Alstom completes successful 5-day hardware in-the-loop test running FORCES Pro generated code on industrial wind turbine control platform
Wind turbine manufacturers must operate the turbine at its limits to remain competitive by maximizing the revenue potential of wind farms. Electronic control is necessary to ensure a safe operation of the turbine while keeping a high energy conversion efficiency, reducing maintenance and prolonging the structural lifetime.
For next generation wind turbines, with many mechanical and electronic actuators, model predictive control is a very attractive technology to systematically handle multiple conflicting control objectives and system constraints. In this scenario, tuning conventional PID controllers would be extremely costly and time consuming, in particular when multiple actuators on the blades are present.
In this project, Alstom’s engineers used embotech’s tools to quickly design, prototype and test optimal multi-variable controllers for an existing wind turbine. By adopting this design methodology they can now systematically handle future turbine upgrades and changes in specifications.
Alstom had developed several model predictive control formulations and evaluated them in simulation studies. The simulations suggested that when the controller can predict further into the future it is possible to improve the performance of the controller.
Unfortunately the long prediction horizon resulted in large optimization problems that had to be solved in real-time on an embedded control unit. The performance of commercial optimization packages was orders of magnitude slower than what was required to close the real-time feedback loop on the wind turbine control unit.
Another challenge was the numerical stability of the optimization algorithms: many existing commercial solutions deliver on rare occasions numerical errors when faced with real-world data, a situation that must be prevented in an automatic control environment that runs without operator interaction.
embotech’s tool FORCESPRO allowed a single Alstom engineer to define the finite horizon optimal control problem with system constraints and automatically generate highly customized code implementing the optimal predictive controller. Subsequent design iterations were significantly simplified by the code generation technology, and simulations could now be carried out more than 10x faster than with any other commercial solution.
Once the controller behaviour was satisfactory in simulation, the code could be directly downloaded to the embedded control unit for hardware-in-the-loop testing and verification of the simulation results. Because the simulation was using the deployed controller code, a perfect match between the actual implementation and the prototype was guaranteed.
Implement predictive controller on industrial control platform.
Using FORCESPRO, controller code was generated for target platform.
- A single control engineer designed controller and implemented it
- Successful hardware-in-the-loop test for 5 days in a row
- No numerical errors in more than 40 million function calls
The controller ran for 5 consecutive days without a single numerical error in about 40 million function calls.
- The MPC controller executes within the allowed time imposed by the real-time control loop: by customizing the optimization solver and the C code implementation to the wind turbine control problem and the embedded microcontroller, embotech’s tools delivered code capable of meeting the tight real-time deadline.
- The MPC is very robust and reliable in highly variable wind conditions: by implementing optimal control actions the controller successfully ran uninterrupted and without fault for a test period of 5 days on the embedded control unit in high turbulence wind conditions. As expected, the results accurately mirrored the predicted simulation results.