Commande avancée pour la régulation des aménagements hydrauliques

Advanced Control for Hydro-Powerplants

¹ EDF/CNEH Savoie Technolac
² Conseiller Scientifique SUPELEC

Abstract

This paper presents new solutions for the control design of Hydropower plants. The chosen examples is a run of the river type operations. Some control problems in France or in foreign countries have, lead to a verification of the existing controls in these plants, and also incited the instigation of advanced controls, carefully compared to the classical one.

The need for adequate models, as well as for powerful tools and algorithms for control design is a long standing issue. Non-linear partial derivative complicate models, which have been carefully tested through experience and used for classical control design using a trial and error approach have been in use for a long time. Subsequently, transfer functions were fulfilled by the identification of the complex numerical model time-responses. Indeed, this approach has not been efficient enough, as some dynamic term effects could not be identified due to the non-linear effects.

The modelization approach using non-linear differential models based on a lumped parameter representation of the physical system is much more powerful. Linearized state space models are deduced from the non-linear ones, which allow the design of convenient classical or modern controls. It is accepted that classical PI are greatly improved by additional loops assuring positivity, but powerful design techniques are missing for a PID plus a stabilising loop. Hence, an original robust design methodology has been implemented which uses dominant poles and modulus margin placement realised by an optimisation technique.

As a generalisation of this idea of additional loops, LQG control using all measurements available is particularly efficient and simulations on non-linear detailed models are presented.

Finally, due to the non-linear behavior of the plant and in order to obtain best performances at any flow rate, as at any set-point level, a gain-scheduling adaptive system using swuching fixed LQG or switching fixed PI with stabilising loop has been implemented and gives excellent results for the adaptive LQG and acceptable ones for the adaptive PI.

Both PI and LQG switching controls have been implemented with very efficient bumpless antisaturation components. Switching criteria have been automated according to the plant flow rate and to the maximum reservoir levels. Due to its general nature, this methodology has not only helped to solve run of the river control problems, but has also been used for speed power control of turbine units with efficient results.