EP0249531B1 - Method and apparatus for controlling a central-heating system - Google Patents

Description

The present invention relates to a process and a device for regulating central heating for an individual or collective residential building.

• - Asservissement de la température de départ du fluide de chauffage à la température extérieure, suivant une loi dite "courbe de chauffe". Ce procédé ne permet pas de prendre en compte les apports internes de chaleurs, les apports solaires, le vent, les inerties thermiques, qui ont un effet sur les températures effectivement atteintes dans les habitations ; d'autre part, il nécessite un réglage manuel délicat des paramètres définissant la courbe de chauffe. Enfin, le coût d'installation d'une sonde de température à l'extérieur d'une habitation est relativement important.
• - Asservissement de la température de départ du fluide à une température intérieure mesurée ; ce procédé n'a pas les inconvénients précédents, mais peut être perturbé par l'influence de conditions locales et/ou passagères (ouverture de fenêtres, fermeture d'un robinet de radiateur, exposition au soleil, proximité d'une source de chaleur, etc.), de sorte qu'il ne convient pas au chauffage d'immeubles collectifs, et demande certaines précautions en chauffage individuel ; il demande encore une liaison entre la chaufferie, le régulateur et le local où la température est mesurée.
• - Modulation de débit par des robinets thermostatiques équipant les émetteurs ; ce procédé, efficace pour des consignes constantes, ne permet pas à lui seul la programmation horaire ; il peut être couplé à une programmation horaire centralisée de la température de départ ; celle-ci doit être suffisante en période normale pour assurer la satisfaction des besoins, et être choisie en période de chauffage réduit, de sorte que, les robinets thermostatiques n'opérant plus, les températures des pièces approchent au mieux la consigne réduite ; ce résultat est difficile à obtenir lorsque la régulation centrale ne connaît que la température extérieure ou une température intérieure.

The methods employed so far fall into three categories, each having various drawbacks. They are based on the following principles:

• - Controlling the flow temperature of the heating fluid to the outside temperature, according to a law called "heating curve". This process does not take into account the internal heat gains, solar gains, wind, thermal inertia, which have an effect on the temperatures actually reached in homes; on the other hand, it requires a delicate manual adjustment of the parameters defining the heating curve. Finally, the cost of installing a temperature sensor outside a home is relatively high.
• - Control of the fluid flow temperature to a measured interior temperature; this process does not have the above drawbacks, but can be disturbed by the influence of local and / or transient conditions (opening of windows, closing of a radiator valve, exposure to the sun, proximity to a heat source, etc.), so that it is not suitable for heating collective buildings, and requires certain precautions in individual heating; it also requires a connection between the boiler room, the regulator and the room where the temperature is measured.
• - Flow modulation by thermostatic valves fitted to the transmitters; this process, effective for constant setpoints, does not by itself allow time programming; it can be combined with a centralized hourly programming of the flow temperature; this must be sufficient during normal periods to ensure satisfaction of needs, and be chosen during periods of reduced heating, so that, with the thermostatic valves no longer operating, the temperatures of the rooms approach the reduced setpoint at best; this result is difficult to obtain when the central control only knows the outside temperature or an inside temperature.

We know from French patent FR-A-2,542,852 a method of regulating a central heating of a building in which at least once a day the temperature setpoint goes to a reduced value, process in which we measure in permanently two temperatures of the heat transfer fluid (flow temperature and return temperature); the heating control is slaved to a value of the interior temperature estimated by successive approximations from the temperature of the heat transfer fluid and the thermal properties of the heating network.

The implementation of this process leads to certain difficulties. Thus, the internal temperatures obtained are subject to too large fluctuations because the thermal properties of the network change too much with the actions of manual adjustment of the users.

It can also be seen that learning is not obtained with sufficient precision when a small difference exists between the two temperatures of the heat transfer fluid.

The object of the present invention is to implement a method which makes it possible to avoid the above drawbacks.

It also aims to reduce the number and complexity of the adjustment interventions required from the user of the heating system.

