Method and device for internet-based optimization of parameters of heating control

Abstract

The present invention relates to a method for determining a set of optimized control parameters (Θ.sub.k) of a closed-loop controller (3) or an open-loop controller for an HVACR (heating, ventilation, air conditioning and refrigeration) system (2). In a first method step, an outside temperature (T.sub.A), an actual room temperature (T.sub.R) of a room (9), a supply temperature (T.sub.VL), a predefined target room temperature (T.sub.R,W), and a predefined target supply temperature (T.sub.VL,W) are detected. From the detected measured values (T.sub.A, T.sub.R, T.sub.R,W, T.sub.VL, T.sub.VL,W) and a time (t.sub.k) of detection data packet (D.sub.k) is generated, which is transmitted via an internet connection to a server (8) where the data packet (D.sub.k) is stored in a storage medium (6, 7) connected to the server (8). In the next method step, a set of optimized control parameters (Θ.sub.k) is calculated on the basis of the measured values (T.sub.A, T.sub.R, T.sub.R,W, T.sub.VL, T.sub.VL,W) of the transmitted and stored data packet (D.sub.k) and on the basis of measured values (T.sub.A, T.sub.R, T.sub.R,W, T.sub.VL, T.sub.VL,W) of a plurality of further data packets (D.sub.0 . . . k) generated at an earlier time (t.sub.. . . k) of a specified period (Δt) and/or at least one of a plurality of previously determined sets of optimized control parameters (Θ.sub.k-1) by executing a calculation algorithm on the server (8). In the following method step, the calculated set of optimized control parameters (Θ.sub.k) is stored in the storage medium (6, 7) connected to the server (8) and is transmitted via the internet connection to the closed-loop controller (3) or the open-loop controller of the HVACR system (2) or to a user (B) of the HVACR system (2).

Claims

1. A method for determining a set of optimized control parameters (Θ.sub.k) of a closed-loop controller (3) or an open-loop controller for an HVACR (heating, ventilation, air conditioning and refrigeration) system (2) comprising the following steps: detecting an outside temperature (T.sub.A); detecting an actual room temperature (T.sub.R) of a room (9); detecting a supply temperature (T.sub.VL); detecting a predefined target room temperature (T.sub.R,W); detecting a predefined target supply temperature (T.sub.VL,W); generating a data packet (D.sub.k) with the detected measured values (T.sub.A, T.sub.R, T.sub.R,W, T.sub.VL, T.sub.VL,W) and a time (t.sub.k) of detection; transmitting the data packet (D.sub.k) to a server (8) via an internet connection; storing the data packet (D.sub.k) in a storage medium (6, 7) connected to the server (8); and calculating a set of optimized control parameters (Θ.sub.k) on the basis of the measured values (T.sub.A, T.sub.R, T.sub.R,W, T.sub.VL, T.sub.VL,W) of the transmitted and stored data packet (D.sub.k) and on the basis of: measured values (T.sub.A, T.sub.R, T.sub.R,W, T.sub.VL, T.sub.VL,W) of a plurality of further data packets (D.sub.0 . . . k) generated at an earlier time (t.sub.. . . k) of a specific period (Δt), and/or at least one of a plurality of previously determined sets of optimized control parameters (Θ.sub.k-1); wherein the set of optimized control parameters (Θ.sub.k) is calculated by executing a calculation algorithm on the server (8) that determines a first deviation (e.sub.R) between the actual room temperature (T.sub.R) and the target room temperature (T.sub.R,W) for each data packet (D.sub.0 . . . k) of a predefined time period (Δt), determines a second deviation (e.sub.VL) between the supply temperature (T.sub.R) and the predefined supply temperature (T.sub.VL,W) for each data packet (D.sub.0 . . . k) of the predefined time period (Δt), and determines and applies a weighting factor between 0 and 1 on the basis of the determined deviations (e.sub.R, e.sub.VL) so that, when calculating the set of optimized control parameters (Θ.sub.k), low deviation data packets (D.sub.0 . . . k) are taken more into account and high deviation data packets (D.sub.0 . . . k) are taken less into account; storing the calculated set of optimized control parameters (Θ.sub.k) in the storage medium (6, 7) connected to the server (8); and transmitting the calculated set of optimized control parameters (Θ.sub.k) from the server (8) to the closed-loop controller (3) or the open-loop controller of the HVACR system (2) via the internet connection and adopting the set of optimized control parameters (Θ.sub.k) by the closed-loop controller (3) or the open-loop controller of the HVACR system (2), or sending a notification to a user (B) of the HVACR system (2) with the calculated set of optimized control parameters (Θ.sub.k) and setting, by the user (B), the set of optimized control parameters (Θ.sub.k) on the closed-loop controller (3) or the open-loop controller of the HVACR system (2).

2. The method according to claim 1, wherein the method further comprises: detecting, at the time (t.sub.k), at least one further value from the group consisting of return temperature (T.sub.RL); mass flow ({dot over (m)}); solar radiation (G.sub.sol); external heat source input (P.sub.FW); wherein the generated data packet (D.sub.k) comprises the further measured value(s) (T.sub.RL, {dot over (m)}, G.sub.sol, P.sub.FW), and the calculation algorithm determines the set of optimized control parameters (Θ.sub.k) on the basis of at least one of the further measured values (T.sub.RL, {dot over (m)}, G.sub.sol, P.sub.FW).

