Experience with the minimal realization approach identifying large MIMO system

A new state-space identification technique for MIMO systems based on their impulse responses is now well established. Following this approach, the authors use pulse responses generated by a transient stability program to identify three models of a large power system: five inputs and five outputs (5×5)-35-th order; 10×10-35-th order and 25×25-85-th order. This paper focuses on the computational constraints of the minimal realization algorithm (time and memory requirements) applied to such huge systems. It also discusses comparative results demonstrating the tradeoffs in identification accuracy using different algorithmic settings. A scalar metric is suggested, which allows quantitative assessment of the precision of the state matrices (a,b,c,d). The best settings are then selected by minimizing this performance measure, keeping in mind the associated computational cost. From the system matrices (a,b,c,d), the authors derive several types of very promising numerical tools, including controllability and observability measures, which illustrate the system natural frequencies and damping as well as the contribution of the hydraulic power plants to common and local modes dynamics