These objects are achieved according to the invention by the method as defined by claim 1.

• - La figure 1 montre très schématiquement un dispositif pour la mise en oeuvre du procédé, intégré à une chaudière de chauffage central.
• - La figure 2 montre un diagramme illustrant les variations de température théorique et pratique, ainsi que la puissance de chauffe au cours d'une journée.

Other characteristics and advantages of the invention will appear during the following description of various embodiments given by way of illustration, but in no way limiting. In the attached drawing:

• - Figure 1 shows very schematically a device for implementing the method, integrated into a central heating boiler.
• - Figure 2 shows a diagram illustrating the theoretical and practical temperature variations, as well as the heating power during a day.

We see in Figure 1, a boiler 1, with its burner control 2, its fluid circulator 3. Reference 4 illustrates the fluid flow pipes to the elements for exchange with the building, while reference 5 illustrates the fluid return pipes to the boiler. A regulator 10 is integrated into the boiler 1; it receives at input 11 information from a member 8 for measuring the temperature of the fluid, and has an output 12 for controlling the burner 2.

• Le modèle "réseau" caractérise le réseau de distribution de chauffage en tant qu'échangeur entre le fluide caloporteur et le bâtiment à chauffer. Il s'écrit de façon générale :
• f (0 i₁, 0f, 't1) = 0, avec :
• 0 i₁ = température intérieure moyenne estimée selon le modèle "réseau"
• θf = température mesurée du fluide caloporteur par l'organe 8
• τ₁ = taux de charge, rapport de la puissance à la puissance maximale, connu d'après la commande définie par le régulateur.

The regulator uses a model for estimating the interior temperature of the building, in fact combining two distinct models called the "network" model and the "building" model, respectively:

• The "network" model characterizes the heating distribution network as an exchanger between the heat transfer fluid and the building to be heated. It is generally written:
• f (0 i ₁ , 0f, 't1) = 0, with:
• 0 i ₁ = average interior temperature estimated according to the "network" model
• θf = measured temperature of the heat transfer fluid by the member 8
• τ ₁ = charge rate, ratio of power to maximum power, known from the command defined by the regulator.

_T'i (t) étant une famille de mémoires exponentielles du taux de charge instantané τ(t), basées sur des temps caractéristiques A ti différents:

du, approché par

nGiven the thermal inertia of the network, τ ₁ should not be treated as an instantaneous value, but as an average value over a sufficient time interval, or better as an average value weighted by the inertial response factors of the network, ie for example :

_T 'i (t) being a family of exponential memories of the instantaneous charge rate τ (t), based on different characteristic times A ti:

from, approached by

not

p(i) = 1, ne sont pas connus a priori, mais peuvent être établis assez rapidement par auto-apprentissagei effet, l'estimation de θ i₁ d'après la loi f (0 i₁, θ f₁, τ₁) = 0 ne doit varier que lentement étant donné l'inertie thermique du bâtiment, alors que le taux de charge instantané τ(t) est susceptible de varier rapidement.Response factors p (i), network characteristics, verifying

p (i) = 1, are not known a priori, but can be established fairly quickly by self-learning effect, the estimate of θ i ₁ according to the law f (0 i ₁ , θ f ₁ , τ ₁ ) = 0 should only vary slowly given the building's thermal inertia, while the instantaneous charge rate τ (t) is likely to vary quickly.

_bility . It is therefore advisable to seek the values {P (i)} which give the estimate of θ i ₁ the best stability.

• δ p_i = λ A δτ'_i où λ est une constante, et A l'erreur sur la variation de température en une minute.

The self-learning of p (i) takes place continuously. Each minute the charge rate τ ' _i is stored; if we call δp _i the weak correction to be made to p _i for a variation δτ ' _i corresponding to two successive values of τ' _i :

• δ p _i = λ A δτ ' _i where λ is a constant, and A the error on the temperature variation in one minute.