3. The method according to claim 1, wherein the time period (Δt) used in calculating the set of optimized control parameters (Θ.sub.k) is predefined on the basis of the number of data packets (D.sub.0 . . . k) stored in the storage medium (6, 7).

4. The method according to claim 1, wherein the measured values are detected regularly after a predefined time interval has elapsed, and the set of optimized control parameters (Θ.sub.k) is calculated recurrently each time a predetermined plurality of predefined time intervals have elapsed, so that a corresponding plurality of generated data packets (D.sub.0 . . . k) for calculating the set of optimized control parameters (Θ.sub.k) are stored in the storage medium (6, 7) of the server (8).

5. A system (1) for determining a set of optimized control parameters (Θ.sub.k) of a closed-loop controller (3) or an open-loop controller for an HVACR (heating, ventilation, air conditioning and refrigeration) system (2) comprising: an outside temperature sensor (11) for detecting an outside temperature (T.sub.A); a room temperature sensor (10) for detecting an actual room temperature (T.sub.R), which is arranged in a room (9); a supply temperature sensor (12) for detecting a supply temperature (T.sub.A); an apparatus (17) for presetting a target room temperature (T.sub.R, w); an apparatus (4) for presetting a target supply temperature (T.sub.VL, W); an apparatus (22) for generating a data packet (D.sub.k) with the detected measured values (T.sub.A, T.sub.R, T.sub.R,W, T.sub.VL, T.sub.VL,W) and a time (t.sub.k) of detection; a transmission apparatus having an interface (20) for transmitting the data packet (D.sub.k) to a server (8) via an internet connection; and a storage medium (6, 7) connected to the server (8) for storing the data packet (D.sub.k); wherein the server (8) has a processor (CPU) configured to: calculate a set of optimized control parameters (Θ.sub.k) on the basis of measured values (T.sub.A, T.sub.R, T.sub.R,W, T.sub.VL, T.sub.VL,W) of the transmitted and stored data packet (D.sub.k) of the point in time (t.sub.k) and on the basis of: measured values (T.sub.A, T.sub.R, T.sub.R,W, T.sub.VL, T.sub.VL,W) of a plurality of further data packets (D.sub.0 . . . k-1) generated at earlier times (t.sub.0 . . . k-1) of a specified time period (Δt); and/or at least one of a plurality of previously determined sets of control parameters (Θ.sub.k-1); execute a calculation algorithm for calculating the set of optimized control parameters (Θ.sub.k) by determining a first deviation (e.sub.R) between actual room temperature (T.sub.R) and target room temperature (T.sub.R,W) for each data packet (D.sub.0 . . . k) of a predefined time period (Δt), determining a second deviation (e.sub.VL) between supply temperature (T.sub.R) and predefined supply temperature (T.sub.VL, W) for each data packet (D.sub.0 . . . k) of the predefined time period (Δt), and determining and applying a weighting factor (w) between 0 and 1 on the basis of the determined deviations (e.sub.R, e.sub.VL), so that, when calculating the set of optimized control parameters (Θ.sub.k), data packets (D.sub.0 . . . k) with low deviation are taken more into account and data packets (D.sub.0 . . . k) with high deviation are taken less into account; store the calculated set of optimized control parameters (Θ.sub.k) in the storage medium (6, 7) connected to the server (8); and transmit the calculated set of optimized control parameters (Θ.sub.k) via an interface (21) via the internet connection to the closed-loop controller (3) or the open-loop controller of the HVACR system (2) and adopt the set of optimized control parameters (Θ.sub.k) by the closed-loop controller (3) or the open-loop controller of the HVACR system (2), or send a notification with the calculated set of optimized control parameters (Ok) to a user (B) of the HVACR system (2) and, if the user (B) confirms the set of optimized control parameters (Θ.sub.k), adopt the set of optimized control parameters (Θ.sub.k) on the closed-loop controller (3) or the open-loop controller of the HVACR system (2).

6. The system (1) according to claim 5, wherein the system (1) further comprises: a return temperature sensor (15) for determining a return temperature (T.sub.RL); and/or a mass flow sensor (13) for detecting a mass flow ({dot over (m)}); and/or a solar radiation sensor (14) for detecting a solar radiation (G.sub.sol); and/or an apparatus for detecting an external heat source input (P.sub.FW); wherein the apparatus (22) for generating a data packet is configured to generate the data packet (D.sub.k) with the further measured value(s) (T.sub.RL, {dot over (m)}, G.sub.sol, P.sub.FW), and the processor (CPU) is designed to calculate the set of optimized control parameters (Θ.sub.k) on the basis of at least one of the further measured values (T.sub.RL, {dot over (m)}, G.sub.sol, P.sub.FW).

7. The system (1) according to claim 5, wherein the processor (CPU) of the server (8) is configured to predefine the time period (Δt) used in calculating the set of optimized control parameters (Θ.sub.k) on the basis of the number of stored data packets (D.sub.0 . . . k).

8. The system (1) according to claim 5, wherein the sensors detect the measured values regularly after a predefined time interval has elapsed, and the processor (CPU) of the server (8) is configured to recurrently calculate the set of optimized control parameters (Θ.sub.k) each time a predetermined plurality of predefined time intervals have elapsed, so that a corresponding plurality of generated data packets (D.sub.0 . . . k) is stored in the storage medium (6, 7) of the server (8) for calculating the set of optimized control parameters (Θ.sub.k).

