CRC — Carbon Reducing Controller · white paper v3 (2022, archival)

Archival copy — verbatim. This is the author’s original 2022 white paper (v3), converted from the source .docx without editorial changes: spelling, figure numbering and the unresolved citation markers [1]–[4] (the original has no bibliography) are preserved as-is. The algorithm described here (CRC©) was authored by R. de Ren in 2022 and open-sourced by the author upon joining Waterschap Brabantse Delta; it is the direct ancestor of the EVOLV machineGroupControl node. A 2026 scientific paper referencing this document is published separately on this track.

A Centralized Energy efficient Method for Flow distribution and machine control.Abstract: For decades we have been controlling rotating equipment (1 ex. Pumps, blowers, compressors, engines, turbines, generators, Archimedes-screws,…), further referred to as machines1, in an inefficient way. We use a single process demand such as pressure or temperature to maintain adequate flow and pressures. This is achieved by turning pumps on and off and controlling them in cascade assuming the machines are equally sized.In this way we depend on the original design parameters assumptions and we don’t take advantage of the actual site conditions.We propose a solution where machine control is recalculated each demand change and then looks for the most optimal solution by combining the best efficiency points (BEP) of all available machines on a certain process demand.The energy gain made by this solution is dependent of the available machine setup however estimated on average the CRC© can shave 10% of the global industrial energy consumption. A Centralized Energy efficient Method for Flow distribution and machine control.Abstract: For decades we have been controlling rotating equipment (1 ex. Pumps, blowers, compressors, engines, turbines, generators, Archimedes-screws,…), further referred to as machines1, in an inefficient way. We use a single process demand such as pressure or temperature to maintain adequate flow and pressures. This is achieved by turning pumps on and off and controlling them in cascade assuming the machines are equally sized.In this way we depend on the original design parameters assumptions and we don’t take advantage of the actual site conditions.We propose a solution where machine control is recalculated each demand change and then looks for the most optimal solution by combining the best efficiency points (BEP) of all available machines on a certain process demand.The energy gain made by this solution is dependent of the available machine setup however estimated on average the CRC© can shave 10% of the global industrial energy consumption. A Centralized Energy efficient Method for Flow distribution and machine control.Abstract: For decades we have been controlling rotating equipment (1 ex. Pumps, blowers, compressors, engines, turbines, generators, Archimedes-screws,…), further referred to as machines1, in an inefficient way. We use a single process demand such as pressure or temperature to maintain adequate flow and pressures. This is achieved by turning pumps on and off and controlling them in cascade assuming the machines are equally sized.In this way we depend on the original design parameters assumptions and we don’t take advantage of the actual site conditions.We propose a solution where machine control is recalculated each demand change and then looks for the most optimal solution by combining the best efficiency points (BEP) of all available machines on a certain process demand.The energy gain made by this solution is dependent of the available machine setup however estimated on average the CRC© can shave 10% of the global industrial energy consumption. A Centralized Energy efficient Method for Flow distribution and machine control.Abstract: For decades we have been controlling rotating equipment (1 ex. Pumps, blowers, compressors, engines, turbines, generators, Archimedes-screws,…), further referred to as machines1, in an inefficient way. We use a single process demand such as pressure or temperature to maintain adequate flow and pressures. This is achieved by turning pumps on and off and controlling them in cascade assuming the machines are equally sized.In this way we depend on the original design parameters assumptions and we don’t take advantage of the actual site conditions.We propose a solution where machine control is recalculated each demand change and then looks for the most optimal solution by combining the best efficiency points (BEP) of all available machines on a certain process demand.The energy gain made by this solution is dependent of the available machine setup however estimated on average the CRC© can shave 10% of the global industrial energy consumption.

Introduction

In many industrial applications, the control of multiple parallel machines1 in a single process, is based on a form of cascade control method.

