the productivity of single wind turbine.
- To develop control strategies for increasing production per area by optimally controlling wake effects through pitch angle and yaw based optimisation.
- To use the wind corridor sophisticatedly.
- To reduce energy/power loss caused by wake effects and turbulence in a wind farm.
Normally each wind turbine in a wind farm operates in such a way that it consumes maximum momentum from the wind it is facing. In this greedy control method the downstream turbines are left with no or very low wind speed which adversely affect the net power production of the wind farm.The proposed solution for this is to use co-ordinated wind farm control strategies. Optimisation techniques will be used to maximise the net power of the co-ordinated wind farm. In wind turbines, there are basically two controllable entities which are pitch control and yaw control. Both of these variables contribute in net power evaluation. Pitch control means to adjust the angle of blades of a turbine. It mainly decides how much wind should the turbine consume.
Yaw control is the angle between the wind direction and the hub of the turbine. Yaw angle controls the angle with which the turbine faces the wind.
Optimisation techniques would be used to try different possible combinations of these inputs (controllable variables) with a suitable step size to find optimum solution. The best solution for the two controllable variables will be the one which gives the maximum net energy of all the wind turbines. Different optimisation techniques would be tested for finding the global best solution.The optimisation with least computing overheads and with small processing delays would be selected. The optimisation process would be carried out on a distributed network of raspberry-pie cluster. Each raspberry pie would be assigned equal number of turbines.The cluster of raspberry pie would be connected through a switch which will create a multiprocessing environment. The data would be finally passed to a central computing unit for decision making purpose.
The testing phase would require Brute Force to be used for bench-marking and then different optimisation technique's performance would be compared with the benchmark.The optimisation technique with best accuracy and highest speed would be deployed.
There are about 21 wind farms in Jhimpir, Sindh, Pakistan. Each of these wind farm is producing 50 to 100 Mega Watt. This makes more than 6% of the total energy requirement of Pakistan. Wind Energy is a source of renewable energy which means its has no or very less contribution to environmental pollution. Another reason which makes wind energy of great importance is that it's comparatively less costly as it don't require a huge infra structure.
The upstream turbines face the free stream wind and slows down the wind leaving it very less for the down stream turbine. This problem is known as wake effect. By using the Data-Analytic and Optimisation techniques, these wake effects can be reduced resulting in a higher power production without any major alteration in the infra structure of the wind farm. This can greatly help in solving the energy crisis problem in our country.
Technical Details of Final Deliverable- A python Jupyter notebook based algorithm for modelling a wind farm.
- A powerful centralised computing cluster of raspberry-pi which will run the optimisation process.
- A generic objective function (minimisation function) which would optimise the wind farm parameters for yielding maximum power.
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 80000 | |||
| Raspberry Pi 4 Model B | Equipment | 4 | 14500 | 58000 |
| GigaBit Switch | Equipment | 1 | 3000 | 3000 |
| Raspberry Pi Cases | Equipment | 4 | 500 | 2000 |
| Spartan 6 Xylinx FPGA | Equipment | 1 | 7000 | 7000 |
| Printing, Stationary, Data Collection, Wires etc | Miscellaneous | 1 | 10000 | 10000 |