Pipeline &
Gas Journal
April issue 2000:
Emerging Technologies For Gas Quality Measurement In The 21st Century
by Daniel A. Zimmerman, Reynolds Equipment Company, Garland, TX
“I am pleased to introduce a low-cost, fully integrated, self-contained energy meter that provides all interested parties with real time information of gas measurement details over wide area networks like the Internet through open system protocols across wireless communications systems.
“The device not only detects pressure, temperature and flow to calculate volume, but at the same time provides detection of specific gravity, percentage of inert gas and Btu. These measurements are generated from a single, in-line solid-state primary sensor connected to ultra low-power micro-miniature electronics. Accuracy is better than plus or minus one therm. There are no moving parts.
“The system provides not only the measurement components for determining therms, but can be programmed to use artificial intelligence for incorporating real time dynamic rate structures based upon time of use demand experience. Control and alarm intelligence is available through infinitely variable outputs. The power supply lasts for 25 years and may be thrown in the trash without harming the environment. The system is suitable for use in hazardous atmospheres, requires no recalibration and one size fits all pipe diameters. The price is less than $100. Availability of this new device is the year 2010.”
Well, we’re not quite ready for the new millennium, but the work required to get us there is well under way.
Technical innovations are ushered in by market forces. This statement is no slight against engineers, it is a compliment. Engineers invent; markets employ.
The advent of electronics in measurement was started by engineers but carried forward by the forces of the market. Gas measurement in general is a reflection of this tendency. In 1967, at the now International School of Hydrocarbon Measurement (ISHM) short course in Oklahoma, an old friend since passed introduced a new concept in a “New Ideas In Measurement” session of the school called flow computers. He referred to odd terms such as I/O and transducers. He commented that he knew of no gas company using flow computers for billing purposes.
In 1967, gas quality was determined primarily by calorimeters and gravitometers. These mechanical beasts were combined with mechanical volume and flow recorders to yield energy measurement (therms) through hand calculations off chart recordings. That was the order of the day.
We’ve certainly come a long way. Today chromatographs do the bulk of the work in determining both heating value and physical properties. They are preferred for their precision, accuracy and repeatability. Chromatograph technology continues to improve by modest enhancement but with little realized cost reduction.
Just as the market forces have dictated the widespread implementation of real time electronic flow and volume instruments for installation at ever-smaller meter points, those same forces are at work to put real-time point of use gas quality indicators at those same sales points.
Current technology (chromatographs) is not a cost-effective answer.
In December 1997, Niagara Mohawk Power Corp. (NMGas) awarded Reynolds Equipment Company a contract to develop new technology based upon intellectual property (a patent) controlled by NMGas. The project goal is to introduce a low-cost Btu meter in the year 2000.
The device may be described as an inferential measurement system. It is totally combustionless as it relies upon the use of novel sensor technologies, microprocessors and leading edge mathematics to provide a real-time, online, point of use gas quality instrument. The device yields values for specific gravity, % inert and Btu. Gas volume is easily added to provide therm calculations.
The Btu meter will be battery powered and have a full complement of communications capability. Multiple protocols including MODBUS will be resident. Accuracy of +-2 Btu is realizable and the cost to the industry is projected to be less than $3,000.
The technology defies the conventional wisdom and its success will depend on a change of mindset among the gas measurement community. This is not a chromatograph. It is not a calorimeter or a titrator. It is totally electronic and inferential.
Chromatographs provide comparative measurements of heating potential and gas composition through chemical analysis and combustion with great accuracy. The Btu meter predicts heating value and aspects of gas composition (within a normal range) based on analog electrical inputs with great accuracy. The predictions are more than guesswork; they are grounded in mathematics.
When using statistical methods of regression and correlation in concert with a large number of samples, we are able to construct a very reliable model for predicting just about anything. For example, to guess your weight I am relying on sensory experience. My mind cannot stray from obvious observation. If I spend years collecting honest answers from people, thousands of people of all shapes and sizes, I can begin to establish tendencies that provide me with the confidence to guess any individual’s weight.
If you blindfold me, the job becomes a little more difficult. But if you provide me with certain specific details, (variables), I can begin to plot them along a curve in my mind’s eye. Different variables have different weighted value in helping me formulate my prediction. Some have little or even negative value. Together, the right combination of variables can yield a very precise and repeatable prediction.