Although there are other forms of control, for example: where an individual machine is limited to use only part of its total capacity to avoid being on the outer edges of its machine curve, they all work according to the same principal. Scale a machine up or down until a certain process demand is achieved.

Given the significant amount of energy machine systems use, even the smallest improvement in energy efficiency is key to reducing the total global energy consumption. By estimation, industrial pumps and compressors alone use about 20% of the worlds energy production. Source : [1],[2],[3],4]Anno 2019 this accounts for approx. 5.2 TWh ( “5,200,000,000,000 kWh” ) source : [1],[2]Fig 2 ( Global energy production / year ) : Applying optimal control can easily shave of 10% of this energy consumption. To put that in perspective:5,200,000,000,000 * 0.1 = 520,000,000,000 kWh savings1 average household in uses approx. 2800 kWh per year [1]. So, achieving a saving of something seemingly trivial as “only” 10% would save the equivalent of 185,714,285 households. That’s more than half of the population of the U.S.

[Native chart in the original: “Produced per year”, series 2000→2019 — not reproduced as image; data preserved in the source docx.]

individual machine curve

In order to understand the solution, we must first understand what a machine efficiency curve is and how it operates.A machine works by converting electrical power to mechanical rotation. The rotation of an impeller, screw, lobes, … increases the pressure or flow in a gas or liquid in order to move it from one point to another in a closed or open system. Although these rotating parts come in many forms and shapes, they are almost all components of which the shape remains the same. Because they are non-dynamic in shape or size we control the vast majority of these machines, if present, with a frequency drive in order to control the rotational speed. Simply put, faster rotations, equal more flow and pressure.This implies that we have a fixed 3D curve for each individual machine telling us how much flow it will produce at a certain speed (RPM) given a certain pressure. For now, we will leave out the density of the liquid or gas and focus only on the efficiency curve for 1 density.In order to understand this better first we will visualize this in a flat 2D curve where we plot the flow (liter per second) over the energy consumption (kWh)

Here we can clearly see that a machine has a sweet spot where it performs best (Between 3290 and 3660 RPM @400mbar). This is the highest point on the curve representing the most flow it can deliver vs the power it requires, the best efficiency point (BEP).

Whenever the pressure increases or decreases the best efficiency point will be different and so will the efficiency curve.Higher delta pressure (pressure losses in the system) equals to more power use per delivered flow unit.

Since the pump is controlled from minimum to maximum speed and the efficiency depends on the resulting pressure in the system by changing variables (ex. Closing or opening of valves), we can conclude that we need the variable pressure in order to precisely predict the flow at a certain speed.

Multiple machine control

Current machine control methodoligy

There are a multitude of machine controllers on the market.These control methods all have the same basic control methodology. They all use a principle called cascade control.The cascade control runs through the available machines by using their preset priority numbers and run from a preset minimum to maximum speed. Depending of the individual program the principle is as follows:Fig 5 : Cascade control principle

The above graph is a simplified representation on how this works. As we need more pressure or flow, we will move right on the horizontal axis until our demand is met and vice versa when we need less pressure or flow.Green, blue and orange represent the individual machine control. As we gradually move right or left on the horizontal axis machines go up or down in RPM to meet the demand.

Using this knowledge and combining it with the individual machine curve examples we had presented in chapter 2 we would have a total cascade efficiency curve. Again, to simplify matters we will take a fixed pressure in order to make it clearer what happens. Keep in mind in reality we cannot know on forehand what the pressure would be in a system because of variables we do not control with the machine controller.Fig 7 : Combined cascade efficiency curve with total flow output on a fixed pressure.

Note: This is quite easy to follow as the curve for each machine is identical. This also means that on the same demand the output will be the same as we are adding flow and power of 2 identical machines which in turn will have identical efficiency. In reality this differs because no 2 machines are truly identical and the pressure varies over the total flow, but again for the sake of simplicity we will assume that they are.

