JP5869406B2 - Heat exchange system and controller - Google Patents

Description

  The present invention relates to a heat exchange system and a controller.

  Currently, even when a dead time or the like occurs in controlling a predetermined control target, a technique for accurately controlling the control target has been developed.

  For example, in Patent Document 1, an operation amount that is an input signal to a control target is passed through a process identification model to compensate for dead time, and the control amount of the control target (in Patent Document 1, an amount output by the control target). A dead time compensation control device that outputs (feedback amount)), and adopts a neural network as the process identification model, and an estimation means for estimating a state quantity of the controlled object from the manipulated variable and the controlled variable; A dead time compensation signal calculation means for calculating a dead time compensation amount in the process identification model that receives only the state quantity, and calculates a deviation between the target value of the controlled object and the controlled quantity to the controlled object Difference calculating means, PI calculating means for PI calculating the output of the difference calculating means, and adding means for adding the output from the PI calculating means to the dead time compensation signal; A dead time compensation control device comprising: main control means for calculating the final manipulated variable for the controlled object by performing PID calculation or fuzzy calculation on the addition signal output from the adding means. ing.

JP 2001-296907 A

  Although the dead time compensation is compensated for by the dead time compensation control device, when the control object is a heat exchange object to be heat exchanged by the heat exchanger, the physical quantity of the heat exchange object before heat exchange. However, since such an effect is not taken into account in the dead time compensation control apparatus, it is impossible to perform a control with high accuracy.

  This invention is made | formed in view of the said situation, and it is providing the heat exchange system and controller which can perform an accurate control with respect to the control object in the system which a dead time produces.

In order to achieve the above object, a heat exchange system according to the first aspect of the present invention includes:
A heat exchanger in which a heat medium flows in and changes the temperature of the heat exchange object by performing heat exchange using the heat medium that has flowed in;
An adjustment mechanism for adjusting the temperature of the heat medium flowing into the heat exchanger;
Control means for controlling the adjustment mechanism;
A heat exchange system comprising:
The control means controls the adjustment mechanism by supplying a control amount for changing the temperature of the heat medium flowing into the heat exchanger to the adjustment mechanism,
The adjustment mechanism changes the temperature of the heat medium flowing into the heat exchanger according to the control amount supplied from the control means,
The control means includes an input value representing a measurement value obtained by measuring a predetermined physical quantity of the heat exchange object before heat exchange, and an input value representing a measurement value obtained by measuring the temperature after heat exchange of the heat exchange object. The heat after the dead time has elapsed from when the control amount is supplied to the adjustment mechanism until the temperature of the heat medium flowing into the heat exchanger changes based on the input value representing the control amount. The neural network that predicts the temperature after heat exchange of the object to be exchanged and outputs it as a predicted value, and the predicted value output by the neural network is compared with a predetermined target value, based on the comparison result, Control amount changing means for changing the control amount to be supplied to the adjustment mechanism so that the temperature after heat exchange of the heat exchange object approaches the target value,
It is characterized by that.

The heat exchange system includes:
A plurality of neural networks having different characteristics;
Selection means for acquiring information representing the operating state of the heat exchange system, and selecting any one of the plurality of neural networks as a neural network that is actually used based on the acquired information representing the operating state; With
The control amount changing means controls the control amount by comparing the predicted value output from the neural network selected by the selection means with the target value.
You may do it.

In the heat exchange system,
The heat exchange object is air,
The heat exchanger performs heat exchange between the air flowing from the outside and the heat medium, and sends the air after heat exchange to the outside.
The measurement value obtained by measuring a predetermined physical quantity of the heat exchange object before heat exchange is a measurement value obtained by measuring at least the temperature of the temperature, humidity, and flow rate of air flowing from the outside,
The measured value obtained by measuring the temperature after heat exchange of the heat exchange object is a measured value obtained by measuring the temperature of the air sent to the outside.
You may do it.

In the heat exchange system,
The measurement value obtained by measuring a predetermined physical quantity before heat exchange of the heat exchange object is a measurement value obtained by measuring at least the temperature of the temperature and humidity of the air flowing from the outside,
The neural network obtains a flow rate of air flowing from the outside after the dead time has elapsed, and further uses the obtained flow rate after the dead time has elapsed to perform the prediction.
You may do it.

In the heat exchange system,
The input value representing the control amount is a value representing a measurement value obtained by measuring the temperature of the heat medium changed according to the control amount within a predetermined period after the control amount is supplied to the adjustment mechanism.
You may do it.

The heat exchange system includes:
A measurement value obtained by measuring the predetermined physical quantity before heat exchange of the heat exchange object obtained by operating the heat exchange system, a measurement value obtained by measuring the temperature after heat exchange of the heat exchange object, A construction means for constructing the neural network based on the value representing the control amount;
You may do it.

In the heat exchange system,
The adjustment mechanism is connected to the first flow path through which the first heat medium adjusted to a predetermined temperature flows, and the first heat medium from the first flow path. By flowing in, the second heat medium having a temperature corresponding to the inflow amount of the first heat medium flows through the second flow path connected to the adjustment mechanism, and disposed in the first flow path. A valve that adjusts the flow rate of the first heat medium flowing into the second flow path by becoming a valve opening according to a control amount supplied from the control means,
The input value representing the control amount is a value representing the valve opening.
You may do it.

In the heat exchange system,
In the adjustment mechanism, the heat exchanger performs heat exchange so that the flow rate of the second heat medium in the second flow path is constant regardless of the inflow amount of the first heat medium. A third flow path into which a part of the heat medium of 2 flows,
You may do it.

In the heat exchange system,
The input value representing the control amount is in accordance with the inflow amount of the first heat medium that flows in the vicinity of the connection portion of the second flow path between the first flow path and the second flow path. Represents the measured value of the temperature of the second heat medium,
The valve is disposed in the vicinity of the connection portion in the first flow path.
You may do it.

