pond level control

I am going to be controlling the level of a pond of volume 20,000m3. the flow in to the pond is unknown and changes up and down between 1 and 40 m3/s and the control is with a pipe out of the pond which can take a maximum of 10m3/s.

Usually the tank level control I do is on a much smaller volume, so changes in level happen much more quickly and a PID can be set up.

The goal of the control is to take as much water as possible out of the pond without lowering the level below a certain limit.

I am concerned the PID will not be able to regulate the steady state well and respond to the big shocks (floods) that will occur or that tuning on such a large volume might be so slow that it could take a very long time.

Does anyone have any experience with "large tanks" that may be relevant?

US Army Corp of Engineers??

Do you have to maintain a certain level always.

Since your outflow is not a controlled item how about a simple overflow for where you want the water AND another say six inches above sized for max flow ie the surges when rain etc?

Dan Bentler

Why use a PID at all?

The Goal is to maintain minimum level so why not use bang bang with some deadband?

I agree with Michael G. PID is poorly suited to very slow response systems. Unless there is an overwhelming reason to try to maintain constant level, a simple deadband control is sufficient. If you have variable speed pumps you can have several setpoints with different pump speeds for each, so that if the level is a little high the pump flow rate is low, and if it is very high the pump is at max speed.

Simple is better!

it is a hydro plant so the outflow is very finely controlled. the exact head pond level doesn't matter as long as it is within the allowable range, so there is a maximum level and a minimum level. If the level is over the maximum all the generators should be at 100% power, and if it is below the minimum they will all be shut off.

The rate at which the volume extracted can be changed is limited to about 25% per hour so bang bang control doesn't exactly fit.

Definitely agree that simple is better but a deadband may be too slow to allow the generators to follow floods up and back off quickly when melt shuts off quickly at night.

The flow in can be guessed from the change in level and the flow out is known so it should be possible to match flow in and flow out and have a constant level.

It isn't clear what magnitude the pond level varies over. Is the minimum to maximum pond level 1cm, 10cm, 100cm, 1m, or 10m?

the maximum change in pond level should be about 1m. the change in volume for 1m change in level is 8000m3 and it can be assumed that is a linear relationship.

Von

For sure I would increase the capacity of the overflow pipe. Given numbers you provided
at 40 m3/sec you are adding water and increasing volume by .2% per second or 12% per minute. If this is continued downspour at this rate for an hour you have overfilled your dam -- You did not give number for volume thru turbines.

From the numbers I would not worry too much about PLC but in rainy weather or snow melt I would for sure have a good operator on hand to put more units on line to prevent overfill.

As an aside here in Washington state windpower is going to be turned off since spring melt is so high and dams are forced to run water thru turbines to prevent fish kill. Wind power folks are screaming bloody murder and betrayal of course.

Dan Bentler

>The flow in can be guessed from the change in level and the flow out is known so it should be possible to match flow in and flow out and have a constant level.

You are proposing regulating the flow out to match the flow in (I'm not how a varying flow rate affects the power production; isn't the pond outlet some sort of gate valve? not pumped, is it?), so using a level sensor to get a level process variable over a full meter seems feasible with PID.

The change in level must be fairly slow, as your initial post indicates, so any integral (reset) factor would have to fairly minimal (large value for 'minutes', very low value for 'repeats/minute'), otherwise you'll face reset windup, where the output gets driven full scale (100% or zero %, depending on which way the error is) and gets 'stuck' there because the PV has to cross the SP (zero error) before reset term backs off.

If you use any derivative factor, you want to assure that weather induced waves and turbulence don't cause exaggerated output swings due to PV oscillation due to turbulence. Is any part of the pond sheltered with a 'sea wall' so that weather induced waves and turbulence can be damped down in order to get a level reading minus the turbulence? Yes, you can also filter or damp an input signal.

For level, I'd probably tend to one of the submersible head pressure transmitters, because they're so reliable.

If the pond level exceeds the maximum, water spills over a weir and this is fine, as long as the generators are at 100% output.

the level sensing is by pressure transducer with mm resolution and I am not concerned about short period wind waves affecting the level measurement. It is definitely possible to set up a long period oscillation in the headpond which would be refelected in the measurements though any control loop will be much too slow to be unstable.

the pond outlet is a penstock, and at the bottom of the penstock are turbines, the flow in to the turbines is precisely controlled but must be changed slowly (at least an hour from 0% to 100%).