Depending on the level of “intelligence” in the cascade control there is always a noticeable step when there is a machine switch. See cascade step 1 & 2. In most cases the machines are chosen to avoid the step as much as possible. Unfortunately, this is in contrast of choosing the machines to have good overlapping efficiency curves and thus making this step bigger.

“Smart” Cascade

Although we have discussed how the majority of cascade controllers operate, there are a handful cascade controllers that have a smart way of removing the characteristic step.

Because we already established the way how the cascade control works, we will now only focus on the results of a cascade control where the step is removed.

Here our transitions between machine changes are a lot smoother. We can still notice the distinct movement of the efficiency curves of the machines when a machine is started. And we still have the disadvantage of not controlling a machine or multiple machines on their best efficiency points.

This methodology is not easy because on different pressures the steps are different. In a process where the pressure does not deviate a lot this is not such a problem. But when the pressure starts to move up or down the step will be at another moment in time and the controller will still have issues with it.

So, while you are perfectly capable of configuring the most advanced cascade controllers for a specific pressure (mostly nominal pressure) it will never work on multiple pressures.

The smoothness of both the flow and the efficiency curve is highly dependent of the pressure on which you have benchmarked it.

The same accounts for the normal cascade: The algorithm doesn’t control on best efficiency point.

When controlling on a demand flow between 2 machines the algorithm wont smooth out the control. (It will turn on and turn off machines without thinking about the control stability.

No feedback of measurements can be given back to the control. (If a machine degrades over time the control doesn’t know and act on it).

Carbon reducing control methodoligy (CRC©)

If we want to control machines efficiently, even when the process demand is dynamic and unpredictable, we will need to be able to achieve the best efficiency independent of flow or pressure. This is done proposing a novel approach by using the machines efficiency at the momentary pressure and comparing all the possible combinations we can make on that moment to obtain a certain flow.

In order to simplify this and compare this on a single pressure we have plotted the same process demand as we had before in the Cascade control methodology.

Fig 8 : Combined CRC© efficiency curve with total flow output on a fixed pressure.

Here we do not see any steps between machine switches at all. This is because the algorithm knows which machines to activate at exactly the right moment in order to achieve the smooth control.

Process demand:The output flow shall be equal to the demand flow when this is mechanically / physically possible.

Process stability:The stability of the process shall be safeguarded by continuously evaluating the right moment to add or remove a machine from the control without conflicting with the first law.

Energy efficiency:We keep energy efficiency as high as physically possible as long this doesn’t conflict with the first and second law.

Each demand change there is an evaluation made whether to switch on another machine to optimize efficiency without impacting the process.

The efficiency line is kept as high as mathematically possible to ensure the highest energy gain possible with multiple machines.

It can feedback information from available measurements to optimize the data as its being controlled (AI).

Although this is usually working on systems with continuous control (and there is no need for such a feature), it has the possibility to instantly calculate how many machines to control at a specific demand point when the corresponding pressure is known. (There is no feedback loop needed).

Comparison control methods in realtime

Now that we have explained the different controls and their specific advantages and disadvantages lets see how this translates and holds up in a setup mimicking a real process with multiple rotary machines. How our setup is built is described in attachment 1.

The system operates between a pressure of 150 – 450 mbar. The pressure in the system is a direct result of the pressure drop caused by the systems components. No valves were operated during the operation time of the pumps or any other components that can have a direct impact on the pressure.

Cascade algorithm:

In our first run we operated the machines in the most basic cascade logic. This means we did not account for the specific curves of the pumps. We started at 0% control and ended at 100% even though the pump operates between 10 – 90 %.

As you can see this is a terrible way to control the pumps. This is exactly what happens when machines get controlled by a cascade controller that doesn’t know or take into account pump curves. This means the control did not only not correct the gap of 10% where the pump doesn’t move because of the 10 – 90 % limit but also wouldn’t correct the gap where machines would be controlled on a higher pressure and can’t deliver flow at all. For example a pump would be controlled at 20% but he wouldn’t deliver flow because the system pressure is too high.

