A pre-scheduled adaptive control scheme based upon system knowledge

Most solar thermal power plants provide thermal energy in the form of oil heated by an array of parabolic trough collectors. The industrial processes connected to such a field require the energy to be supplied at a constant temperature despite fluctuating levels of solar intensity, so a control scheme that varies the oil flow rate is applied to the collector field. This scheme generally utilises a combination of feedback and feedforward control, but the design of the feedback controllers is nontrivial because of the existence of low frequency resonance dynamics. High-order plant models must be used if reliable, fast, well damped control is to be achieved. In addition, the dynamics of a collector field are heavily dependent upon the operating conditions, which can vary widely. Any feedback control scheme implemented on such a plant must therefore incorporate an adaptive mechanism to operate effectively. However, online estimation based upon high order transfer function models tend to perform poorly with slow parameter convergence being the norm. Additionally, lower order transfer function models tend to yield unreliable performance and inadequate control bandwidths. Pre-scheduling is therefore an attractive proposition for this type of plant where model dynamics can be accurately defined. This paper details the development of a pre-scheduled adaptive mechanism based upon a physical model of the Acurex solar power plant