44923018_sThe possibility of an autonomous flat heating system, i.e. being able to set the desired indoor temperature independently of the residents in the other flats in the building and to pay on the basis of the actual energy consumption of each flat, is a fundamental necessity to achieve both a higher level of comfort and the freedom to drastically decrease consumption of heating fuel. Heating conversion systems, in fact, are aimed at converting old, centralized plants that supplied flats, where it was not possible to set an independent and known local temperature (in °C or F) and where heating costs were allocated on the unfair basis of the flats floor areas. Allocating heating costs based on the real consumption of each flat is not just fair, it is also a way to naturally encourage residents to adopt virtuous behavior in using the heating system: the more you heat, the more you pay; if you switch off the heating system when you are out, you will pay less. And, incredibly, this makes it possible to increase the thermal comfort because, when there is no local temperature control, the flats closer to the heating source are generally overheated while the ones farthest away are cold.

Measuring thermal energy transfers, the “heating”, is difficult; it cannot be done with a single sensor because it is not a fundamental physical quantity, it is derived from others. Particularly, the transfer of thermal energy through a heating body, such as a radiator, having, as the means of transportation, a heat-bearing fluid medium such as water, is a function of the difference of the inlet and outlet temperatures and the flow rate of the medium through the radiator i.e. the quantity of medium that passes in a unit of time (liters per second, for instance). Temperature (particularly with electronic sensors) and flow rate measurements are quite difficult; achieving high levels of accuracy is not trivial. The physical relationship representing the thermal energy released to the indoor environment (Q) through the radiator involves the three physical quantities (Ti inlet temperature, To outlet temperature, q flow rate) in this way (where c is the specific heat capacity of the fluid medium):


Q = c  *  (Ti – To)  *  q


therefore, the errors in the temperature and the flow rate are additive when determining Q. For this reason, accuracies from 2% to 5% when measuring thermal energy transfers are considered good and almost good respectively.


HEAT METERThe instrument that should ideally be used for this type of measurement is vulgarly called a heat meter, or calorimeter. It is the ideal device because it has three probes that measure the two temperatures and the instant flow rate. Its processor reads the signals from the three probes and combines them to retrieve thermal power and thermal energy in a desired time range (for instance, the current winter season or the previous ones independently). Nevertheless, this device cannot be used in buildings where the centralized heating hydraulic distribution circuit is of a vertical type. These circuits are characterized by the usage of pairs of warm water (outgoing) and colder water (returning) pipes, up the height of the building for every radiator on the same floor. In other words, every radiator on the same floor, sometimes just two of them, is connected to a distinct pair of vertical pipes. This type of fittings was built up to the ’70s and ’80s, in some countries even later, and it requires every single radiator to be adapted to transform the heating system from a centralized to a functionally “autonomous” one. The two basic functions necessary for functional autonomy are: thermostatic or electronic valves for each radiator, to set the desired indoor temperature independently of the other flats, and a measurement device for each radiator to allocate the heating expenses on the basis of the actual energy consumption in each flat.

On the other hand, the modern heating plant context is quite different. These heating systems use a so-called horizontal hydraulic distribution circuit, characterized by a single couple of vertical pipes for each apartment and a star or loop local horizontal circuit for each flat, connected to the only vertical pair of pipes. With these modern, centralized heating systems, it is convenient and easy to install a separate meter in each flat at the junction of the vertical pipes and the local, horizontal circuit. On the contrary, heat meters are very inconvenient and difficult to install on every radiator in vertical distribution circuits, for the following reasons:

  • they have a considerable cost and this cost is multiplied for the number of radiators since, in this case, every radiator needs a meter
  • the flow rate sensor of the heat meter is bulky and, physically, there is no room to mount it between the radiator and the wall from which the inlet pipe comes out
  • the sensor, the simplest type of which is mechanic (similar to a water meter), have to be changed every 6/7 years making it necessary to replace all the calorimeters with the same frequency.
 heat meters for horizontal distribution circuits  heat cost allocators for vertical distribution circuits

Modern horizontal horizontal distribution heating systems, it is possible to use heat meters.

Old Vertical distribution heating systems, it is not possibl eto use heat meters, heat cost allocators are used.