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) Further advantageous embodiments are described in more detail below on the basis of an exemplary embodiment which is shown in the drawings but to which the invention is not, however, limited.

(2) The drawings show schematically:

(3) FIG. 1 shows a structural design of a system or a method for the internet-based optimization of control parameters of a heating system according to a first exemplary embodiment of the invention.

(4) FIG. 2a and FIG. 2b show further details of the first exemplary embodiment of the invention.

(5) FIG. 3 shows four exemplary heating characteristic curves.

(6) FIG. 4 shows an example of a heating characteristic fitted to a plurality of measured values.

(7) FIG. 5a and FIG. 5b each show a long time series of measured values of a room temperature and a corresponding course of optimized control parameters.

(8) FIG. 6a illustrates a simplified signal flow of the physical process of a heating circuit. FIG. 6b illustrates the approximation of the process by a mathematical model.

DETAILED DESCRIPTION OF THE INVENTION ON THE BASIS OF EXEMPLARY EMBODIMENTS

(9) In the following description of a preferred embodiment of the present invention, equal reference signs designate equal or similar components.

First Exemplary Embodiment

(10) FIG. 1 shows a first exemplary embodiment of a system 1 for calculating a set of optimized control parameters Θ or Θ.sub.k of a closed-loop controller 3 or an open-loop controller 3 for an HVACR (heating, ventilation, air conditioning and refrigeration) system 2 with the characteristic signal flow according to the method of the invention. The vectors Θ={ϑ.sub.1 . . . ϑ.sub.q} or Θ.sub.k={ϑ.sub.1 . . . ϑ.sub.q}.sub.k represent a set of a total of q control parameters. The index k refers to the current point in time and indicates that this is an iteratively optimized set of control parameters. The set of control parameters Θ without the running index k, for example, is used in the conventional least squares method without an iterative calculation method.

(11) As an example of an HVACR system 2, a heating system 2 is considered in the following. The heating system 2 is installed in a building. The following is an example of a single room 9, the temperature T.sub.R of which is controlled by the heating system 2. This room 9 can also be used as a reference room for the temperature control of the entire building. Alternatively, several reference rooms can be used, or a temperature control can be conducted for each room in the building.

(12) The heating system 2 continuously detects relevant system states, such as the outside temperature T.sub.A, the room temperature T.sub.R, the supply temperature T.sub.VL, the return temperature T.sub.RL and the mass flow {dot over (m)} of a carrier medium through a supply pipe or through a return pipe of the heating system 2 and transmits these measured values summarized as data packet D.sub.k of the measurement time t.sub.k via the internet to a server 8. In addition, the solar radiation G.sub.sol, which contributes to the heating of the building or the room, can be detected via a suitable sensor.

(13) The server 8 is designed to store and evaluate the received data packets D.sub.k and thus to determine a set of optimized control parameters Θ.sub.k for a heating controller 3 of the heating system 2. The set of optimized control parameters Θ.sub.k can then be transmitted to a user B, for example the heating contractor or the end customer, who then has the possibility to decide whether to accept the control parameters Θ.sub.k. In this case, the user B sets the control parameters Θ.sub.k on the controller 3. Alternatively, the optimized control parameters Θ.sub.k can also be transmitted directly via the internet to the heating controller 3 and adopted by it.

(14) The method according to the invention can therefore also be used for heating controllers of older heating systems, the control parameters of which cannot be adjusted via the internet. The central server 8 can also simultaneously determine optimized control parameters for several different remote HVACR systems. For this purpose, the server receives 8 data packets of several HVACR systems, evaluates them and transmits the calculated optimized control parameters to the respective user or controller.

(15) FIG. 2a and FIG. 2b show further details of the first exemplary embodiment of the invention. FIG. 2a shows part of system 1 for determining a set of optimized control parameters Θ.sub.k. The system comprises several sensors for recording relevant system states. A supply temperature sensor 12 for detecting the supply temperature T.sub.VL is arranged in the supply of the heating system 2. In the return of the heating system 2 there is a return flow temperature sensor 15 for detecting the return temperature T.sub.RL. A mass flow sensor can also be arranged in the supply pipe and/or in the return pipe to detect the mass flow {dot over (m)}. A room temperature sensor 10 arranged in the room 9 detects the room temperature T.sub.R. The heating system 2 supplies the supply flow with a carrier medium that transports heat to a radiator 16 arranged in the room 9. A thermostat 17 in room 9 can be used to predefine a target room temperature T.sub.R,W.

(16) Outside the building there is an outside temperature sensor 11 for detecting the outside temperature T.sub.A and a solar radiation sensor 14 for detecting solar radiation G.sub.sol. In addition, at least one further sensor (not shown) for detecting an external heat source input P.sub.FW can be arranged in the room. A source for an external heat source input P.sub.FW can be e.g. an electrical appliance that emits heat, such as a refrigerator or an electric stove.