In order to achieve the above object, a controller according to the second aspect of the present invention provides:
A heat exchanger in which a heat medium flows in and changes the temperature of the heat exchange object by performing heat exchange using the heat medium that has flowed in;
An adjustment mechanism for adjusting the temperature of the heat medium flowing into the heat exchanger, and a controller for controlling the adjustment mechanism in a heat exchange system,
The controller controls the adjustment mechanism by supplying a control amount for changing the temperature of the heat medium flowing into the heat exchanger to the adjustment mechanism;
The adjustment mechanism changes the temperature of the heat medium flowing into the heat exchanger according to the control amount supplied from the controller,
The controller is an input value representing a measured value obtained by measuring a predetermined physical quantity of the heat exchange object before heat exchange; an input value representing a measured value obtained by measuring the temperature after heat exchange of the heat exchange object; The heat exchange after a lapse of dead time from the supply of the control amount to the adjustment mechanism to the change of the temperature of the heat medium flowing into the heat exchanger based on the input value representing the control amount A neural network that predicts the temperature after heat exchange of an object and outputs it as a predicted value, and compares the predicted value output from the neural network with a predetermined target value, and based on the comparison result, Control amount changing means for changing the control amount to be supplied to the adjustment mechanism so that the temperature after heat exchange of the exchange object approaches the target value,
It is characterized by that.

  According to the present invention, accurate control can be performed on a control target in a system in which dead time occurs.

It is a lineblock diagram of the heat exchange system concerning one embodiment of the present invention. It is an example of the hardware block diagram of the controller of FIG. FIG. 2 is a configuration diagram of a controller when the heat exchange system of FIG. 1 is in a trial operation mode. FIG. 2 is a configuration diagram of a controller when the heat exchange system of FIG. 1 is in a main operation mode. It is a conceptual diagram of the neural network used with the heat exchange system of FIG. It is a simulation result of PID control without using a neural network. It is a simulation result of PID control without using a neural network. It is a simulation result of PID control without using a neural network. It is a simulation result of PID control without using a neural network. It is a simulation result at the time of using a neural network in the heat exchange system of FIG. It is a simulation result at the time of using a neural network in the heat exchange system of FIG. It is a simulation result, such as the case where a neural network is used at the time of start-up (modification) of the heat exchange system of FIG. It is a simulation result, such as the case where a neural network is used at the time of start-up (modification) of the heat exchange system of FIG. It is a block diagram of a controller when the heat exchange system concerning a modification is this operation mode.

  A heat exchange system according to an embodiment of the present invention will be described.

  As shown in FIG. 1, the heat exchange system 100 according to the present embodiment is configured as a bleed-in air conditioning system. The heat exchange system 100 includes an outward path 11, a circulation path 13, a return path 15, a controller 101, a valve 103, a pump 105, a heat exchanger 107, a heat source device 109, a heat medium thermometer 121, A measurement unit 123 and an outlet thermometer 125 are provided.

  The forward path 11 includes, for example, a plurality of pipes, and causes a heat medium (for example, hot water or cold water) that has reached a predetermined temperature by the heat source device 109 to flow into the circulation path 13 from the heat source device 109.

  The circulation path 13 includes, for example, a plurality of pipes, and forms a passage for circulating the heat medium (in FIG. 1, the heat medium circulates clockwise, see a dotted arrow). The circulation path 13 includes a first connection portion 13A and a second connection portion 13B configured from pipe joints and the like. The forward path 11 is connected to the first connecting portion 13A, and the return path 15 is connected to the second connecting portion 13B. The heat medium flowing in from the forward path 11 flows into the circulation path 13 from the first connection portion 13A. On the other hand, a part of the heat medium circulating in the circulation path 13 flows into the return path 15 from the second connection portion 13B.

  The return path 15 includes, for example, one or more pipes, and returns the heat medium flowing from the circulation path 13 to the heat source device 109.

  The controller 101 controls the valve 103. Details of the controller 101 will be described later.

  The valve 103 is a two-way valve provided in the middle of the forward path 11. The valve 103 is driven by an appropriate driving method such as an electric type, a hydraulic type, or a pneumatic type. The valve 103 is disposed between the two pipes constituting the forward path 11 and is provided in the middle of the forward path 11 by being connected to the two pipes. In particular, the valve 103 is provided in the middle of the forward path 11 so as to be positioned in the vicinity of the first connecting portion 13A. The valve opening degree of the valve 103 is controlled by the controller 101. By controlling the valve opening, the flow rate of the heat medium flowing from the forward path 11 into the circulation path 13 is controlled. The valve 103 maximizes the flow rate of the heat medium flowing into the circulation path 13 when the valve opening degree is maximum (when the valve 103 is fully open), and when the valve opening degree is “0” (for example, the valve 103 is In the closed state), the flow rate of the heat medium flowing into the circulation path 13 is minimized (for example, the heat medium does not flow).

  The pump 105 is provided in the middle of the circulation path 13. For example, the pump 105 is disposed between two pipes constituting the circulation path 13 and is provided in the middle of the circulation path 13 by being connected to the two pipes. The pump 105 is caused to flow in the heat medium in the circulation path 13. That is, the heat medium flows (circulates) through the circulation path 13 by the pump 105.

  The heat exchanger 107 is provided in the middle of the circulation path 13. For example, the heat exchanger 107 is disposed between the two pipes constituting the circulation path 13 and is provided in the middle of the circulation path 13 by being connected to the two pipes. The heat medium from the circulation path 13 flows from one side into the heat exchanger 107, and the heat medium flowing into the heat exchanger 107 flows through a predetermined path and flows out from the other side of the heat exchanger 107 to the circulation path 13. In this way, the heat medium flowing through the circulation path 13 flows through a predetermined path in the heat exchanger 107. Further, the heat exchanger 107 exchanges heat between, for example, air that flows in from an inlet (such as an air inlet) by intake air and the heat medium that flows through a predetermined path in the heat exchanger 152, and heats the air that flows in Alternatively, it is cooled and heated or cooled air (heat exchanged air) is sent out from the outlet (air blowing port).

  The heat source device 107 is connected to the forward path 11 and the return path 15, and has a role of heating or cooling the heat medium flowing in from the return path 15 to a predetermined constant temperature set in advance in the heat source apparatus 107 and flowing it to the forward path 11.