This inability to use heat meters in a vertical central heating system confirms the so-called heat cost allocators as, essentially, the only technology for measuring heat usage in these cases for more than a century. In fact, it is said that in 1908 a company in northern Europe invented the first heat cost allocator, using an ampoule with a graduated scale and filled with colored water. The water evaporated more or less rapidly as a function of the temperature of the radiator on which the allocator was placed; at the end of the winter season a company employee manually noted the level of the liquid in each allocator in the building to make the proper calculations and determine how to allocate the heating costs for each flat. Obviously every 3/5 years, the allocator was replaced because the ampoule became empty. The physical principle has not changed since then, what has changed the most is the means by which the principle is realized. Today heat cost allocators are electronic devices with two temperature sensors, a battery, electronics and a small LCD display.

Regretfully, despite the promises of heat cost allocator manufacturers and commercials, these devices are inaccurate because they do not directly measure the physical quantities involved in the thermal exchange; they measure physical quantities related to the effects of the thermal exchange, such as the radiator surface temperature and the surrounding air temperature. This introduces unknowns, because the same effects on the radiators and on the indoor environment could be obtained in several different ways. An exhaustive explanation of all the drawbacks of heat cost allocators can be read in this article.


ACCURACIES COMPARISON TABLESummarizing, there are two main measuring technologies: the heat meters that are the ideal instruments for measuring thermal energy transfers (with errors between 2.5% and 5%), but they cannot be used for the functional conversion of old, centralized heating plants with vertical distribution hydraulic circuits. In these cases, only heat cost allocators can be used (with errors between 9% and 40%) and they are the technology currently applied to approximately 150 million radiators. The first technology, the heat meters, really measures the energy utilized, the physical quantity of which is the kWh. The other, the heat cost allocator, does not measure any physical quantity; it measures something dimensionless that is called “units”. The sum of the units of all the heat cost allocators in the building is considered equal to the heating expenses for the whole building; these expenses are then proportionally allocated on the basis of the sum of the units of the heat cost allocator in each flat.


In 2008, a small high tech startup was established at the I3P incubator, supported at the beginning by two departments of the Politecnico di Torino, one of the most important Italian engineering schools (the Energy and Automatic Controls and Computer Science departments), and later the Italian National Metrological Research Institute (INRIM). The aim was extremely challenging: “Realize a wireless, remote controllable, full Internet of Things system to convert old centralized heating plants into autonomous ones, performing heating allocation in kWh per each radiator without using heat meters and with an accuracy as close as possible to the direct one performed by the same heat meters.” Basically, have a heat meter, without installing a heat meter. The physics and technological idea was simple to imagine, but extremely hard to realize: apply the concept of soft sensor, used in other engineering fields, in a system that could have hundreds to thousands possible sensors. A soft sensor (sometimes referred to by the misleading term of virtual sensor) is a complex mathematical model, representing a physical system or a physical phenomenon. The more faithful the model is, the more accurate are the soft sensor’s predictions of the phenomena that actually happen. The soft sensor math model generally takes a reduced number of measurements from physical probes and gives as output a measurement from one sensor or more that is/are not physically installed. When it is correctly realized, the soft sensor technique is extremely robust and accurate, so much so that it is used in many safety critical applications (automotive, railways, avionics, etc.). Generally, soft sensors are used when there are critical working conditions for every possible tangible probe (strong vibrations and high temperatures for instance) or when a tangible probe does not exist to retrieve that specific physical quantity (velocity vector for a vehicle to be used for ESP – cars Electronic Stability Program – systems for instance). In the specific case of heat cost allocation, the EcoThermo soft sensor system requires as tangible probe input just the global flow rate of warm water that circulates in the heating plant, passing through the heating source (heater) and the circulation pump. Since this is the only measure coming from a tangible sensor used by the EcoThermo soft sensor, it must be very accurate: it uses an electromagnetic flow rate sensor with 0.1% accuracy. The output of this soft sensor is the flow rate that circulates at every moment in every radiator in the building with an open valve. In this way EcoThermo gets the flow rate of the heat-bearing fluid medium (warm water) passing through each radiator without installing a tangible flow rate probe for each radiator.


SLIDE - METROLOGICAL SOLUTION - THE SOFT SENSOR BENEFITSThis fixes the main problems of mounting a heat meter on each radiator: the absence of enough room to mount the heat meter’s flow rate probe of the heat meter, the cost of the heat meter and the necessity of replacing the flow rate probe every 6/7 years. Two other set of measures are necessary for each radiator thermal energy transfer measurement: the inlet and outlet water temperatures. The electronic wireless valve of EcoThermo manages these measurements with two temperature probes clipped on the inlet and outlet pipes of the radiator. The temperatures, measured with tangible sensors, and the accurately estimated flow rates of each radiator are then elaborated by a remote central server, where the soft sensor software runs, to determine the energy absorbed by every radiator in the building and to allocate the heating costs for every user.

