A Simple Multi-objective Control for Cascaded Hydro Power Plants

G. Robert*, F. Michaud*

*EDF Hydro Engineering Centre, Savoie Technolac, 73373 Le Bourget du Lac - France (e-mail: gerard.robert@edf.fr, frederic.michaud@edf.fr)

Abstract: This paper presents a simple multi-objective control technique for cascaded hydro power plants, based on a centralized distributed power control system and local water level controllers. This effective power management solution enables to guarantee the generation program following with a very good accuracy, to participate to power-frequency Ancillary Services with the dynamic performance required and to control water levels in every canals.

Keywords: Distributed control, multi-objective, hydro power plant, water level, power management.

1. INTRODUCTION

Hydroelectricity is the main renewable energy source in France, representing 21% (20 GW) of the total national installed capacity including nuclear and thermal power plants. As a power producer, EDF (Électricité de France) uses water resources to generate green electricity in France (239 dams, 447 hydropower stations with 20.300 MW of installed capacity, including 4.500 MW in pumped-storage) and in international projects.

EDF‘s Hydroelectric installations in the Durance and in the Verdon rivers constitute two chains able to generate 2.000 MW thanks to the two big reservoirs at Serre-Ponçon and Sainte-Croix (see Fig. 1).

Fig. 1. Durance-Verdon framework

Within this framework, three missions are entrusted at EDF:

1. electrical production;

2. water supply of the cultures (irrigation) and the cities;

3. regulation of the floods.

The Durance-Verdon unit produces 6 to 7 billion kWh per year (10% of the French hydroelectric production). It is divided in 8 lines where each line consists of a small reservoir feeding EDF cascaded run of the river power plants separated by canals. The lines provide drinking water to all the area, and irrigate all Provence (a third of the French irrigation). This irrigation constitutes an hydraulic disturbance for the water level control system.

Fig. 2. The Mallemort hydro power plant (line 5) Cascaded hydroelectric power plants transfer water flow from upstream to downstream without spillage and convert the

Preprints of the 18th IFAC World CongressMilano (Italy) August 28 - September 2, 2011

Copyright by theInternational Federation of Automatic Control (IFAC)

4960

flow into electric energy through hydraulic turbines. A supervision system (SCADA) manage the 19 Durance-Verdon cascaded power plants from the generation program calculated by EDF Optimisation Department a day ahead. This remote control system sends the production program (programmed power converted to a flow setpoint trajectory) in open loop. A water level controller is implemented in each power plant in order to maintain the water level constant in the canals in spite of agricultural withdrawals. This controller acts on the gate opening of the hydraulic turbine, modify the generator output power and therefore create a deviation between the generation program and the generated power. Recently, EDF Hydro Generation & Engineering Division launched a control system modernization project in order to cancel this power deviation and offer power-frequency Ancillary Services to the grid.

In France, producers must offer a certain amount of power-frequency Ancillary Services (primary power-frequency control and secondary power-frequency control). The Transmission System Operator (TSO) prescribes the requested participation (volume & dynamics). The provision of this service consists in modulating the generated power according to frequency variations on the electrical power system.

Since it is not possible with only one control input (gate opening of the turbine) to manage both the water level in the canals and the generated power, EDF Hydro Engineering Centre has designed a multi-objective control technique based on a distributed power controller centralized in the supervisor and existing water level controllers localized in power plants control system. The paper presents this global control approach. The first part will concern the water level control system and the second part, the distributed power control system.

2. WATER LEVEL CONTROL SYSTEM

Automatic water level control in open channels like in Durance is difficult because of the features of the hydraulic system:

non-uniform long canal (tens of kilometers) with variable roughness and profile

slow water flow dynamics described by 2 coupled nonlinear partial differential equations (Saint-Venant equations), where the unknown functions are time and space dependent

multivariable nonlinear finite-dimensional system,

unknown disturbances, which are due to natural inflows or agricultural withdrawals.

In order to ensure hydraulic safety of the system, plants of a same line are managed with the “flow equality” principle (water volume conservation): at any time, the balance between all water flows through a canal must be kept equal to zero. Thus, neither the canal is filled in, nor it’s drained. This action is completed by a water level regulator in each power plant: a linear combination of three levels (upstream, middlestream and downstream), representative of the volume,

is maintained to a specific setpoint, by changing the flow passing through the generation units. The controller is PI type, tuned with regards to previously pointed-out specificities of canals. The regulating scheme is shown in Fig. 3.

Fig. 3. Water level control system for one line

The flow production program (in m3/s) is sent to every power plants of the line as a “feedforward term”, obtained from the power generation program (in MW). The conversion from electrical power to flow isn’t perfect, bringing a first gap between the generation program and the power provided.

When a canal is upon a perturbation (for instance, an unexpected withdrawal), the level regulator will restore the balance, making the production program not being respected. Moreover, all downstream power plants will lower their flows (spreading of the perturbation from a canal to the next).

3. DISTRIBUTED POWER CONTROL SYSTEM

3.1 Principle

A typical power generation management system relative to a run of the river power plant is based on the following synoptic (Fig. 4).