If there is varying flow in to the headpond then ideally power out of the turbine is adjusted to match it. any mismatch results in changing headpond level which is the signal to change the turbine output.

The problem with maintaining an exact set head pond level with PID is that first the level is too low so the turbine backs off to allow the level to rise. Then the turbine flow has to be increased to match the flow in to the pond to keep the level constant at the setpoint. It would be fine if the turbine just matched the flow and stayed at the slightly lower or higher level.

The exact requirements haven't been specified, It should be fine as long as
1) if water is spilling over the weir, turbine flow is 100%.
2) if the water level reaches the minimum level, the turbine output is also the minimum.

This seems to fit a droop characteristic, where output changes from 0-100% depending on the level.

Alternatively the PID loop headpond setpoint could be reset to the current headpond level whenever a maxima or minima in the level (indicating flow in = flow out) is reached.

Some other folks have other ideas as well, but I thought I'd canvas the wizards here to get some more

thanks for all the replies

Kiss

V0N_hydro said:

I am concerned the PID will not be able to regulate the steady state well and respond to the big shocks (floods) that will occur or that tuning on such a large volume might be so slow that it could take a very long time.

Click to expand...

If the inflow is four times your max out flow you are in trouble at the start. The controller can't maintain a set point if the in flow is greater than the out flow the controller can generate. The valve will open fully and the integrator will saturate. At this point you are controlling nothing.

I would simply have a proportional band realizing the floods may increase the water level above the proportional band and the valve is wide open.

I would filter the level significantly so the waves on the water does make the valve respond each wave.

Assuming the average depth of the pond is 2m the average surface are is going to be 10000m^2 so at maximum inflow the water level will rise at the rate of (40-10)/10000=0.003m/s. Your water level sensor will see even the fastest in rush of water as noise because it is less that the waves so you must filter over a period of time like one minute. Then you might be able to use a derivative term too so the valve opens up faster when the water level is increasing faster.

Given the constraints, this is not going to be a work of art. Keep it simple.

For large tanks, if there is an outlet flowmeter, I use cascade control, with the inner loop controlling flow, and the outer loop controlling level. The level controller provides a setpoint to the flow controller. It's easy to apply limiting and/or rate-limiting to the flow setpoint, which prevents large disturbances to the outlet flow. The level control can be proportional only, or PI.

Some time ago, I got called back to a job where too agressive tuning of a level controller produced surges they could see 30 miles away.

It seems similar to the tuning of a large float bank (mining) I found using very little gain with quite a lot of reset (intergral) worked best in that application. You really don't want to start opening the valve until you exceed setpoint or you are wasting valuable energy so your integral keeps the valve shut. Once you exceed level it ramps open.

Sometimes P&ID dosen't work that great perhaps you could just have a simple routine like 5% over open valve x%, 10% over open valve xx% and so on.
That way when the level drops to normal you can be sure the valve will be shut.

Roy Matson said:

It seems similar to the tuning of a large float bank (mining) I found using very little gain with quite a lot of reset (intergral) worked best in that application.

Click to expand...

The integrator isn't needed unless controlling to a set point. In VON_hydro's application the inflow can easily outpace the outflow. In this case the integrator would simply wind up.

Sometimes P&ID

Click to expand...

That is a diagram, not a control loop.

dosen't work that great perhaps you could just have a simple routine like 5% over open valve x%, 10% over open valve xx% and so on.
That way when the level drops to normal you can be sure the valve will be shut.

Click to expand...

Yes, that sounds like a simple proportional controller. A SCP block would do.

the PID block I am using PIDP1 in unity pro xl, has a parameter for the maximum output. When the PID loop output is at the maximum, the integrator no longer integrates so as to avoid windup.

When the 'valve' is saturated and is 100% open and no longer able to regulate the pond level, this is not a problem as it means the generators are at 100% output and increased pond level spills over the weir.

The situation to avoid is water spilling over the weir without the generators at 100%.

What is an SCP block?