(17) The controller 3 of heating system 2 has a heating circuit controller 4 and a boiler controller 5. The heating circuit controller 4 is supplied with at least the measured variables, room temperature T.sub.R, outside temperature T.sub.A, target room temperature T.sub.R,W and supply temperature T.sub.VL. Optionally, at least one of the measured variables, i.e. solar radiation G.sub.sol, mass flow {dot over (m)} and/or external heat source input P.sub.FW, can be supplied to the heating circuit controller. The heating controller 4 determines a target supply temperature T.sub.VL,W, which is output to a boiler controller 5, using a control method predefined for the heating controller 4. The control method used by heating controller 4 can, for example, be based on a heating characteristic, as described in more detail below. In general, the heating controller 4 uses a set of control parameters Θ={ϑ.sub.1 . . . ϑ.sub.N}. For a heating characteristic, for example, the set of control parameters includes the two parameters slope ϑ.sub.1=θ.sub.S and level ϑ.sub.2=θ.sub.L, for example.

(18) FIG. 2b illustrates the transmission of the measured values in the form of data packets D.sub.k to a server 8 for calculating optimized control parameters Θ.sub.k. The heating controller 3 comprises a device 22 for generating the data packet D.sub.k with the measured values and a point in time t.sub.k for detecting the measured values. Via an interface 20 to the internet, the data packet D.sub.k is transmitted to a server 8, which is also connected to the internet via an interface 21.

(19) The server 8 comprises a buffer 6, in which the received data packet D.sub.k is first intermediately stored, and a permanent memory 7 with a database in which received data packets and calculated control parameters are stored. The server 8 also has a CPU processor that is configured to calculate a set of optimized control parameters Θ.sub.k from the data packets. The set of optimized control parameters Θ.sub.k can then be retransmitted to the controller 4 or to a user of the heating system 2 via the interface 21.

(20) The heating controller 4 of the heating system 2 of the first exemplary embodiment can be controlled, for example, by means of a heating characteristic curve. The heating characteristic describes a relationship between the outside temperature T.sub.A and the target value of the supply temperature T.sub.VL,W of heating system 2. A heating characteristic is characterized by the two parameters level θ.sub.N and slope θ.sub.S. The set of heating characteristic parameters Θ.sub.k to be optimized is therefore Θ.sub.k={θ.sub.N, θ.sub.S}.sub.k.

(21) With a heating characteristic curve, a weather-guided operation of the heating system 2 can be achieved. In the optimization method according to the invention, the system states, i.e. supply temperature T.sub.VL, outside temperature T.sub.A and temperature in the reference room T.sub.R, are detected using the corresponding temperature sensors 10, 11 and 12 shown in FIG. 2a. The target room temperature T.sub.R,W, for example, can be predefined by a user via the thermostat 17.

(22) FIG. 3 shows four exemplary heating characteristics a to d for different values of the parameters, slope θ.sub.S and level θ.sub.N, at a target room temperature T.sub.R,W of 20° C. The three heating characteristics a to c differ only in the parameter slope θ.sub.S. It can be seen that the heating characteristic is steeper when the slope θ.sub.S is increased (from a to c). A change in the level θ.sub.N parameter shifts the characteristic curve in the vertical direction accordingly. This is illustrated using the two heating characteristic curves c and d as examples. Depending on the outside temperature T.sub.A and the adjusted room temperature T.sub.R,W, the heating controller 4 uses the heating characteristic to calculate the target value of the supply temperature T.sub.VL,W, which is passed on to the boiler controller 5 as shown in FIG. 2a.

(23) In the case of generic heating controllers of the prior art, the two parameters, slope and level, of the heating characteristic can be adjusted, e.g. via a configuration menu on the heating controller. Often, an installer of a heating system only sets the parameters depending on the building properties when the heating system is commissioned. In some cases, heating controllers are even operated with the parameters set in the as-delivered state. The parameters are usually only adapted during the heating operation if there is a considerable deviation from the expected comfort of the heating system. In order to avoid an undersupply of heat, heating systems are often operated with a higher supply temperature than necessary. This can lead to excessive energy consumption.

(24) The parameters, slope θ.sub.S and level θ.sub.N, of the heating characteristic can be optimized with the system 1 according to the invention. This means that the heating system 2 can be operated in a particularly efficient way. An excessive energy requirement or an insufficient amount of heat provided is avoided.

(25) As described above, the time-dependent measured values or target values of the relevant status variables are detected at regular intervals and transmitted to the server 8 as data packets. The measured values can here be represented as data vectors which comprise all measured values of a time series. For example, the measured values can be recorded over a period of one day with a resolution t.sub.S of 15 minutes, so that 96 data packets D.sub.1 . . . D.sub.96 are generated in 24 hours and transmitted to the server. The data vector thus comprises 96 measured values of the outside temperature:
T.sub.A=[T.sub.A(1) . . . T.sub.A(96)]

(26) Therefore, the server 8 can calculate the optimized control parameters again at regular intervals t.sub.S. In particular, the control parameters can be calculated iteratively. This means that either the data packets of past points in time or previously calculated control parameters are used in the calculation of the control parameters.

