Heat source system, target operating capacity estimation method, target operating capacity estimation program

Abstract

To improve operating efficiency of a heat source system.SOLUTION: A cooling tower control device 30 for a heat source system 1 includes: a reinforcement learning section 34 learning a relation between input of an operating condition and operational capability of a cooling tower 20 and output of operation capacity; a state observation section 31 acquiring state variables at least including the operating condition, the operational capability and the operation capacity; an evaluation data acquisition section 32 acquiring energy consumption of the cooling tower 20 and a cooling water pump 40 for evaluating a control result of the cooling tower 20; and a reward calculation section 33 calculating reward on the basis of the energy consumption. The reinforcement learning section 34 acquires a relation between input and output of a target operation capacity model 36 through reinforcement learning by using the reward calculated by the reward calculation section 33 so that the energy consumption is minimized in the operating condition and a condition of the operational capability.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present disclosure relates to a heat source system, a method for estimating a target operating capacity, and a program for estimating a target operating capacity.

Background Art

[0002] Conventionally, in a heat source system used for chilled water supply in a semiconductor factory or district heating and cooling, a plurality of cooling towers are provided so as to have an operating capacity corresponding to the maximum load of the refrigerator for cooling an external load (a cooled part) (see, for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the heat source system provided with a plurality of cooling towers described in Patent Document 1, it is desirable to be able to appropriately select the number of operating cooling towers in order to improve the operating efficiency.

[0005] An object of the present disclosure is to provide a heat source system, a method for estimating a target operating capacity, and a program for estimating a target operating capacity that can improve the operating efficiency.

Means for Solving the Problems

[0006] A heat source system according to a first aspect of the present disclosure comprises a unit to be cooled, a cooling water pump that supplies cooling water to the unit to be cooled, and a cooling tower that cools the cooling water introduced from the unit to be cooled by the cooling water pump by bringing it into contact with the outside air, wherein a plurality of cooling towers are provided so as to have an operating capacity corresponding to the maximum load of the unit to be cooled, and comprises a reinforcement learning unit that learns the relationship between the input of the operating conditions and operating capacity of the cooling tower and the output of the operating capacity, a state observation unit that acquires state variables including at least the operating conditions, the operating capacity and the operating capacity, an evaluation data acquisition unit that acquires the energy consumption of the cooling tower and the cooling water pump to evaluate the control result of the cooling tower, and a reward calculation unit that calculates a reward based on the energy consumption, wherein the reinforcement learning unit acquires the relationship between the input and the output by reinforcement learning so as to minimize the energy consumption under the operating conditions and operating capacity conditions, using the reward calculated by the reward calculation unit.

[0007] According to a first aspect of this disclosure, a heat source system capable of improving operating efficiency can be provided.

[0008] In a heat source system according to a second aspect of the present disclosure, the part to be cooled is a refrigerator.

[0009] In a heat source system according to a third aspect of this disclosure, the reward calculation unit calculates a higher reward the lower the energy consumption.

[0010] According to a third aspect of this disclosure, the reinforcement learning unit can obtain higher rewards as energy consumption decreases, thereby enabling it to quickly and appropriately learn the relationship between the input of cooling tower operating conditions and operating capacity and the output of the operating capacity.

[0011] A heat source system according to a fourth aspect of the present disclosure comprises an external load and a chilled water pump that supplies chilled water cooled by heat exchange within the cooled section to the external load, wherein the state variables and the input include the load of the cooled section, the evaluation data acquisition unit acquires the energy consumption of the cooled section and the chilled water pump in addition to the energy consumption of the cooling tower and the chilled water pump, and the reward calculation unit calculates the reward based on the energy consumption of the cooling tower and the chilled water pump and the energy consumption of the cooled section and the chilled water pump.

[0012] According to a fourth aspect of this disclosure, in addition to the energy consumption of the cooling tower and chilled water pump, the energy consumption of the cooled part and chilled water pump can also be taken into consideration, and the energy consumption of the entire heat source system can be reflected in the reward calculation. This allows the reinforcement learning unit to proceed in a more appropriate direction and enables the reinforcement learning unit to estimate the operating capacity with higher accuracy.

