The most popular controller in industry is of course the PID controller and most of the tuning methods do not require a model of the system be known. This is in fact a great advantage of the basic PID algorithm, since models can be expensive to construct and require a level of expertise, which may not be available. A good question is therefore why is model based control becoming so popular.
Most of the model based advanced control algorithms involve optimal control or optimization in some way and this includes many of the predictive control laws. The use of a cost function to judge performance provides a useful benchmark for comparison, and it provides tuning inputs, which have a physical meaning. However, despite all the advantages of using optimal methods there are not many occasions when the cost-function truly actually reflects the needs of the actual system. The cost-function is therefore used as the way to incorporate design information, rather than a physically motivated criterion to be optimized. This is not therefore the reason model based control has been so popular.
Models are of course required for modelling physical disturbances such as wind speed variations in wind turbines and wave disturbances in marine systems. The prediction equations required in predictive control do of course require reasonable models to provide the predictive capability. Thus, although models are essential for representing the system behaviour adequately, the benefits of model based control are still not obvious.
The main benefit experienced is rather indirect and not so obvious. In fact, it is the understanding that is gained through building models of the system, and the understanding of behaviour and performance, which is achieved. The very act of building models and producing a simulation adds to the engineers understanding of how to design a control a system. It reveals where the major disturbances, interactions, nonlinearities and problems are likely to arise. It is this rather hidden property of model-based control, which has made it so successful and why when high performance is essential that this is the approach which is recommended.
If on some future occasion I am to be sent to a desert island and can take only one control technique with me then it will have to be the very trustworthy simulation capability. PID control would of course be helpful on the sandy beaches but simulation is probably the most important tool for the control engineer. In years gone by this was certainly not the case and in fact only 30 years ago simulation facilities were what might be described as primitive. However, in recent years simulation has become a valuable tool, certainly for those employed in advanced control systems design.
In many industries simulation is needed to reduce the time is takes implementing advanced model based controls. The tuning of such systems can be performed in a safe and fast manner and what if questions can be answered. Because of the rapid model change in automotive applications there is a real need to reduce time consuming activities like engine calibration. The generation of engine maps and the tuning of controllers can be reduced in time if good simulation tools are available. Since the trend is to use model based control, the models needed for simulation will of course exist and tools like the LabVIEW based simulation facilities and the ubiquitous Matlab/Simulink are very common.
The need for simulation in for example energy systems is often to improve the total system control. The requirements for modelling and control in individual wind turbine control are surprisingly similar to those in automotive engine control. However other problems like total wind farm power control, whilst minimising fatigue and wear and tear requires a rather different approach. For individual wind turbines high fidelity models are required for each turbine here as for wind farm control the complexity of the models needs to be reduced and in any case sophisticated models are nor needed for the much slower dynamics regulating total wind farm power outputs. In this case problems arise due to the scale of the modelling and simulation problem.
Getting back to the sandy beach on the desert island you may have wondered why simulation was so important. The main reason is often associated with the comfort a good simulation provides. By knowing how the system is very likely to perform and by having a range of possible controller designs all if which give adequate performance the visit to the application is a lot less worrying and fraught. In my case since I can choose the desert island I think I will make it one in Hawaii and I am sure there are many applications to explore.