  The heat medium thermometer 121 is provided in the circulation path 13 in the vicinity of the first connection section 13A between the first connection section 13A and the pump 105, and the heat medium flowing in the forward path 11 and the heat medium flowing in the circulation path 13 Measure the temperature of the heat medium immediately after the merging. The thermometer 121 for heat medium is arrange | positioned by providing in the piping which comprises the circulation path 13, for example. The temperature of the heat medium measured by the heat medium thermometer 121 is supplied to the controller 101 in the form of an electrical signal. The temperature of the heat medium to be measured by the heat medium thermometer 121 may be referred to as the heat medium temperature.

  The measurement unit 123 is disposed in the vicinity of the inlet (such as an intake port) in the heat exchanger 107, and measures the temperature, humidity, and flow rate of air flowing from the inlet into the heat exchanger 107. Specifically, the measurement unit 123 includes a thermometer that measures the temperature, a hygrometer that measures the humidity, and a flow meter that measures the flow rate, and the temperature of the air measured by each measuring instrument, Humidity and flow rate are independently supplied to the controller 101 in the form of electrical signals. Note that the temperature, humidity, and flow rate of air to be measured by the measurement unit 123 may be referred to as inlet air temperature, inlet air humidity, and inlet air flow rate, respectively.

  The outlet thermometer 125 is arranged in the vicinity of the outlet (sending port) in the heat exchanger 107, and measures the temperature of the air (air after heat exchange) sent from the outlet. The temperature of the air measured by the outlet thermometer 125 is supplied to the controller 101 in the form of an electric signal. The temperature of air to be measured by the outlet thermometer 125 may be referred to as outlet air temperature.

  Since the heat exchange system 100 is a bleed-in system, the flow rate (and the flow speed) of the heat medium flowing (circulating) through the circulation path 13 depends on the inflow amount of the heat medium flowing into the circulation path 13 from the forward path 11. It does not change and becomes constant. That is, the flow rate of the heat medium flowing in the forward path 11 and the flow rate of the heat medium flowing in the return path 15 are the same. Since the heat medium flowing in the forward path 11 is brought to a predetermined constant temperature by the heat source device 109, when this heat medium joins the heat medium flowing in the circulation path 13, it circulates from the second connection portion 13B toward the first connection portion 13A. The temperature of the heat medium before merging flowing through the path 13 is changed by this merging. The degree of this temperature change is that the flow rate of the heat medium flowing (circulating) through the circulation path 13 is constant, so the flow rate of the heat medium flowing from the forward path 11 into the circulation path 13 (that is, the valve opening degree of the valve 103). Depends on. That is, the temperature of the heat medium flowing through the circulation path 13 from the first connection portion 13A toward the heat exchanger 107 is determined by the flow rate of the heat medium flowing into the circulation path 13 from the forward path 11 (that is, the valve opening degree of the valve 103). It changes by changing. And since the heat medium which flows through the circulation path 13 toward the heat exchanger 107 from the 1st connection part 13A flows into the heat exchanger 107 and is used for heat exchange, the temperature of this heat medium is the heat exchanger 107. It also affects the temperature after heat exchange of the air to be heat exchanged. As described above, in the heat exchange system 100, the temperature of the heat medium flowing into the heat exchanger 107 can be controlled by controlling the valve opening degree of the valve 103, and further, the air after the heat exchange (from the outlet) can be controlled. Air) temperature (outlet air temperature) can be controlled.

  The valve opening degree of the valve 103 changes, and the heat medium whose flow rate has changed passes through the forward path 11 and the circulation path 13 to reach the heat exchanger 107, that is, after the flow rate has changed in the valve 103, Time L is required until the temperature of the heat medium flowing into the heat exchanger 107 changes after the change is reflected. This time L becomes longer as the distance from the valve 103 to the heat exchanger 107 (distance on the forward path 11 and the circulation path 13) is longer. When the distance from the valve 103 to the heat exchanger 107 (distance on the forward path 11 and the circulation path 13) is H [m], and the flow rate (flow velocity) of the heat medium is V [m / s], the heat exchanger 1 The time L required until the cold water (hot water) having the temperature changed to is reached is represented by the following mathematical formula (1).

  The time L is a “dead time” in the heat exchange system 100. The controller 101 acquires the outlet air temperature measured and supplied by the outlet thermometer 125, and based on the acquired outlet air temperature, the outlet air temperature is a preset target value (for example, the outlet air temperature set by the user). If the valve opening degree of the valve 103 is controlled by feedback control (for example, PI (Proportional Integral) control or PID (Proportional Integral Differential) control) so as to approach the set temperature of Since the control is actually reflected in the outlet air temperature after the dead time L has elapsed, it becomes difficult to control the outlet air temperature.

  Therefore, the controller 101 of the present embodiment performs control in consideration of the dead time using a neural network.

  As shown in FIG. 2, the controller 101 includes, for example, an input / output unit 111, a one-chip first microcomputer (my computer) 112, and a one-chip second microcomputer (my computer) 113.

  Each of the first microcomputer 112 and the second microcomputer 113 changes (adds or deletes) stored information (programs and data) such as a ROM (Read Only Memory) that stores programs and data, a hard disk, and a flash memory. Non-volatile storage device (hereinafter may be referred to simply as non-volatile storage device) and a ROM or non-volatile storage device based on a program stored in the ROM or non-volatile storage device. A CPU (Central Processing Unit) that actually executes processing of each microcomputer using stored data, and a RAM (Random Access Memory) serving as a main memory of the CPU are provided. The first microcomputer 112 and the second microcomputer 113 can exchange information with each other.

  The input / output unit 111 includes various input devices that receive operation from the user, including operation buttons, for example, generate operation information representing the content of the received operation (operation content), and supply the operation information to the first microcomputer 112 And a display device that is controlled by the first microcomputer 112 and appropriately displays an operation screen or the like. The operation from the user includes an operation for specifying the target temperature of the outlet air temperature and an operation for specifying the operation mode of the heat exchange system 100.