Fig. 4. Typical power generation management system for a run of the river power plant

The conversion of flow setpoint to power setpoint for each generating unit is done using a flow/power abacus, introducing an additional gap from the initial power demand. Nevertheless, this conversion is needed since units can only receive power setpoints when participating to Ancillary Services.

Preprints of the 18th IFAC World CongressMilano (Italy) August 28 - September 2, 2011

4961

To achieve our purpose of power generation program following, the idea was to complete the current scheme with a Distributed Power Controller (DPC). This control system aims at automating the behaviour of an on-watch operator, who would naturally increase the flow of all plants in a line when it produces less than the demand.

The DPC consists of a numerical PI controller, implemented in the supervisor, which calculates the suited flow adjustment in order to obtain the power target for the line (cf. Fig. 5). The flow adjustment is sent to all power plants, in compliance with the “flow equality” principle.

Resetting the adjustment, the operator can turn back to the current control, when flow management has priority over electrical power (during a flood, for example).

Fig. 5. Distributed power control system

3.2 Simple illustration

This design permits to take advantage of mutual regulating efforts from each plant in the line. As an illustration, let’s consider a line of 5 plants, where 1 m3/s give 0,5 MW (Fig. 6.).

Fig. 6. An illustration of power control with DPC upon an hydraulic perturbation

If the canal between plant n°3 and plant n°4 is upon a withdrawal of 25 m3/s, the water level control system makes plants n°4 and n°5 lower their flows, causing a 25 MW gap.

To compensate this power deviation, DPC will calculate a +10 m3/s adjustment to each plant in order to re-establish the target power setpoint.

The flow adjustment only impacts the line upstream reservoir whose operating capacity (~ 1 hm3) is sufficient. As a consequence, the level evolution will be slightly different from the expected one (obtained if the line strictly follows its flow production program).

3.3 Controller tuning

The system to control is unusual ; it is a computer process: the variable to control is the sum of setpoint calculated by industrial computers equipping power plants. To tune the parameters of PI controller in DPC, the system could be represented by a simple model: a variable gain (to represent flow to power conversion) submitted to an unknown communication delay. The minimum and maximum slope of the flow/power abacus allows us to estimate the gain. The controller is then tuned to present a delay margin of 3.Te (with Te the sample time of the supervisor), and to respect the dynamic criterion for Ancillary Services.

4. SIMULATION RESULTS

A simulator of all hydro power-plants along Durance river has been created, using Matlab-Simulink, to prove the feasibility of DPC, based on real scenarios.

The next graph (Fig. 7) shows the efficiency of the distributed power control system, for the fifth line of Durance, for 3 hours. The closed-loop system behaves like a 1st order system with a time constant of 100 s as required by the TSO.

Fig. 7. Tracking performance simulation with/without

DPC on the 5th line of Durance

The blue solid line is the actual power setpoint obtained with the current control system: it undergoes the wavering of withdrawals, causing up to a 50 MW gap with the target power setpoint (black dashed line). The red solid line is the effective power setpoint with use of DPC, following the generation program, regardless of hydraulic perturbations.

Preprints of the 18th IFAC World CongressMilano (Italy) August 28 - September 2, 2011

4962

The impact on line upstream reservoir level for the same scenario can be seen on Fig. 8. Since the effective power setpoint (without DPC) was too low compare to the generation program, the flow adjustment is positive (rise of flows along the line), therefore the reservoir level with DPC (red solid line) is lower than the expected one (blue solid line): the reservoir has slightly been emptied. Yet the difference is acceptable (15 cm).

Fig. 8. Impact on reservoir level with/without DPC

on the 5th line of Durance

Similar results have been observed when lines participate to power-frequency control.

5. CONCLUSION

A multi-objective control system was developed for cascaded hydro power plants. The control technique do not use sophisticated control law which implies long tuning tests. On contrary, the control system designed presents a lot of advantages for an industrial point of view:

- suited for power plants upgrading or new project,

- easily implementable in a remote control system,

- simple architecture (distributed power control centralized in the supervisor and local water level controllers),

- simple control law (PI controllers).

This effective power management solution is planned to be installed in the Durance-Verdon framework soon. Without disturbing the existing water level control in canals, the solution would enables to guarantee the generation program trajectory following with a very good accuracy and to participate to power-frequency Ancillary Services with the dynamic performance required.

REFERENCES

Dang Van Mien H., Automatisation des systèmes industriels, Eyrolles, 1999.

Dumur D., Libaux A., Boucher P., Robust RST control for Basse-Isère run-of-river cascaded hydro-electric plants, Proc. IEEE Internat. Conf. Control Applications, Mexico City, 2001.

Georges D., Litrico X., Automatique pour la gestion des ressources en eau, Hermès, 2002.

Join C., Robert G., Fliess M., Model-free water level control for hydroelectric power plants, IFAC Conf. Control Methodologies Technology Energy Efficiency, Vilamoura, 2010.

Join C., Robert G., Fliess M., Vers une commande sans modèle pour aménagements hydroélectriques en cascade, Actes 6ème Conf. Internat. Francoph. Automatique, Nancy, 2010.

Zhuan X., Xia X., Models and control methodologies in open water flow dynamics: A survey, Proc. IEEE Africon, Windhoek, Namibia, 2007.

Preprints of the 18th IFAC World CongressMilano (Italy) August 28 - September 2, 2011

4963