To estimate the flow passing through each radiator accurately, the EcoThermo soft sensor models the entire heating system, not just the radiators, as is done by the heat cost allocators. In this way, EcoThermo knows at every moment what flow rate is passing in every hydraulic distribution circuit branch down to each radiator. This information can be also used for diagnostic purposes: if there is an obstruction in a pipe on the distribution circuit, the EcoThermo soft sensor can indicate in which branch of the circuit this obstruction is located, limiting the intervention of the workers to a specific range of the building.


The soft sensor of EcoThermo is not a simple math model because of the extent of the pipe network, that characterizes even a small building, and because of the hydraulic physical phenomena represented with non-linear equations. However, nowadays, computers and servers have enough computational power to solve even these mathematical problems.  But another problem needed to be solved and it was even more challenging: the warm water hydraulic distribution circuit is different in every building, how could the EcoThermo soft sensor math model be so flexible to adapt itself to all possible cases? And how could it possibly retrieve the physical constants that represent the behavior of each hydraulic circuit branch of every heating system, without direct access to each pipe and hydraulic component (which are mostly located inside the walls) for inspections and direct measurements of them? These two great conceptual difficulties became two further strengths of EcoThermo, once they had been solved. The first question is resolved with an intuitive graphical drag-and-drop tool to sketch the topologic schema of the hydraulic circuit. This takes from a few minutes to a few hours as maximum and can be done on site by the installers or remotely, as back office activity, by a skilled technician. This activity reduces the total time of installing heat cost allocators, for which it is necessary to write down the dimensions, type and characteristics of every radiator and calculate or retrieve some configuration parameters. The saved time is 15 minutes to 20 minutes per radiator, also considering possible back office activities for the data check done by just the most serious heat cost allocator companies. It means from 1 man/weeks to 3 man/weeks saved per installation, depending on the building size. The second issue is solved with a completely automated operation to retrieve all the physical characteristics of the hydraulic circuit. This operation, that for simple and single devices could called calibration, for such a complex soft sensor is named math identification procedure by those in the automatic control engineering field. This procedure determines the constants of the fluidic equation of every branch of the hydraulic distribution circuit, stimulating the hydraulic circuit by setting the circulation pump on different working points and opening and closing the radiator valves in a precise sequence. As a result, this approach permits a very precise representation of the fluid behavior of each part of the heating system and it prevents any mistakes in data retrieval or configuring and determining the parameters, which are done statistically by the heat cost allocator installers. The potential for human error is removed, because human intervention is not allowed in the metrological system configuration.


CASSINO - HEATING PLANT AND MEASURAMENT SCHEMAAnother advantage of EcoThermo is the metrological traceability that essentially does not exist, or is very hard to implement, for heat cost allocators. Traceability is the basis of every metering device and makes it possible to compare and calibrate a measuring instrument with a more accurate one. At the vertex of the pyramid are the national standards, these are unique instruments created and maintained by the national research institutes in each country, such as INRIM for Italy. Since it is not possible to install other types of heating metering devices in a normal building with a vertical distribution hydraulic circuit, the users, in case of doubts, cannot ask for a verification of the accuracy of the measurement process. The measurement becomes an act of faith, the opposite of a scientific procedure. In the few cases, where, for experimental reasons, a building has been deeply modified by adding a heat meter for each radiator or each junction to vertical pipes as a higher accuracy instrument in comparison to heat cost allocators, such as the Cassino University experiment, typically errors higher than 9% and maximum errors of 40% have been detected. Since EcoThermo refers to the three physical quantities directly involved in the thermal energy exchange, the two temperatures can be compared with one from a temporarily installed high accuracy thermometer while the flow rates can be measured with provisory external ultrasound flow meters.




The soft sensor based metrological innovation of EcoThermo is disruptive to traditional allocators, therefore it needed an indisputable validation. That is why the world’s first thermal-fluid­-dynamic mockup has been designed and realized at the National Metrological Research Institute of Italy. The aim of this unique test-bed environment was two-fold: to validate robust accuracy in a real use case, not in an ideal test carried out in very precise conditions on a single metering device, such as has been done by the reference industry for decades, to hide the disputable accuracy performances of the heat cost allocators. Second, to investigate physical phenomena and increase the accuracy of the soft sensor and the robustness of the entire procedure.