For instance, if someone told me your height I could likely guess your weight within 10%, 90% of the time. If I combined this with knowledge of what sex you were I could achieve results approaching 97%. And if I knew your age and penchant for ice cream I could shave off enough uncertainty to guess right 99% of the time. If, in addition, you told me your hair color or inseam or shoe size, I might find that data misleading or extraneous. Lesser-weighted variables may likely diminish a high precision model but could, on the other hand, enhance another with less than acceptable precision.
Nothing, however, is always as it should seem. If I grew up spending all my time at the horse track and in the gym, my observed tendencies may be distorted from the real world. You throw me a curve by exposing me to frequent encounters with horse jockeys and basketball players. This is when my inference-based model shows its shortcomings. It’s especially apparent compared to a positive measurement system like a weight scale. In other words, the regression model works best within a normal population of samples. Move outside of the normal range and the results become unpredictable.
In terms of Btu, a properly configured chromatograph is as accurate at 500 Btu as it is 2,000 Btu and all points in between. The inferential Btu meter model is more specific to normal samples between 950 Btu and 1,100 Btu. Outside of those ranges it doesn’t work as well, if at all.
With a chromatograph, if you took all of the gas components from many heterogeneous samples and plotted them, eventually you would unveil a pattern that might allow you to predict Btu with minimal error based only on a fraction of awareness of the total composition.
This is the assumption that we are relying upon in our system. The gas composition variables we have determined to best fit within our system by weighted contribution (most to least) are:
Thermal Conductivity
Specific Heat
Viscosity x Molecular Weight
The selection process is driven less by chemistry and physics than by the performance and suitability of available electronic sensor technologies. The sensors we have selected respond well to variations in total gas composition and through theoretical and empirical analysis, we infer proportionality to a suitable (best-fit) gas property. The sensor outputs are analog and may be represented as voltage, resistance, capacitance or frequency.
By plotting the results of indicated sensor measurements of a specific mixed gas sample and tying them back to the known Btu and known gas composition values of that sample, and repeating this over and over with different samples, the system “learns” to predict Btus with precision. If you’re thinking artificial intelligence (AI) here you’re right on track. Regression is the cornerstone of AI mathematics. Genetic Algorithms and Neural Networks are sophisticated learning systems that take basic regression curves and use experience to back-propagate and overlay second and third dimensions to enhance precision and repeatability.
In addition to the novel sensors utilized for indicating gas properties, conventional gas pressure and gas temperature transducers are incorporated for correcting to standard Btu.
The sample data taken to construct the regression equations have the potential through AI techniques to improve over time. Theoretically, the higher number of normal samples it is exposed to, the higher degree of accuracy it will attain (neither regression or AI systems are “expert systems”, therefore 100% accuracy will never be achieved).
The sensor inputs to the main circuit are powered by precision micro-supplies and are conditioned through high performance analog to digital converters and then routed to the microprocessor for interpolation, calculation, unit conversion and data processing and storage according to the programmed firmware.
A product does not exist, however, until the various engineered components are integrated into a field-worthy package and a user-friendly, man-machine interface (host software) is developed.
The Btu meter will not be suitable for all applications. Low Btu landfill gas or high Btu imported LNG are examples of gasses outside the “norm” that it excels in. The target for the Btu meter is the typical commercial/industrial end user metering site where gas between 950 Btu and 1,100 Btu is delivered.
Of course, in 1967, the prevailing industry mindset viewed flow computers as curiosities. Recent industry focus has been on detailed analysis of gas compositions (e.g. A.G.A. #8) that only a gas chromatograph can deliver. It is unlikely that gas chromatographs can be reduced enough in cost to meet the demands of the emerging market forces. Some compromises will have to be made if this exciting technology is to be accepted not only by the gas industry measurement community, but by the legal (gas contracts) and government (regulatory) interests as well.
The progress I’ve witnessed first hand gives me the confidence to say that gas measurement advances are certainly in store for the new millennium. P&GJ
Mr. Zimmerman has been with Reynolds Equipment Company since 1982. He is an associate member of the A.G.A. Distribution Measurement and Transmission Measurement Committees. He is a graduate of Austin College in Sherman, TX with a BA in Liberal Arts.
This article was presented at the American Gas Association’s Operations Conference in May 1999.