  The controller 101 constructs a neural network used in the actual operation when the heat exchange system 100 performs a trial operation. That is, system identification of this heat exchange system 100 is performed. In this case, the controller 101 operates with the configuration shown in FIG. For example, when the heat exchange system 100 performs a test operation, if the user performs an operation for instructing the operation mode of the test operation to the input / output unit 111 and an operation for specifying a desired outlet air temperature as a target value, these operations are performed. Operation information to be represented is supplied to the first microcomputer 112. When this operation information is supplied, the first microcomputer 112 sets a target value designated by the operation represented by the operation information (for example, target value data is held in a RAM or the like), and this target value is used. It operates as a valve control unit 115 that executes PID control (may be PI control) on the valve 103. Further, the first microcomputer 112 supplies a predetermined signal to the second microcomputer 113 to cause the second microcomputer 113 to operate as a neural network construction unit 116 for constructing a neural network.

In the neural network construction unit 116, a neural network in an initial state is constructed in advance. In this embodiment, this neural network employs a three-layer neural network of an input layer, an intermediate layer, and an output layer (see FIG. 5). The input layer and the intermediate layer, and the intermediate layer and the output layer are coupled by a coupling load. The input values input to the input layer are five values of the heat medium temperature, the inlet air temperature, the inlet air humidity, the inlet air flow rate, and the outlet air temperature. The number of elements in the intermediate layer is “11”. The output is a predicted value of the outlet air temperature of the dead time elapsed. When the input value to the input layer is x _i , the coupling load between the input layer and the intermediate layer is w _ij , the coupling load between the intermediate layer and the output layer is w _jk , and the input to the element of the intermediate layer is s _j , the intermediate layer input the s _j to the element, expressed by the following equation (2), the output q _j from the intermediate layer to the output layer is expressed by the following equation (3) and (4), the output y _k of a neural network Is expressed by the following formula (5). Here, the subscripts i, j, and k are element numbers of the input layer, the intermediate layer, and the output layer, respectively, and n and m are the number of inputs and the number of intermediate elements, respectively. Since the output of the neural network is one, y = 1 is set. Further, x ₀ = 0, s ₀ = 0, q ₀ = 0, x ₁ is the heat medium temperature, x ₂ is the inlet air temperature, x ₃ is the inlet air humidity, x ₄ is the inlet air flow rate, and x ₅ is the outlet Let air temperature.

The neural network constructing unit 116 measures each of the heat medium temperature, the inlet air temperature, the inlet air humidity, the inlet air flow rate, and the outlet air temperature from the heat medium thermometer 121, the measurement unit 123, and the outlet thermometer 125. Values are sequentially input at predetermined intervals in real time. The neural network construction unit 116 sequentially inputs each input measurement value as a set of input values x _{1 to} x ₅ to the neural network, and causes the neural network to sequentially perform operations as exemplified by the above formula. the output value y ₁ are sequentially output.

The neural network construction unit 116 supplies the output value y ₁ output from the neural network to the valve control unit 115. The valve control unit 115 performs a PID calculation (may be a PI calculation; the same applies hereinafter) based on the currently set target value and the sequentially supplied outlet air temperature, and controls the valve 103 according to the calculation result. By performing the control, PID control (PI control may be performed. The same applies hereinafter) that causes the outlet air temperature to approach the target value. In this embodiment, the neural network construction unit 116 calculates the operation amount to be supplied to the valve 103 by the above calculation, and controls the valve opening degree of the valve 103 by supplying the calculated operation amount to the valve 103. The operation amount may be an amount that specifies how much the valve 103 is opened and closed by operating the valve 103 (how much the valve opening is changed, or a single PID control. It may be an amount corresponding to the valve opening after being changed in step). When the operation amount is supplied from the neural network construction unit 116, the valve 103 has a valve opening degree corresponding to the supplied operation amount. Output value _{y 1,} since sequentially output from the neural network construction unit 116, are sequentially supplied to the valve control unit 115, the valve control unit 115 controls the sequential valve 103 by the PID calculation. For example, the neural network construction unit 116 sequentially calculates the operation amount and sequentially supplies the operation amount to the valve 103 to sequentially control the valve opening degree of the valve 103.

  The measured values of the heat medium temperature, the inlet air temperature, the inlet air humidity, the inlet air flow rate, and the outlet air temperature are sequentially supplied at predetermined intervals in real time, and according to this, the neural network construction unit 116 and the valve control unit 115 are supplied. Thus, the valve opening degree of the valve 103 is sequentially controlled, so that the valve opening degree changes in real time in response to each measurement value sequentially supplied. After the dead time elapses after the valve opening changes, each measurement value reflecting the change in the valve opening is supplied to the neural network construction unit 116. Based on the supplied measurement values, the neural network construction unit 116 and the valve control unit 115 perform the above processing, the above control is performed, and the valve opening of the valve 103 changes again.

Here, the neural network construction unit 116 causes the neural network to learn using the outlet air temperature as a teacher signal. This learning is performed by the error back propagation method. For example, the neural network construction unit 116 stores the output value y ₁ in a RAM or the like, outputs the stored output value y _1, and outputs the output value y ₁ as the output value y ₁ ( that is, since the like outlet air temperature became Moto of obtaining the output value y ₁ is supplied), and the outlet air temperature of the measurement values supplied after a dead time L, the squared error The coupling load is corrected retroactively from the output layer side to the upstream side (input layer side) so that the sum E is minimized. As described above, the dead time L is calculated from the distance H [m] from the valve 103 to the heat exchanger 107 and the flow velocity V [m / s] of the heat medium. Since it can be easily obtained from measurement or the like, the dead time L is calculated from these and is set in advance in the neural network construction unit 116 as a fixed value. In the error back propagation method, the neural network construction unit 116 corrects the weight by the gradient method. The neural network construction unit 116 performs such learning for a predetermined period, and constructs a neural network that is used during the actual operation of the heat exchange system 100 (stored in a nonvolatile storage device or the like). Update the data necessary for the neural network).

  Further, since the first connecting portion 13A is disposed in the vicinity of the valve 103, the junction of the heat medium flowing in the forward path 11 and the heat medium flowing in the circulation path 13 is located in the vicinity of the valve 103. For this reason, the heat medium whose flow rate is controlled by the valve 103 immediately reaches the position where the thermometer 121 for the heat medium near the first connection portion 13A is located. For this reason, the heat medium temperature measured by the heat medium thermometer 121 is a temperature representing the valve opening degree of the valve 103 at the time of the measurement. Since the valve opening is a value that depends on the degree of control of the valve 103 (for example, a value that depends on the amount of operation), the heat medium temperature is a value that depends on the current degree of control of the valve 103. Thus, each measured value includes a value representing the current control state of the valve 103 necessary for predicting the outlet air temperature of the dead time lapse degree.