(27) The advantage of the iterative approach is now illustrated by a comparison of a conventional method with the illustrated iterative calculation according to the invention. The calculation of the set of optimized control parameters can, for example, be carried out using known methods. In the following, two exemplary embodiments of the calculation procedure are described. With a conventional least squares method for optimizing control parameters of a heating controller, measured data must first be recorded over a long period of time. The upper diagram in FIG. 5a shows an example of a measured curve of the room temperature T.sub.R over a period of 60 days. This means that when 96 data packets are detected per day, a total of 60.Math.96=5760 data packets are stored and evaluated when the control parameters are calculated. The lower diagram in FIG. 5a shows an initial value Θ.sub.0 of the control parameters as well as the wanted set of optimum control parameters Θ.sub.opt. The conventional method requires 60 days to deliver a result here. The large number of data packets can also make the calculation of the control parameters very time-consuming.

(28) The advantage of the iterative execution of the calculation of optimized control parameters is described on the basis of FIG. 5b. The upper diagram again shows a curve of the room temperature T.sub.R over a period of 60 days. The lower diagram accordingly shows the curve of the control parameters based on the same initial value Θ.sub.0 and the wanted set of optimum control parameters Θ.sub.opt. In the illustrated example, a daily calculation of the set of optimized control parameters is carried out, each time with 96 data packets transmitted and stored, as well as starting from the set of control parameters calculated the day before. This means that a much smaller amount of data has to be processed during the calculation. In addition, the data packets can be deleted every day after calculating the optimized control parameters, so that there is a much smaller memory requirement compared to the conventional method. As the progression of the control parameters in the lower diagram in FIG. 5b illustrates, the result iteratively approaches the wanted optimum set of control parameters. This means that the controller can be operated with optimized control parameters much earlier than in the comparison example in FIG. 5a. In addition, the calculation effort is considerably lower.

(29) Standard Least Squares Method

(30) A first exemplary embodiment of a calculation algorithm for calculating a set of optimized control parameters uses a least squares method. A conventional least squares method determines optimized control parameters Θ by solving the linear equation system:
AΘ=b  (1)
in the sense of

(31) $\arg \underset{\Theta }{\min }.Math.A\Theta -b.Math.$

(32) Matrices A and b here represent the so-called data matrices of the underlying estimation problem. Typically, more data points and thus lines of the data matrices are available as parameters to be estimated, so that the present linear equation system is overdetermined. Such a conventional least squares method is described e.g. in Stoer/Bulirsch: Numerische Mathematik 1, Springer Verlag, Berlin, 2007, page 250 et seq.

(33) Equation (1) initially discloses
A.sup.TAθ=A.sup.Tb.

(34) Assuming that the ATA is regular, the parameters to be estimated follow from
θ=(A.sup.TA).sup.−1A.sup.Tb
with
θϵ.sup.q,Aϵ.sup.N×q,bϵ.sup.N,N>>q.

(35) A heating characteristic is characterized by a set with two parameters, slope θ.sub.S and level θ.sub.N, as described above. The number of parameters is therefore q=2. The result therefore is:
A.sup.TAϵ.sup.q=q,A.sup.Tbϵ.sup.q

(36) With the abbreviations L:=A.sup.TA, r:=A.sup.Tb
the solution can also be compactly implemented through
θ=(A.sup.TA).sup.−1A.sup.Tb=L.sup.−1r

(37) This representation later forms the starting point for an iterative formulation of the estimation problem.

(38) Weighted Least Squares Method

(39) The standard least squares method can be further generalized by weighting the individual lines of the data matrices and thus the data points. A detailed scientific paper of the methods described in this and the next section is described e.g. in Ljung, L.: System Identification—Theory for the User, Prentice Hall, Upper Saddle River (1999) and R. Isermann and M. Münchhof: Identification of Dynamic Systems, Springer (2010).

(40) The weighted least squares method later forms the theoretical starting point for applying the least squares method to the optimization of the heating characteristic. The weighting matrix can be represented as follows

(41) $W\text{:=}\left[\begin{array}{ccc}{w}_1& & \\ & \ddots & \\ & & w_N\end{array}\right],,$
wherein w.sub.i>0 and Wϵ.sup.q×q.

(42) Then, the weighted least squares problem has to be solved:

(43) $\arg \underset{\Theta }{\min }.Math.W.Math.A\Theta -W.Math.b.Math.$

(44) Assuming that A.sup.TWA is regular, the equation (1) then discloses the solution to
θ=(A.sup.T.Math.W.Math.A).sup.−1A.sup.T.Math.W.Math.b  (2)
or with the abbreviations L:=A.sup.TWA, r:=A.sup.TWb
the compact presentation
θ=(A.sup.TWA).sup.−1A.sup.TWb=L.sup.−1r.
Iterative Least Squares Method

(45) With the two approaches described above, it proves to be disadvantageous that first all data points and thus the data matrices A and b must be determined over the entire period, for example a measurement of the data packets D.sub.0 to D.sub.N, the data packet D.sub.N determining the end of the measurement period, which can require a great deal of computing effort. For the large amounts of data, a great deal of storage space and great computer power must be provided. In contrast, it is therefore advantageous to determine a set of optimized control parameters already during the measurement, i.e. when a new data packet is available. The memory space required for this purpose and the necessary computing power can be considerably smaller. This is made possible by an iterative formulation of the least squares problem.