• θ i₁ = θf - a. τ1, expression dans laquelle intervient un paramètre à identifier a, caractéristique du réseau.

The simplest formulation of the law f would be:

• θ i ₁ = θf - a. τ1, expression in which a parameter to be identified intervenes, characteristic of the network.

• τ1 = p₁ ( θ f- θi) + p₂ ( θ f- θi)², dont on peut déduire θi, connaissant θ f et τ1. Dans toute la suite on utilisera cette formulation pour le modèle"réseau".

Given the significant variations in the exchange coefficients and the flow rate with the temperature levels, this formulation risks being insufficiently precise, so that it may be useful to resort to a slightly more complex model, for example the implicit model:

• τ1 = p ₁ (θ f- θi) + p ₂ (θ f- θi) ² , from which we can deduce θi, knowing θ f and τ1. In the following we will use this formulation for the "network" model.

This model involves two parameters p ₁ and p ₂ . These can also be established by self-learning, taking advantage of the hourly programming, as will be explained below.

Indeed, the regulator can interrupt the heating - generally for several hours - when the setpoint changes from the normal value to the reduced value; τ1 tends to zero, so that θ i becomes well known, because it is close to θ f.

The value obtained constitutes a precise reference, which can be injected into the "building" model (see below), in order to estimate with good precision what was the value of θ i during the change of setpoint, and of deduce a condition from p ₁ and p ₂ .

Similarly, the well-known value of θ i at the end of the heating shutdown period, combined with the "building" model, gives a precise estimate of θ i when the heating has restarted for a short time, for example an hour , while τ 1 and θ f have already reached high values, which gives a new condition on p ₁ and p ₂ . We therefore see that each transition of the setpoint gives information on p ₁ and p ₂ , so that these can be evaluated fairly quickly.

The law f (θi ₁ , θ f, τ ₁ ): 0 can be disturbed by local actions, such as for example limiting actions thermostatic valves. In this case, the temperature of the fluid 0 f is higher at a given charge rate. The application of the law f leads to an overestimation of θ i, and therefore to a reduction in heating. This reduction is useful because the action of thermostatic valves means that the demand for heating is lower than the supply. We can therefore consider that the regulator mainly regulates the average interior temperature of the building, with however a partial adaptation to the demand expressed by the local adjustment actions.

avec :

• . θi₂ (t) = température intérieure moyenne au temps t, selon le modèle "batiment"
• . θi₁ (to) : température intérieure moyenne selon le modèle "réseau", mémorisée depuis l'heure to de fin de la dernière période d'arrêt prolongé du chauffage
• . i₂ (t), τ₂ (to) = mémoire exponentielle du taux de charge instantané τ(t), de temps caractéristique Δto égal à 3 heures environ.
• . _T2 est calculé par :
  
• . p3 = paramètre définissant la sensibilité de la température du bâtiment vis-à-vis du chauffage.
• . g (t), g (to) = fonction de correction horaire de la température intérieure, pour tenir compte de la part prévisible des fluctuations horaires des conditions climatiques et d'occupation ; g (t) est minimal vers 8h, et maximal vers 15h.
• . p4 : paramètre définissant l'amplitude des corrections horaires de température à effectuer, pour le bâtiment considéré.

The "building" model is the second means available to the regulator for estimating the indoor temperature. It characterizes the building as a space, the temperature of which is sensitive to heating inputs, as well as to other more or less known stresses (climate, occupation). It can be written for example:

with:

• . θi ₂ (t) = average interior temperature at time t, according to the "building" model
• . θi ₁ (to): average interior temperature according to the "network" model, memorized since the time to end of the last prolonged period of heating shutdown
• . i ₂ (t), τ ₂ (to) = exponential memory of the instantaneous charge rate τ (t), with characteristic time Δto equal to approximately 3 hours.
• . _T2 is calculated by:
  
• . p3 = parameter defining the sensitivity of the building temperature vis-à-vis the heating.
• . g (t), g (to) = time correction function of the indoor temperature, to take into account the predictable share of hourly fluctuations in weather and occupancy conditions; g (t) is minimum around 8 a.m., and maximum around 3 p.m.
• . p4: parameter defining the amplitude of the hourly temperature corrections to be made, for the building in question.