  The controller 101 performs PID control using the neural network constructed in the trial operation when the heat exchange system 100 performs the main operation. In this case, the controller 101 operates with the configuration shown in FIG. For example, when the heat exchange system 100 performs the main operation, when the user performs an operation to instruct the operation mode of the main operation to the input / output unit 111 and an operation to specify a desired outlet air temperature as the target value, these Operation information representing an operation is supplied to the first microcomputer 112. When this operation information is supplied, the first microcomputer 112 sets a target value designated by the operation represented by the operation information (for example, target value data is held in a RAM or the like), and this target value is used. It operates as a valve control unit 115 that performs PID control on the valve 103. Further, the first microcomputer 112 supplies a predetermined signal to the second microcomputer 113 to cause the second microcomputer 113 to operate as the currently constructed neural network 117 (see above for the contents of the neural network 117).

The neural network 117 has measured values of the heat medium temperature, the inlet air temperature, the inlet air humidity, the inlet air flow rate, and the outlet air temperature from the heat medium thermometer 121, the measurement unit 123, and the outlet thermometer 125. These are sequentially supplied as a set of input values x _{1 to} x ₅ at predetermined intervals in real time, and the neural network 117 sequentially performs the above-described operations and sequentially outputs the output value y ₁ .

An output value y ₁ output from the neural network 117 is supplied to the valve control unit 115. The valve control unit 115 performs PID calculation based on the currently set target value and the sequentially supplied outlet air temperature, and controls the valve 103 according to the calculation result, thereby setting the outlet air temperature to the target value. Execute PID control to approach. Output value _{y 1,} since sequentially output from the neural network construction unit 116, are sequentially supplied to the valve control unit 115, the valve control unit 115 controls the sequential valve 103 by the PID calculation. For example, the neural network 117 sequentially calculates the operation amount and sequentially supplies the operation amount to the valve 103 to sequentially control the valve opening degree of the valve 103.

  The measured values of the heat medium temperature, the inlet air temperature, the inlet air humidity, the inlet air flow rate, and the outlet air temperature are sequentially supplied at predetermined intervals in real time, and according to this, the neural network construction unit 116 and the valve control unit 115 are supplied. Thus, the valve opening degree of the valve 103 is sequentially controlled, so that the valve opening degree changes in real time in response to each measurement value sequentially supplied. As a result, the outlet air temperature approaches the target value. In particular, since the predicted value of the outlet air temperature after the dead time L has elapsed from the present time, which is output from the neural network 117, is input to the valve control unit 115, the dead time is used in the control of the valve 103 in the actual operation. Compensated and control accuracy is improved. Furthermore, since the predicted value of the outlet air temperature after the dead time L has elapsed, not only the outlet air temperature but also the heat medium temperature, the inlet air temperature, the inlet air humidity, and the inlet air flow rate related to the outlet air temperature are considered. The accuracy of the predicted values has been improved. For this reason, the accuracy of control of the valve 103 in the actual operation is improved.

  6 to 9, when the target value is changed in the heat exchange system 100 when the valve 103 is controlled only by PID control in which the target value and the outlet air temperature are fed back without using the neural network 117. The simulation result about the time change of the outlet air temperature and the valve opening degree of is shown. FIGS. 6 and 7 show the case where the proportional gain when the valve opening of the valve 103 is changed once is large. FIGS. 8 and 9 show the proportional gain when the valve opening of the valve 103 changes once. It is a simulation result about the time change of the outlet air temperature and the valve opening degree when is small.

  FIGS. 10 and 11 show simulation results for the time variation of the outlet air temperature and the valve opening when the target value is changed when the valve 103 is controlled by the method of the above embodiment in the heat exchange system 100. . Note that the learning coefficient during learning (construction) of the neural network 117 was “0.01” and the inertia term was “0.8”.

The inlet air flow rate (air volume) is 10,000 [m ³ / h], the inlet air temperature is 30 [° C.], the inlet air humidity (relative humidity) is 60 [%], and the distance H [m] from the valve 103 to the heat exchanger 107 ] Was set to 50 [m], the flow velocity V [m / s] of the heat medium was set to 1.0 [m / s], and the dead time L was set to 50 / 1.0 = 50 [s].

  As shown in FIG. 6 to FIG. 11, by performing control by inputting the measured values of the heat medium temperature, the inlet air temperature, the inlet air humidity, the inlet air flow rate, and the outlet air temperature to the neural network 117, the outlet air temperature is Since the target value is approached more quickly than when the neural network 117 is not used, it can be seen that the control using the neural network 117 has good control accuracy. It can also be seen that the control using the neural network 117 makes it difficult for the valve opening to change significantly in a short time. For this reason, it is possible to prevent the valve 103 from deteriorating quickly.

  In the above embodiment, with the above configuration, the heat exchange system 100 changes the temperature of the heat exchange object (here, air) by performing heat exchange using the heat medium that has flowed in. The heat exchanger 107, an adjustment mechanism for adjusting the temperature of the heat medium flowing into the heat exchanger 107 (here, the valve 103, the forward path 11, the heat source device 109, the circulation path 13, etc.), and the controller 101 that controls the adjustment mechanism (A controller 101 for controlling the valve 103).

  In the above embodiment, the controller 101 (valve control unit 105) has a control amount (adjustment mechanism (here, the valve 103) for changing the temperature of the heat medium flowing into the heat exchanger 107 by the above configuration. Information to specify the content of the control for the adjustment mechanism (in this case, the operation amount) is supplied to the adjustment mechanism to control the adjustment mechanism. Changes the temperature of the heat medium flowing into the heat exchanger according to the control amount supplied from the controller 101 (here, the temperature of the heat medium is changed by changing the valve opening degree of the valve 103). become.

