Exploring Energy Storage Combinations & Optimal Control Schemes

[Liferider]

I'm currently working on combining energy storage systems like flywheels and batteries to balance consumption/regeneration. I've been looking at using an optimal control scheme so that a cost function can be tailored to our wishes.

I'm curious about what other people in this field have been working on but it's hard to find papers that isn't behind pay walls. Any links/tips to research material is highly appreciated.

 

[Bystander]

I'm currently working on combining energy storage systems like flywheels and batteries to balance consumption/regeneration. I've been looking at using an optimal control scheme so that a cost function can be tailored to our wishes.

"Optimal" and "energy storage systems" and "balance consumption/regeneration" are, or have been so far, mutually exclusive terms as far as tailoring "cost functions" for anything other than "pulling 'green' wool over suckers' eyes." Can you be a bit more specific as far as your "consumption/regeneration" application? Without of course letting too many cats out of the bag?

 

[magoo]

I would suggest your looking at some of the studies that have been done on pumped hydro projects. Considering the cost of pumping and generating at a pumped hydro installation would involve understanding the cost functions during both operating modes such that you can formulate an optimum strategy based on the system load duration curve for which it is being employed. In other words, how many hours should be used for pumping and for generating. Looking at some of this might help you better understand Bystanders comments.

Duke Energy, Dominion Energy and others have some fairly large (2000 - 3000 MW) pumped hydro installations that should help your study.

Although this may not seem as exotic as batteries and flywheels and micro grids, it will reveal to you the significant parameters involved with distributed generation.

 

[Liferider]

Thank you Magoo. My question wasn't meant to be a "What is 2 + 2" question and probably seemed quite vague.. Bystander, let me elaborate some more.

One or more generator (they generate using fossil fuels) sets are supplying power to a bus where there are components that can both consume and regenerate, their power curve can sometimes be regular and predictable but also they can be irregular at times. Now, let's introduce a flywheel and a battery to buffer energy for us. The flywheel is excellent at compensating fast and for high power peaks, while the battery has a huge energy bank, but relatively low power capability.

Now, there is an associated cost of keeping the flywheel charged, at least we have wear and tear, this cost is a function of speed and time. While using the battery it has a cost (reducing lifetime) when you have a large depth of discharge, DoD, especially when combined with high power input/output. But, there is also a cost associated with not compensating what's happening on the common bus because that will cause unwanted effects on the generator sets.. Generators like to operate at a continuous point with highest efficiency. Also, a very high cost is placed on the need to start a new generator if consumption becomes too large.

Now, how to find a good compromise between bus compensation using flywheel and battery (we control power input/output) and the cost of using them?

I hypothesize that it is possible to form a cost functional which we can solve using eg Quadratic Programming. I am wondering how optimal control has been used in the power grid field. I know for example that flywheels have been used to smooth out peaks on a grid, but that is a straight forward control problem.

 

[Liferider]

It would be a cost function that we minimize at time intervals where we look back in time on how the power consumption has been. So let's say that we minimize each second based on measurments obtained over the last 10 minutes say.

 

[russ_watters]

I would probably try to model this with Excel, since you can easily manipulate the dozen or so variables you need and you can use real-world data on the bus's performance as the main input (speed and acceleration). You'll need to data log a real bus for that input data.

[edit] By the way, I made that sound easy but as @anorlunda's post indicates, in the real world these models are actually very complicated. Depending on what your actual goal is here, it may be worth it to pay for some papers (and read some textbooks). Perhaps some patents, too. But a lot of this information is going to be proprietary.

 

[anorlunda]

If you are searching for literature use the terms "economic dispatch" and "security constrained economic dispatch". Yes there are papers behind firewalls, but there are numerous textbooks that are not.

In the old days, the optimum solution was simple, we simple dispatched according to equal incremental costs. That was simple and effective. Today, we optimize systems of linear programming using the Simplex Algorithm. Custom software sets up the equations (on the order of 250000 simultaneous equations and constraints) and commercial solver engines calculate the optimum.

The full problem includes multiple optimum solutions looking forward in time different periods. For example, thermal power plants need to know if they will be needed tomorrow so that the appropriate staff can be told to come to work. Getting them started, warmed up and ready takes 4-18 hours. Therefore, we make one optimum plan for a whole day 24 hours in advance. Then in the actual day, once every hour we make a new optimum solution for the deviations between reality and the day-ahead solution. Then every 5 minutes we make another optimum solution for the deviations between reality and the hour-ahead solution. The actual closed-loop control uses the 5 minute optimum solution, but feedbacks and control actions are based on frequency and power flows only; no more economics.

In all those cases, the actual capabilities of each component must be modeled including startup/shutdown times, ramp rates, weather (for solar/wind), and capacity limits (batteries, flywheels, hydro). In addition, the capacity of the transmission system to securely transport power from every point A to every other point B is modeled. In some cases, laws and regulations must also be modeled.

Of course, for the purposes of learning you can start much simpler. But I would describe it as an optimization problem, not optimal control.

Good luck

You may enjoy this PF Insights article. https://www.physicsforums.com/insights/what-happens-when-you-flip-the-light-switch/

 

[donpacino]

Are you trying to decide whether to buy flywheels vs batteries, or optimize their use assuming both of them are in your system?

if it is the former, an easy way to help you make a decision is to make a decision matrix. assign all the parameters (cost, efficiency, maximum storage, recurring costs, lifetime, etc)a weight, the higher the weight the more important it is to you. Then go through each option and assign a score. Multiply the scores by the weights and youll get a total score for both options. The higher the score the better the choice is.

example below
http://deseng.ryerson.ca/dokuwiki/design:weighted_decision_matrix