Even the first three months of experimentation have been incredibly satisfactory: EcoThermo has reached a first accuracy result very close to that of a heat meter, with average errors between 3% and 5% and, in a very few cases on single radiators, with maximum errors around 10%. Some second order phenomena have also been discovered that could further increase quality of the whole measurement system once they have been represented in the math model of the soft sensor. Creation of the TFD Mockup has been co-financed by European (FP7) and Regional (FinPiemonte) funds and cost around 500k euro, including design, realization and experimentation.


VOLUMETRIC FLOW METER - LIMESCALE FOMRATIONFinally, two curious anecdotes.  Every time that a knowledge paradigm is changed, above all in science and technology, a conservative, sometimes badly interested, part of the community, in notorious cases of the entire society, reacts in a strong, absurd, obstinate, sometimes denigrating way. As an extreme example, think of Galileo Galilei whose astronomical observations changed the earthcentric conception of the universe to the heliocentric one and the reactions to him from other “scientists” and the Catholic Church. EcoThermo, above all during the first conceptual and R&D phase, experienced a similar reaction, obviously at a much smaller scale, from some “technicians”, who tried to spread the idea that a tangible sensor, that actually cannot be installed in the use cases that interest EcoThermo and heat cost allocators, is more reliable than a “software-based” math model. That is a contradiction in itself. First, because it is well known that, in many contexts, tangible sensors cannot be applied or simply do not exist to measure that specific physical quantity. Soft sensors are used in many safety critical applications and machines for which people safety could be at risk if they were not robust and reliable. Even more, what these people forget is that the majority part of tangible sensors is also based on a math model. And this math model is their weakness. In the specific and very typical case of the EcoThermo metrological innovation, one of the possible types of tangible flow probe of a heat meter, the device that cannot be installed in vertical distribution centralized heating systems, is a mechanical one: a turbine spins, every loop is equal to a certain amount of water that has been passed through. In this case there is a math model that correlates the thrust of the water to the turbine and the consequent number of loops of the turbine to the quantity of water that has passed through the probe. The math model is codified with a very fine and fixed gears-down mechanism that is moved by the turbine at one end and that creates magnetic or electric pulses at the other end (the same tangible probe is used also as a water meter that, on the opposite end of the turbine, has summing rotating indicators such as the one used in an odometer).


Well, since after some years limescale and sand deposits change the physics of the turbine that experiences more friction, while the math model codified in the gear-down mechanism remains the same, the sensor becomes inaccurate and must be changed. The math model of the tangible sensor becomes its weakness because the physical part of the sensor changes over time. Pure soft sensors cannot experience this problem, so becoming more robust than tangible ones.


The other anecdote. Years ago it was proposed to one of the big companies manufacturing and selling heat cost allocators, based in north Europe, that it should participate in a project to commercially offer both technologies: their traditional one as an entry product and the innovative one as top-of-the-line system with a much higher market position. The answer from the company astonished the Italian innovators that proposed it: “We like the IoT part of EcoThermo and the possibility of increasing heating savings from 2 to 2.5 times and remotely controlling the system. For the metrological innovation, if we become a shareholder of yours we will ask for 51% of the startup equity, we will pay for the international patent maintenance, but we will not allow you to finish and commercialize that metrological innovation. We have invested the equivalent of hundreds of millions of euros worldwide in marketing to explain how accurate the heat cost allocators are. If you want to succeed, you will have to explain how much more accurate your system is. This will be disruptive for us, even if EcoThermo is a product of ours.” The proposers had a shock hearing these words. The first was a cultural shock: as Italians, they were not used to such explicit declarations. They were positively surprised that a leading company would compliment and show interest in the IoT infrastructure and remote controlling functions. Also positively surprised that the metrological innovation was considered such a dangerous solution. And happy at such a sincere declaration. But this shock was suddenly overtaken by one 10 times bigger: the sincerity for the entrepreneurship proposal was incredibly small compared to the misleading marketing that this, and probably all other, sector industries have promoted for decades declaring a high accuracy that was not true. And even more, they still think that it is better to invest in misleading marketing campaigns than to innovate and improve their products and services.