(46) For the iterative approach, the matrices are represented as follows:

(47) $A=\left[\begin{array}{c}{A}_{k-1}\\ A_k\end{array}\right];b=\left[\begin{array}{c}{b}_{k-1}\\ b_k\end{array}\right];W=\left[\begin{array}{cc}{W}_{k-1}& 0\\ 0& W_k\end{array}\right]$

(48) For example, the division is to be interpreted in such a way that variables with the index k−1 represent results of optimized control parameters from past data packets (data packets up to the time point k−1). The index k thus points to a newly generated data packet D.sub.k, which is used to update new control parameters Θ.sub.k.

(49) From equation (2) thus follows:

(50) ${\Theta }_k={\left(\left[A_{k-1}^T{A}_k^T\right]\left[\begin{array}{cc}{W}_{k-1}& \\ & W_k\end{array}\right]\left[\begin{array}{c}{A}_{k-1}\\ A_k\end{array}\right]\right)}^{-2}\left(\left[A_{k-1}^T{A}_k^T\right]\left[\begin{array}{cc}{W}_{k-1}& \\ & W_k\end{array}\right]\left[\begin{array}{c}{b}_{k-1}\\ b_k\end{array}\right]\right)$${\Theta }_k={\left(A_{k-1}^T{W}_{k-1}{A}_{k-1}+A_k^T{W}_k{A}_k\right)}^{-1}\left(A_{k-1}^T{W}_{k-1}{b}_{k-1}+A_k^T{W}_k{b}_k\right)$

(51) With the abbreviations L.sub.k:=A.sub.k.sup.TW.sub.kA.sub.k r.sub.k:=A.sub.k.sup.TW.sub.kb.sub.k
equation (1) can thus be compactly represented by
Θ.sub.k=(L.sub.k-1+L.sub.k).sup.−1(r.sub.k-1+r.sub.k).

(52) The course of a procedure for calculating Θ.sub.k can be summarized as follows using the example of data packets D.sub.0 to D.sub.N. 1. For the data packet D.sub.0, in a first step
L.sub.0:=A.sub.0.sup.TW.sub.0A.sub.0 und r.sub.0:=A.sub.0.sup.TW.sub.0b.sub.0 is calculated. This can be used to determine
θ.sub.0=(L.sub.0).sup.−1(r.sub.0) 2. In a second step, the following is calculated for the data packet D.sub.1
L.sub.1:=A.sub.1.sup.TW.sub.1A.sub.1 und r.sub.1:=A.sub.1.sup.TW.sub.1b.sub.1.
and
θ.sub.1=(L.sub.0+L.sub.1).sub.−1(r.sub.0+r.sub.1) is determined therefrom. 3. Accordingly, the following is calculated in a third step for the data packet D.sub.k
L.sub.k:=A.sub.k.sup.TW.sub.kA.sub.k and T.sub.k:=A.sub.k.sup.TW.sub.kb.sub.k. The following can be calculated therefrom
θ.sub.k=(L.sub.k-1+L.sub.k).sup.−1(r.sub.k-1+r.sub.k) 4. Then, L.sub.k-1.fwdarw.L.sub.k-1+L.sub.k and r.sub.k-1.fwdarw.r.sub.k-1+r.sub.k are set and the run variable k is increased by one. 5. If k=N+1, the procedure is terminated. Otherwise return to step 3.

(53) The following follows for the example the heating characteristic optimization: A.sub.kϵ.sup.N×q L.sub.kϵ.sup.q×q r.sub.kϵ.sup.q b.sub.kϵ.sup.N θ.sub.kϵ.sup.q

(54) For a daily calculation with a time resolution of 15 minutes, the parameters are N=96; q=2. Only the parameters θ.sub.k, L.sub.k-1, r.sub.k-1 need to be stored in the storage medium 7 of the server 8, so that there is considerably less memory requirement compared with a conventional method. Basically, all parameter optimization problems of a control, which can be represented by equation (1), can be efficiently managed and calculated on the server 8 using the above described procedure according to the invention.

(55) Adaptation to a Heating Characteristic Curve

(56) It is shown below how the method described above can be used to optimize a heating characteristic. A heating characteristic describes a static relationship between outside temperature T.sub.A, target room temperature T.sub.R,W and target supply temperature T.sub.VL,W and is described by equation (3):

(57) $\begin{array}{cc}{T}_{\mathrm{VL},W}=T_{R,W}+{\theta }_L-{\theta }_S\left(\underset{\alpha =1}{\overset{3}{.Math.}}{\left(T_A-T_{R,W}\right)}^{\alpha }.Math.a_{\alpha }\right)& \left(3\right)\end{array}$

(58) Here, a.sub.α are known, previously defined parameters of the heating curve.