The first reason for the gap is easily removed by skipping the first 10% and avoiding the last 10%.The second reason for the gap is not removed at all for a cascade control unless you put it as an unchangeable parameter in the source code where to stop and where to begin at a certain pressure. Resulting in so many variables to program that it would be really labor intensive and prone to mistakes. This also result in higher setup costs and specific programming for each individual case. (Not standardized) The way around this is to have individual machines indicate when it is at its maximum rotating force. This gets passed as an input parameter to the casacade algorithm so it will not keep trying to raise its output % because this has no use.

the sake of our analysis (and because we have cheap pumps that do not have any feedback possibilities) we made our cascade algorithm / simulation so that it avoided the gaps in the code. We simply ran 1 cascade control and took away those gaps where the machine doesn’t act useful.

Note the distinct steps when the second and third machine get turned on as predicted in chapter 3.A and illustrated in fig 7. This time it’s not on a prediction, this is feedback from the measured flow and power on a dynamic pressure of the test setup.

CRc algorithm:

To make the comparison clearer we will retain the self-improving part of this algorithm and discuss it in a separate paper. For now, we will be focusing only on the difference between the different types of control. We feed our control with the theoretical pump curves and not allow the algorithm to improve itself for the test.

The control works by using the theoretical curves as a baseline and comparing the best efficiency points of all the machines with each other for each individual process demand. Evaluating the outcome using the 3 laws that govern the CRC© it calculates the most optimal control strategy for all connected machines.

This result also follows the prediction made previously. And as with the cascade, this is over a dynamic pressure.

CRc VS Cascade:

Plotted on top of each other its immediately clear that the CRC© uses the pump curve to its advantage. The pump curve is relatively efficient early on and delivers a lot more flow when aided by other pumps on the same rotational speed.

The absolute minimum and maximum measured total flow should be for both algorithms roughly the same. We did not build in a certain waiting period to average the measurements over for a certain time. This would have decreased the error margin of the measurements. This explains the negative number in the beginning and end of the comparison even though they perform on the same flow and almost on the same control.

Even though there is a slight error margin for the measurements the error is present for both algorithms. There is no doubt that, however the results might slightly deviate negative or positive, the results are overwhelmingly positive in favor of the CRC©..

The total area of efficiency is bigger in the CRC©. than in the “smart” cascade control. The difference between both are not always mathematically the largest feasible in all cases on the total line of control. This is a direct cause of the 3 laws that govern the CRC©.. (See chapter B point 1,2,3)

The CRC©. efficiency line is smoother and more consistent than the cascade control especially true when you compare it to the vast majority of the cascade controllers which do not account for the specific machine curves and pressures. Note that this also leads to a much stabler overall process control.

The “smart” cascade control doesn’t look at efficiency it will turn on a machine when its preset limit is reached.

IMpact on Pump Selection – Future impact of optimal control

The current market and the way companies select machines is still based on the “old” style of machine control. The fact that we have got an algorithm that works independent of machine brand or size can pave the way to select different machines for the job. A selection where not 1 machine needs to be very efficient, but a range of machines need to be more efficient.

A well-chosen setup (without an advanced control) can also save up to 30% in energy reduction when chosen correctly and for the specific process histogram. Add that to the potential gain of a good control and we could save a significant amount of energy and help achieve our goals to help the ever-growing demand of cost-efficient solutions.

Avoiding that this paper becomes too hefty to read we will go a lot deeper in detail on the self-learning part of the algorithm in another paper.

If you have any questions or comments, thoughts, or want to start saving energy right now feel free to contact us.

ATTACHMENT 1 : Pilot setup specs

Provenance converted 2026-07-17 from “CRC - whitepaper v3.docx” (sha256 43011b140f43da05…), python-docx walker, figures embedded verbatim as data-URIs.

Status unpublished white paper (2022); archived as citable source.

Contact R&D-lab · lab.wbd-rd.nl