  In the above embodiment, the controller 101 measures the predetermined physical quantities (in this case, the inlet air temperature, the inlet air humidity, and the inlet air flow rate) before heat exchange of the heat exchange object in the above configuration. Neural network to which a value, a measured value obtained by measuring the temperature after heat exchange of the heat exchange target object (here, outlet air temperature), and an input value (here, heat medium temperature) representing a control amount are input Based on these, the neural network 117 that predicts the temperature after heat exchange of the heat exchange target after the dead time has elapsed (here, the outlet air temperature) and outputs it as a predicted value, and the neural network 117 The output predicted value is compared with a predetermined target value (including a target value set by the user operating the input / output unit 111), and based on the comparison result, the heat exchange target Temperature after the heat exchange is, the control amount change means for changing a control amount supplied to the adjusting mechanism so as to approach the target value (in this case, the valve control unit 115 which performs PID control) and, on the secondary side.

  In the present embodiment, the neural network 117 outputs a predicted value after the dead time has elapsed based on each input value, and the controlled variable changing means changes the controlled variable based on the predicted value. Since each dead time is compensated, and each value that is the basis of the predicted value includes a predetermined physical quantity before the heat exchange of the heat exchange object that affects the heat exchange control, accurate control is possible. It becomes possible. In addition, control that greatly changes the adjustment mechanism (here, control of the valve 103) is not required.

  In particular, in the above-described embodiment, the heat exchange object is air, and the heat exchanger 107 performs heat exchange between the air flowing in from the outside and the heat medium, sends the air after heat exchange to the outside, and performs heat exchange. The measured value obtained by measuring a predetermined physical quantity before heat exchange of the object may be air flowing in from outside (air before flowing in depending on the position of the measuring unit 123, or air after flowing in) May be measured values of the temperature, humidity, and flow rate of the heat exchange target, and the measured values of the temperature after the heat exchange of the heat exchange object are the air sent to the outside (the position of the outlet thermometer 125) The measured air temperature may be air before flowing in or air after flowing in.). As a result, particularly in the control of the temperature of the air, the control is performed in consideration of the temperature, the humidity, and the flow rate that have an influence, so that it is possible to perform the control with high accuracy. In controlling the temperature of the air, among the temperature, humidity, and flow rate of the air flowing in from the outside, the temperature of the air is particularly important, so a predetermined physical quantity before heat exchange of the heat exchange object was measured. Even if the measured value is at least one of the temperature, humidity, and flow rate of the air flowing in from the outside, the control can be performed with a certain degree of accuracy.

  In particular, in the above-described embodiment, the input value representing the control amount is obtained by measuring the temperature of the heat medium changed according to the supplied control amount within a predetermined period (immediately after supply) after the control amount is supplied to the adjustment mechanism. Therefore, the value input to the neural network 117 can be easily obtained by measuring the temperature of the heat medium. Moreover, since the temperature of the heat medium directly affects the outlet air temperature, the measurement accuracy of the control can be improved by using the measured value of the temperature of the heat medium as an input value representing the control amount. it can. In particular, even if the valve opening degree of the valve 103 is the same, the temperature of the heat medium is the temperature of the pipe through which the heat medium flows, or the temperature of the heat medium before joining that flows through the circulation path 13 toward the first connection portion 13A. (Still, the temperature of the heat medium does not change to represent the controlled variable.) By using the measured value of the temperature of the heat medium as an input value that represents the controlled variable, Accuracy can be improved. The input value representing the control amount is measured by various sensors and input to the neural network 117, or is input from the valve control unit 115 to the neural network 117 based on the operation amount of the valve control unit 115, etc. The valve opening may be 103.

  In particular, in the above-described embodiment, a measurement value obtained by measuring a predetermined physical quantity before heat exchange of the heat exchange object, obtained by operating the heat exchange system 100 (here, obtained by trial operation), and heat A measured value obtained by measuring the temperature of the object to be exchanged after heat exchange, a value indicating a control amount (for example, a value similar to the input value indicating the control amount, the temperature of the heat medium, the valve opening of the valve 103, etc.), The construction means for constructing the neural network 117 (here, the neural network construction unit 116) is realized on the basis of the neural network 117, and the neural network 117 is obtained by operating the heat exchange system 100. The accuracy of control using is improved. The construction means for constructing the neural network 117 may be realized by a server or the like that can communicate with the controller 101. For example, when the heat exchanging system 100 is performing a trial operation (may be a main operation), the controller 101 supplies each measurement value input to the controller 101 to the server, and the server supplies the measurement value from the controller 101. Using each measured value, a neural network is constructed in the same manner as the neural network construction unit 117, and the constructed neural network is supplied to the controller 101. The controller 101 is a neural network that uses the supplied neural network thereafter. The network 117 may be used.

  Moreover, in the said embodiment, by the said structure, an adjustment mechanism makes the 1st heat medium join the outgoing path 11 through which the 1st heat medium adjusted to predetermined | prescribed temperature by the heat source apparatus 109 and a 1st heat medium join, and is 1st. The circulation path 13 through which the second heat medium having a temperature corresponding to the inflow amount of the heat medium flows, and the valve opening degree according to the control amount (in this case, the operation amount) supplied from the controller 101. 13, and a valve 103 that adjusts the flow rate of the first heat medium flowing into the control unit 13. The input value that represents the control amount is a value that represents the valve opening. This facilitates control of the temperature of the heat medium flowing into the heat exchanger 107.

  In the above embodiment, according to the above configuration, the adjusting mechanism causes the heat exchanger 107 to heat the flow rate of the second heat medium flowing through the circulation path 13 so that the flow rate is constant regardless of the inflow amount of the first heat medium. A return path 15 into which a part of the exchanged second heat medium flows is further provided. This facilitates control of the temperature of the heat medium flowing into the heat exchanger 107.

  Moreover, in the said embodiment, with the said structure, the input value showing control amount is the said 2nd of the temperature according to the inflow amount of the 1st heat carrier which flows through 13 C of 1st connection parts in the circulation path 13. FIG. The valve 103 is arranged in the vicinity of the first connection portion 13A, and the value easily input to the neural network 117 by the measurement of the temperature of the heat medium. Can be obtained. Moreover, since the temperature of the heat medium directly affects the outlet air temperature, the measurement accuracy of the control can be improved by using the measured value of the temperature of the heat medium as an input value representing the control amount. it can. In particular, even if the valve opening degree of the valve 103 is the same, the temperature of the heat medium is the temperature of the pipe through which the heat medium flows, or the temperature of the heat medium before joining that flows through the circulation path 13 toward the first connection portion 13A. (Still, the temperature of the heat medium does not change to represent the controlled variable.) By using the measured value of the temperature of the heat medium as an input value that represents the controlled variable, Accuracy can be improved.

