Optimal control of a gas engine with consideration of grid dynamics

Abstract

Small-scale electrical grids with few power generating and power consuming components, so-called “microgrids”, are prone to frequency fluctuations caused by single grid participants. Internal combustion gas engines combined with generators are part of such microgrids. The relative slowness of power build-up from gas engines combined with the need to quickly compensate for rising or falling loads necessitate extensive control actions. These factors increase the risk of oscillations, both in engine power and grid frequency, occurring, particularly if the engine dynamics are not well known. Cases have been observed, where multiple engines shift the load between each other, even when the total load stays constant. To combat this, a way to abstract the electrical grid with its components into a low order linear state space system, a “virtual grid”, is proposed. In conjunction with the engine dynamics, it represents the dynamics of the whole grid. As all grid entities share the same frequency, it is chosen as input to the virtual dynamics. The output is a virtual torque acting on the torque balance. Parameters for this model are estimated from measurement data of the grid via a grey-box estimation process. A linear quadratic regulator for a gas engine based on the combination of engine and virtual grid dynamics is designed, which makes it possible to account for the grid behaviour in the cost function. To use the LQR, two observers are designed, which estimate the virtual grid states from the reference engine states and the measured engine states. As the engine has a time delay, an additional predictor based on the virtual grid and engine dynamics is implemented. The resulting controller, a combination of the observers, the predictor and the LQR, is tested in simulation against the existing controller of the engine. Both are developed with a deviation between simulation model parameters and control design model parameters to emulate not perfectly known engine dynamics. The simulation results show a significant reduction in the oscillations without losing performance.