(59) Equations (1) and (3) lead to the data matrices:

(60) $b\text{:=}{\stackrel{\_}{T}}_{\mathrm{VL},W}-{\stackrel{\_}{T}}_{R,W}$$A\text{:=}\left[\left[\begin{array}{c}1\\ \mathrm{.Math.}\\ 1\end{array}\right],-1.Math.\left(\underset{\alpha =1}{\overset{3}{.Math.}}{\left({\stackrel{\_}{T}}_A-{\stackrel{\_}{T}}_{R,W}\right)}^{\alpha }.Math.a_{\alpha }\right)\right]$
with the parameters to be determined

(61) $\Theta \text{:=}\left[\begin{array}{c}{\theta }_L\\ {\theta }_S\end{array}\right]$
Rule Deviations
e.sub.VL:=T.sub.VL,W−T.sub.VL
e.sub.R:=T.sub.H,W−T.sub.R

(62) With the temperature sensors shown in FIG. 2a, the measured values T.sub.VL(t), T.sub.A(t), T.sub.R(t), T.sub.VL,W(t), T.sub.R,W(t) are recorded at regular time intervals and transmitted in data packets as time series to the server 8 where they are stored and evaluated. FIG. 4 shows how a heating characteristic can be determined from the measured values. In FIG. 4, the measured values are plotted in a diagram. The plotted data points obtained from the measured values can also be filtered using a weighting function, so that only those data points are used for which the deviations e.sub.VL and/or e.sub.R are very small. The heating characteristic curve fitted to the data points then supplies the optimized control parameters Θ.sub.k which can be transferred to the heating controller 3.

(63) For example, a Gaussian distribution curve can be used as a weighting function:

(64) $W_{\left\{\mathrm{VL},R\right\},i}=\exp \left(-\frac{e_{\left\{\mathrm{VL},R\right\},i}^2}{2{\sigma }_{\left\{\mathrm{VL},R\right\}}}\right)$

(65) The setting parameters σ.sub.VL and σ.sub.R are essentially freely selectable. Depending on the quantity of the data packets available for the evaluation or depending on the scatter of the measured values, the setting parameters can be set larger or smaller in order to determine a set of optimized control parameters Θ.sub.k.

Second Exemplary Embodiment: System Identification

(66) In contrast to the heating characteristic curve of the first exemplary embodiment, the second exemplary embodiment deals with an explicit control of the temperature T.sub.R of a reference room 9 in a building. The second exemplary embodiment describes how a dynamic model M can be determined using the internet and the stored data packets. The model M is a mathematical description of the relationships between the supply temperature T.sub.VL and the outside temperature T.sub.A to the room temperature T.sub.R. The model is determined on the basis of the measured data.

(67) The data packets are generated and transferred to a server 8 as described above using the first exemplary embodiment. In the following, therefore, only the details of the mathematical method for calculating the optimized control parameters Θ.sub.k are dealt with.

(68) A mathematical building model allows the control parameters of a controller 3 of an HVACR system 2 to be specifically optimized, for example by means of a P, PI or PID controller, on the basis of the model. The optimization can be carried out e.g. by standard control methods such as root locus curves, frequency line methods, etc. The determination of the parameters of the building model and the calculation of the optimized control parameters Θ.sub.k are carried out as in the first exemplary embodiment on a server 8 that communicates with the controller via an internet connection. The controller 3 itself works locally, for example in the heat generator of a heating system 2, and is automatically reparameterized at regular intervals (e.g. daily, weekly or monthly).

(69) According to the second exemplary embodiment of the invention, the control parameters are optimized using a model M of the controlled system. For this purpose, a calculation algorithm executed on server 8 determines a mathematical relationship between the input and output data of the control process on the basis of the measurement data obtained, i.e. the data packets D.sub.k.

(70) FIG. 6a shows the initial situation for an exemplary process P, here a heating circuit. The outside temperature T.sub.A is a measurable disturbance variable. The supply temperature T.sub.VL of the heating circuit is the input variable. The room temperature T.sub.R is the output or controlled variable. Further, non-measurable, disturbance variables are designated by G=[g.sub.1, . . . , g.sub.q].

(71) The aim of the method is to determine a mathematical model M so that the output variable T.sub.R supplied by model M corresponds as exactly as possible to the measured room temperature T.sub.R, as schematically shown in FIG. 6b. The aim is therefore to minimize the deviation e between the output variable T.sub.R and the measured room temperature T.sub.R.

(72) As a simple approach for model M, for example, a time-discrete transfer function in the frequency domain for the supply temperature G.sub.VL(Z) and a time-discrete transfer function for the outside temperature G.sub.A(Z) can be selected, from which
T.sub.R(Z)=G.sub.VL(Z).Math.T.sub.VL(Z)+G.sub.A(Z).Math.T.sub.A(Z)
follows. The time t is expressed in the following as a multiple of a sampling time t.sub.S, so that: t=k.Math.t.sub.s. The time-discrete transfer function in the frequency domain is expressed here as a time-discrete transfer function in the image domain of a Z-transformation.

(73) The transfer function in the frequency domain corresponds in the time domain to the difference equation

(74) 0$\begin{array}{cc}{T}_R\left(k.Math.t_S\right)=-\underset{i=0}{\overset{N_p-1}{.Math.}}{a}_i{T}_R\left(\left(k-N_p+i\right)t_S\right)+\underset{i=0}{\overset{N_{Z,\mathrm{VL}}}{.Math.}}{b}_{\mathrm{VL},i}{T}_{\mathrm{VL}}\left(\left(k-N_{Z,\mathrm{VL}}^{-1}+i\right)t_S\right)+\underset{i=0}{\overset{N_{Z,A}}{.Math.}}{b}_{A,i}{T}_A\left(\left(k-N_{Z,A}^{-1}+i\right)t_S\right)k\in \{0,1,\mathrm{.Math.}\};N_{Z,\mathrm{VL}}\le N_p;N_{Z,A}\le N_p& \left(4\right)\end{array}$

(75) N.sub.P is here the number of pole positions of the transfer function and N.sub.Z,VL or N.sub.Z,A is the number of corresponding zero positions. In principle, these are freely selectable, but can usually be determined in advance from physical considerations.