  The present invention is not limited to the above embodiment, and the above embodiment can be modified as appropriate. Although the example is shown below as a modification, the following modification can also combine at least one part.

(Modification 1)
The valve 103 may be a three-way valve. Even if the valve 103 is a three-way valve, the above configuration and operation are the same as appropriate.

(Modification 2)
The controller 101 may be configured by a single computer. In this case, one computer operates as the neural network construction unit 115, the valve control unit 116, the neural network 117, and the like. Further, at least a part of the controller 101 may be configured by a dedicated electronic circuit that does not depend on a program.

(Modification 3)
At the time of starting up the heat exchange system 100, if there is a dead time L, in normal PID control, the valve starts to open slowly and undershoot and overshoot are likely to occur, but on the other hand, the neural network 117 is used. In the conventional control, it is possible to quickly start up by predicting the state after the dead time L. Of the inputs to the neural network 117 at this time, the inlet air flow rate is the inlet air flow rate after the dead time L has elapsed from that time (current). If the rise time of the heat exchanger 107 is determined in advance, the inlet air flow rate after the dead time L (which may change depending on the result time from the rise time) can be obtained from calculation. . For this reason, the neural network 117 uses the inlet air flow rate after the dead time L has passed as a fixed value (which may be a fixed value that changes according to the result time from the start-up time). The measured value of the air flow rate may not be input to the neural network 117. The inlet air flow rate after the dead time L has elapsed may be obtained by calculation or the like when processing by the neural network 117 is performed. That is, the inlet air flow rate after the dead time L has elapsed may be set in advance or may be obtained by calculation or the like. The inlet air flow rate after the dead time L has elapsed may be supplied to the neural network 117 from the outside. The neural network 117 obtains an inlet air flow rate after a dead time L has been obtained, which is obtained by calculation, which is set in advance, or is supplied from the outside (consisting of a nonvolatile storage device of a microcomputer, a ROM, etc. It may be used after being read from the storage unit, obtained by calculation, or supplied from the outside.

12 and 13 show simulation results when the heat exchange system 100 is started up. The inlet air flow rate (air volume) is 10,000 [m ³ / h], the inlet air temperature is 30 [° C.], the inlet air humidity (relative humidity) is 60 [%], and the distance H [m] from the valve 103 to the heat exchanger 107 ] Is set to 50 [m], the flow velocity V [m / s] of the heat medium is set to 1.0 [m / s], and the dead time L is set to 50 / 1.0 = 50 [s]. The initial temperature to flow in is 25 [° C.], and the heat exchange system 100 is started up at a rate of 180 [s] in proportion to an air volume of 10,000 [m ^{3 /} h] (3.333 [kg / s]). To do.

  As shown in FIGS. 12 and 13, by performing control using the neural network 117, the outlet air temperature is higher than when the neural network 117 is not used (a graph with a large proportional gain and a small proportional gain). Since the target value is quickly approached, it can be seen that the control using the neural network 117 has good control accuracy. It can also be seen that the control using the neural network 117 makes it difficult for the valve opening to change significantly in a short time.

  Thus, the measured value obtained by measuring a predetermined physical quantity of the heat exchange object before heat exchange is a measured value obtained by measuring at least the temperature of the temperature and humidity of the air flowing from the outside, and the neural network is wasteful. By acquiring the flow rate of the air flowing in from the outside after the lapse of time, and further using the acquired flow rate after the lapse of the dead time, making a prediction, even at the time of starting up the heat exchange system, Accurate control is possible. Further, it is not necessary to perform control that greatly changes the adjustment mechanism.

(Modification 4)
For example, the controller 101 acquires a plurality of neural networks 117 having different characteristics and information indicating the operation state of the heat exchange system 100, and based on the acquired information indicating the operation state, A selection unit 118 that selects one as a neural network to be actually used, and the neural network 117 to be used may be changed according to the operation state of the heat exchange system 100 (see, for example, FIG. 14). . These are comprised by the 2nd microcomputer 113, for example. In this way, by switching the neural network 117 to be used according to the operating state of the heat exchange system 100, the neural network 117 having favorable characteristics can be used depending on the operating state of the heat exchange system 100, and the control accuracy is further improved. To do. In particular, when the operating state of the heat exchanging system 100 changes suddenly, this sudden change can be dealt with by switching the neural network 117 to be used.

  The operation state of the heat exchange system 100 includes, for example, a valve opening range of the valve 103, an inlet air temperature range, a startup time, a normal time, and the like. The driving state may be a combination of different types. In this case, the heat exchange system 100 is trial-run so that the heat exchange system 100 is in each operation state, and each neural network 117 is constructed by the above method. Thereby, a plurality of specific different neural networks 117 are constructed.

  As an example, the selection unit 118 acquires the valve opening of the valve 103 by various sensors, for example, and selects the neural network A when the valve opening is in the range of 0% to less than 50%, When the opening is in the range of 51% or more and 100%, the neural network B may be selected and each measured value may be input to the selected neural network. In this case, the selected neural network outputs a predicted value.

  As another example, for example, when the inlet air temperature is acquired and the inlet air temperature is in the range of 0 ° C. or higher and lower than 20 ° C., the selection unit 118 selects the neural network A, and the inlet air temperature is When the temperature is in the range of 20 ° C. or higher and lower than 40 ° C., the neural network B may be selected and each measured value may be input to the selected neural network. In this case, the selected neural network outputs a predicted value.

  As another example, for example, the selection unit 118 determines that a predetermined period set in advance from the startup of the heat exchange system 100 such as when the power is turned on is the startup (in this case, the operation at the startup) The state is acquired.), A neural network A (for example, the neural network 117 of the modified example 3) is selected, and a predetermined period of time elapses after the heat exchange system 100 is started up such as when the power is turned on. The later period is determined as normal time (in this case, it is assumed that the operation state of normal time is acquired), the neural network B (for example, the neural network 117 of the above embodiment) is selected, and each measured value May be input to the selected neural network. In this case, the selected neural network outputs a predicted value.