The formulation of the "building" model is justified by the following considerations: - θ i ₁ (to) constitutes a precise evaluation of the interior temperature at time to, even when the parameters {P (i)} P ₁ , P ₂ are wrong known

The call for an exponential memory of the charge rate, of characteristic time equal to 3 hours, makes it possible to take into account approximately the combined effect of the thermal inertias of the network and of the building.

Experience shows that this formulation generally gives fairly good results. The long-term effect (more than 24 hours) of thermal inertia is not to be taken into account, because we have at least once a day a new precise estimate θ i ₁ (to).

The function g (t) intervenes in a corrective term whose amplitude is of the order of 1 to 2 ° C. It reflects the fact that the outside temperature, the solar gain, the internal gain, are statistically higher during the day than at night. Being a corrective term, we admit on g (t) a certain degree of approximation, allowing the use of a preprogrammed function.

It is possible to define the sensitivity parameter p ₃ by self-learning; in fact, the interior temperature is well known thanks to the "network" model for some time (about two hours) after the heating stops due to each drop in setpoint. It then suffices to compare this temperature with the value given by the "building" model, and to correct p ₃ accordingly.

• p₄ peut être ajusté pour que la différence θ i₁ (15h) - 0 i₂ (15h) des températures données par les deux modèles à 15 heures soit statistiquement de moyenne nulle.

Self-learning of parameter p ₄ is also possible but relatively slow:

• p ₄ can be adjusted so that the difference θ i ₁ (15h) - 0 i ₂ (15h) of the temperatures given by the two models at 15h is statistically of mean zero.

• Le modèle "réseau" est particulièrement précis durant les périodes d'arrêt ou de réduction du chauffage ; il l'est moins lorsque la température du fluide est élevée, et dans les régimes transitoires.

The controller can use any combination of the "building" and "network" models at any time to access the best possible estimate of the interior temperature θ i ₃ , and act on the heating control accordingly. This combination must give greater weight to the model that is known to be the most precise under the operating conditions considered:

• The "network" model is particularly precise during periods of shutdown or reduction in heating; it is less so when the temperature of the fluid is high, and in transient regimes.

The "building" model remains precise when the temperature of the fluid is high, provided that the reference 0 i ₁ (to) is sufficiently recent. We can then limit the role of the "network" model to the detection of indoor temperature excursions linked to unpredictable changes in climatic conditions and occupancy.

The "building" model also makes it possible to implement a heating anticipation function, when the setpoint is reduced, in anticipation of the future transition to a normal setpoint: the temperature evolution should be estimated on the assumption of 'operation at full load, and to initiate said operation when the temperature estimated for the time of transition coincides with the normal set point.

It is possible to obtain acceptable, although not optimal, operating conditions upon commissioning, by giving the parameters involved in the models the most plausible initial values. Adapting to the specifics of the building and its heating system takes only a few days.

However, it is desirable that the regulator continues to regularly use the self-learning functions, in order to adapt to any changes in the definition of the context.

• {p (i)} , p₁, p₂ du modèle "réseau".

In particular, the parameter p ₄ can undergo significant seasonal variations (higher values in mid-season than in winter), and the provisional condemnation of certain convectors can modify the parameters

• {p (i)}, p ₁ , p ₂ of the "network" model.

Figure 2 represents a test result of a regulator according to the invention, applied to the heating of an inhabited dwelling, immediately after commissioning.

On the abscissa we have a time scale t (Hours); the origin is fixed at 4.30 pm on the 58 ^th day of the year and the measurement ends at 4.30 am on the 60 ^th day of the year.

The interior temperature setpoint (curve T), the interior temperature estimated by the regulator (curve 0 i ₃ ), the temperature of the fluid (curve θ _f ) has been shown; As an indication, the effective internal temperature 0 i and the sunshine (W / m ^{2 -} curve S) are also given, which are not measured by the regulator.