(76) The mathematical model M calculates the model output using historical values of the outside temperature T.sub.A and room temperature T.sub.R, which are stored as data packets in the storage medium 7 of server 8. The calculation of the model parameters a.sub.i, b.sub.VL,i, b.sub.A,i can again be presented as a least squares problem as in the first exemplary embodiment: AΘ=b. The matrices A and b can be generated from the data packets D.sub.i. The model can be extended accordingly if further disturbance variables are known, such as the solar radiation G.sub.sol or an external heat source input P.sub.FW.

(77) In order to calculate the model parameters, the problem is converted into a least squares problem. The data packets D.sub.1 to D.sub.N are used for this purpose. With a sampling time t.sub.s of 15 minutes, for example, N=96 is used for one day. Thus: k=1, . . . , 96 N.sub.p=2 N.sub.Z,VL=1 N.sub.Z,A=1

(78) It follows from equation (4) that:
T.sub.R(k)=−a.sub.1T.sub.R(k−1)−a.sub.0T.sub.R(k−2)+b.sub.VL,1T.sub.VL(k−1)+b.sub.VL,0T.sub.VL(k−2)+b.sub.A,1T.sub.A(k−1)+b.sub.A,0T.sub.A(k−2)

(79) From D.sub.1, . . . D.sub.N thus follows that:

(80) $\begin{array}{c}\left[\begin{array}{c}{T}_R\left(1\right)\\ \mathrm{.Math.}\\ T_R\left(n\right)\end{array}\right]=\left[\begin{array}{c}-a_1{T}_R\left(0\right)-a_0{T}_R\left(-1\right)+b_{\mathrm{VL},i}{T}_{\mathrm{VL}}\left(0\right)+b_{\mathrm{VL},0}{T}_{\mathrm{VL}}\left(-1\right)+b_{A,1}{T}_A\left(0\right)+b_{A,0}{T}_A\left(-1\right)\\ \mathrm{.Math.}\\ -a_1{T}_R\left(N-1\right)-a_0{T}_R\left(N-2\right)+b_{\mathrm{VL},1}{T}_{\mathrm{VL}}\left(N-1\right)+b_{\mathrm{VL},0}{T}_{\mathrm{VL}}\left(N-2\right)+b_{A,1}{T}_A\left(N-1\right)+b_{A,0}{T}_A\left(N-2\right)\end{array}\right]\\ =\left[\begin{array}{cccccc}-T_R\left(0\right)& -T_R\left(-1\right)& T_{\mathrm{VL}}\left(0\right)& T_{\mathrm{VL}}\left(-1\right)& T_A\left(0\right)& T_A\left(-1\right)\\ \mathrm{.Math.}& \mathrm{.Math.}& \mathrm{.Math.}& \mathrm{.Math.}& \mathrm{.Math.}& \mathrm{.Math.}\\ -T_R\left(N-1\right)& -T_R\left(N-2\right)& T_{\mathrm{VL}}\left(N-1\right)& T_{\mathrm{VL}}\left(N-2\right)& T_A\left(N-1\right)& T_A\left(N-2\right)\end{array}\right].Math.\left[\begin{array}{c}{a}_1\\ a_0\\ b_{\mathrm{VL},1}\\ b_{\mathrm{VL},0}\\ b_{A,1}\\ b_{A,0}\end{array}\right]\\ =A.Math.\Theta \end{array}$

(81) The matrices A and b can therefore be generated from the data packets D.sub.i transferred to the server 8. The transfer functions G.sub.VL(Z) and G.sub.A(Z) can be calculated from the control parameters θ=[a.sub.1, a.sub.0, . . . , b.sub.A,0]. If the transfer functions G.sub.VL(Z) and G.sub.A(Z) are known, a controller 3 (e.g. a P, PI or PID controller) can be designed to control the room temperature T.sub.R. This can be done on the basis of the model using standard control techniques, for example a root locus curve, a frequency characteristic curve or the like.

(82) Since the parameters of model M are updated cyclically, e.g. daily, weekly or monthly, the result is an adaptive or optimized control of the heating system 2 adapted to the process (here a heating circuit). The control parameters of the controller 3 are updated cyclically accordingly. This allows continuous, internet-based optimization of the controller 3.

(83) The advantage and innovation of this method is that there is no need to maintain and process the parameters for the purpose of identifying them on the embedded systems of the heat generator since these calculations are carried out on the corresponding capacities of the central platform on the internet.

(84) The features disclosed in the above description, the claims and the drawings can be important both individually and in any combination for the realization of the invention in its various embodiments.

LIST OF REFERENCE SIGNS

(85) 1 System for determining optimized control parameters 2 HVACR system 3 closed-loop/open-loop controller 4 heating circuit controller 5 boiler controller 6 buffer (storage medium) 7 database (storage medium) 8 server 9 room 10 room temperature sensor 11 outside temperature sensor 12 supply temperature sensor 13 mass flow sensor 14 solar radiation sensor 15 return temperature sensor 16 radiator 17 thermostat 20 interface of the controller 21 interface of the server 22 apparatus for generating data packets CPU processor B user