(Modification 5)
The heat medium may be other than water, and the heat exchange object may be other than air.

DESCRIPTION OF SYMBOLS 11 Outbound path 13 Circulation path 15 Return path 100 Heat exchange system 101 Controller 103 Valve 105 Pump 107 Heat exchanger 109 Heat source apparatus 121 Thermometer for heat medium 123 Measurement unit 125 Outlet thermometer

Claims (10)

A heat exchanger in which a heat medium flows in and changes the temperature of the heat exchange object by performing heat exchange using the heat medium that has flowed in;
An adjustment mechanism for adjusting the temperature of the heat medium flowing into the heat exchanger;
Control means for controlling the adjustment mechanism;
A heat exchange system comprising:
The control means controls the adjustment mechanism by supplying a control amount for changing the temperature of the heat medium flowing into the heat exchanger to the adjustment mechanism,
The adjustment mechanism changes the temperature of the heat medium flowing into the heat exchanger according to the control amount supplied from the control means,
The control means includes an input value representing a measurement value obtained by measuring a predetermined physical quantity of the heat exchange object before heat exchange, and an input value representing a measurement value obtained by measuring the temperature after heat exchange of the heat exchange object. The heat after the dead time has elapsed from when the control amount is supplied to the adjustment mechanism until the temperature of the heat medium flowing into the heat exchanger changes based on the input value representing the control amount. The neural network that predicts the temperature after heat exchange of the object to be exchanged and outputs it as a predicted value, and the predicted value output by the neural network is compared with a predetermined target value, based on the comparison result, Control amount changing means for changing the control amount to be supplied to the adjustment mechanism so that the temperature after heat exchange of the heat exchange object approaches the target value,
A heat exchange system characterized by that. A plurality of neural networks having different characteristics;
Selection means for acquiring information representing the operating state of the heat exchange system, and selecting any one of the plurality of neural networks as a neural network that is actually used based on the acquired information representing the operating state; With
The control amount changing means controls the control amount by comparing the predicted value output from the neural network selected by the selection means with the target value.
The heat exchange system according to claim 1. The heat exchange object is air,
The heat exchanger performs heat exchange between the air flowing from the outside and the heat medium, and sends the air after heat exchange to the outside.
The measurement value obtained by measuring a predetermined physical quantity of the heat exchange object before heat exchange is a measurement value obtained by measuring at least the temperature of the temperature, humidity, and flow rate of air flowing from the outside,
The measured value obtained by measuring the temperature after heat exchange of the heat exchange object is a measured value obtained by measuring the temperature of the air sent to the outside.
The heat exchange system according to claim 1 or 2, wherein The measurement value obtained by measuring a predetermined physical quantity before heat exchange of the heat exchange object is a measurement value obtained by measuring at least the temperature of the temperature and humidity of the air flowing from the outside,
The neural network obtains a flow rate of air flowing from the outside after the dead time has elapsed, and further uses the obtained flow rate after the dead time has elapsed to perform the prediction.
The heat exchange system according to claim 3. The input value representing the control amount is a value representing a measurement value obtained by measuring the temperature of the heat medium changed according to the control amount within a predetermined period after the control amount is supplied to the adjustment mechanism.
The heat exchange system according to any one of claims 1 to 3, wherein A measurement value obtained by measuring the predetermined physical quantity before heat exchange of the heat exchange object obtained by operating the heat exchange system, a measurement value obtained by measuring the temperature after heat exchange of the heat exchange object, A construction means for constructing the neural network based on the value representing the control amount;
The heat exchange system according to any one of claims 1 to 5, wherein The adjustment mechanism is connected to the first flow path through which the first heat medium adjusted to a predetermined temperature flows, and the first heat medium from the first flow path. By flowing in, the second heat medium having a temperature corresponding to the inflow amount of the first heat medium flows through the second flow path connected to the adjustment mechanism, and disposed in the first flow path. A valve that adjusts the flow rate of the first heat medium flowing into the second flow path by becoming a valve opening according to a control amount supplied from the control means,
The input value representing the control amount is a value representing the valve opening.
The heat exchange system according to any one of claims 1 to 6, wherein In the adjustment mechanism, the heat exchanger performs heat exchange so that the flow rate of the second heat medium in the second flow path is constant regardless of the inflow amount of the first heat medium. A third flow path into which a part of the heat medium of 2 flows,
The heat exchange system according to claim 7. The input value representing the control amount is in accordance with the inflow amount of the first heat medium that flows in the vicinity of the connection portion of the second flow path between the first flow path and the second flow path. Represents the measured value of the temperature of the second heat medium,
The valve is disposed in the vicinity of the connection portion in the first flow path.
The heat exchange system according to claim 7 or 8, characterized in that. A heat exchanger in which a heat medium flows in and changes the temperature of the heat exchange object by performing heat exchange using the heat medium that has flowed in;
An adjustment mechanism for adjusting the temperature of the heat medium flowing into the heat exchanger, and a controller for controlling the adjustment mechanism in a heat exchange system,
The controller controls the adjustment mechanism by supplying a control amount for changing the temperature of the heat medium flowing into the heat exchanger to the adjustment mechanism;
The adjustment mechanism changes the temperature of the heat medium flowing into the heat exchanger according to the control amount supplied from the controller,
The controller is an input value representing a measured value obtained by measuring a predetermined physical quantity of the heat exchange object before heat exchange; an input value representing a measured value obtained by measuring the temperature after heat exchange of the heat exchange object; The heat exchange after a lapse of dead time from the supply of the control amount to the adjustment mechanism to the change of the temperature of the heat medium flowing into the heat exchanger based on the input value representing the control amount A neural network that predicts the temperature after heat exchange of an object and outputs it as a predicted value, and compares the predicted value output from the neural network with a predetermined target value, and based on the comparison result, Control amount changing means for changing the control amount to be supplied to the adjustment mechanism so that the temperature after heat exchange of the exchange object approaches the target value,
A controller characterized by that.