This figure shows that satisfactory performance is obtained quickly after commissioning.

The set temperature T is 20 ° C except between 11 p.m. and 6 a.m., where it is 17 ° C.

A commissioning the 58 ^th day of the year at 16 hours 30, the heating is stopped. This stopping period is necessary to obtain a first estimate of the interior temperature 0 i ₃ .

At 6 pm θ i ₃ being lower than the set temperature, the heating is switched on at full speed, θ _f increases and the regulator operates on the basis of the "building" model which is then the most precise.

At 23 hours θ i ₃ is greater than the value T (17 ° C) and the regulator stops the heating. The temperature θ drops relatively slowly due to the thermal inertia of the network. If we refer to the formula of the "building" model for to = 1 am, θ i ₁ (to) = θ i ₂ (to), the "building" model coincides with the "network" model.

We can then correct the value of p ₃ which could be underestimated.

Around 1:30 am, the regulator implements the "building" model and controls the heating at full load so that the temperature θ i ₃ of 20 degrees is reached at 6 hours.

From 5 am the regulator implements a combination of the two models "building" and "network" until 11 pm.

The "building" model provides a maximum correction g (t) around 3 p.m., hence the minimum temperature of the fluid, in accordance with the reduction in requirements, linked in particular to solar gains (curve S).

At 11 p.m., the regulator orders the heating to stop.

Claims (4)

1. A method of regulating the central heating of a building in which the reference temperature switches to a low value at least once a day, in which method a temperature of the heat-conveying fluid is measured continuously and heating is servo-controlled to an inside temperature value estimated by successive approximations from the temperature of the heat-conveying fluid and the thermal properties of the heating network, the method being characterized by the fact that in order to estimate the inside temperature, two models are used in combination, a "network" model which is used during periods when the heating is stopped or when heating is being reduced, and a "building" model which is used when the temperature of the fluid is high, said "network" model being represented by the relationship:
f(θi₁, θf, τ1) = 0 where:

- θi₁ = the mean inside temperature of the building according to the " network " model;

- θf = the temperature of said heat conveying fluid; and

- τ1 = the load factor, i.e. the ratio of the heating power to maximum power; said "building" model being represented by the relationship:

- θi₂(t) = θi₁(to) + p3[τ2(t) - τ2(to)] + p₄[g(t) - g(to)] where

- θi₂(t) = the mean inside temperature at time t according to the "building" model;

- θi₁ (to) = the mean inside temperature according to the "network" model as stored from time to at the end of the previous prolonged period during which heating was stopped;

- τ2(t), r2(to) = an exponential memory of the instantaneous load factor τ(t) having a characteristic time Δ to;

- p₃ = a parameter defining the sensitivity of the building temperature to heating;

- g(t), g(to) = a time correction function of the inside temperature to take account of the foreseeable portion of hourly fluctuations in climate and conditions of use; and

- p₄ = a parameter defining the amplitude of the time temperature corrections to be performed for the building under consideration.

2. A regulation method according to claim 1, characterized by the fact that the "network" model is represented by the following equation:

τ1 = p₁(θf - θi₁) + P₂(θf - θi₁)²

3. A regulation method according to claim 1 or 2, characterized by the fact that the various parameters of said models are given non-optimized, plausible initial values when the method is put into operation, which values are optimized by automatic learning over a period of a few days.

4. Apparatus for implementing the method according to any preceding claim, the apparatus comprising a regulator (10) integrated in the central heating boiler (1), an input (11) to the regulator being connected to a single member (8) for measuring fluid temperature and an output (12) of the regulator being connected to control the burner (2) of the boiler or to control a three-way or a four-way valve, the apparatus further including means for estimating the inside temperature at all times, the apparatus being characterized in that the inside temperature is estimated using two models in combination, namely the "network" model and the "building" model mentioned in claims 1 and 2.

Images