[0013] In a heat source system according to a fifth aspect of this disclosure, the state variables and the input or output include the temperature of the cooling water.

[0014] According to a fifth aspect of this disclosure, when performing control that varies the cooling water temperature, the learning process can be carried out while taking the cooling water temperature into consideration, thereby improving the learning accuracy.

[0015] A method for estimating a target operating capacity according to a sixth aspect of the present disclosure is a heat source system comprising a unit to be cooled, a cooling water pump that supplies cooling water to the unit to be cooled, and a cooling tower that cools the cooling water introduced from the unit to be cooled by the cooling water pump by bringing it into contact with the outside air, wherein a plurality of cooling towers are provided so as to have an operating capacity corresponding to the maximum load of the unit to be cooled, wherein the cooling tower control device performs the steps of: acquiring state variables including at least the operating conditions, operating capacity, and operating capacity of the cooling tower; acquiring the energy consumption of the cooling tower and the cooling water pump to evaluate the control result of the cooling tower; calculating a reward based on the energy consumption; and using the calculated reward, acquiring the relationship between the input of the operating conditions and operating capacity of the cooling tower and the output of the operating capacity by reinforcement learning so as to minimize the energy consumption under the operating conditions and operating capacity.

[0016] A target operating capacity estimation program according to a sixth aspect of the present disclosure is a heat source system comprising a unit to be cooled, a cooling water pump that supplies cooling water to the unit to be cooled, and a cooling tower that cools the cooling water introduced from the unit to be cooled by the cooling water pump by bringing it into contact with the outside air, wherein a plurality of cooling towers are provided so as to have an operating capacity corresponding to the maximum load of the unit to be cooled, and the program causes the cooling tower control device to perform the following steps: acquire state variables including at least the operating conditions, operating capacity, and operating capacity of the cooling tower; acquire the energy consumption of the cooling tower and the cooling water pump to evaluate the control result of the cooling tower; calculate a reward based on the energy consumption; and use the calculated reward to acquire, by reinforcement learning, the relationship between the input of the operating conditions and operating capacity of the cooling tower and the output of the operating capacity, such that the energy consumption is minimized under the operating conditions and operating capacity. [Brief explanation of the drawing]

[0017] [Figure 1] This figure shows a schematic configuration of the heat source system according to the embodiment. [Figure 2]It is a diagram showing an example of the hardware configuration of a cooling tower control device. [Figure 3] It is a diagram for explaining the functional configuration of the cooling tower control device according to the first embodiment. [Figure 4] It is a flowchart showing the flow of the target operating capacity estimation process according to the embodiment. [Figure 5] It is a diagram showing an example of the simulation result of the system COP of the heat source system. [Figure 6] It is a diagram for explaining the functional configuration of the cooling tower control device according to the second embodiment.

Mode for Carrying Out the Invention

[0018] Hereinafter, embodiments will be described with reference to the accompanying drawings. For ease of understanding of the description, the same reference numerals are given to the same components in each drawing as much as possible, and duplicate descriptions are omitted.

[0019] [First Embodiment] The first embodiment will be described with reference to FIGS. 1 to 4. First, the overall configuration of the heat source system 1 will be described with reference to FIG. 1. FIG. 1 is a diagram showing a schematic configuration of the heat source system 1 according to the embodiment.

[0020] As shown in FIG. 1, the heat source system 1 includes a refrigerator 10 (cooled part), three cooling towers 20a, 20b, 20c that are connected in parallel to the refrigerator 10 and have fans rotating for cooling, and a cooling tower control device 30 that controls the cooling towers 20a, 20b, 20c. In FIG. 1, the case of having three cooling towers is illustrated, but the number of cooling towers is not limited. In the following, unless otherwise specified, the cooling tower is described as the cooling tower 20.

[0021] The refrigerator 10 includes a condenser 11 and an evaporator 12. The condenser 11 is connected to the cooling water pipes of the supply pipe 41 and the return pipe 42, and the cooling water W1 cooled by the cooling towers 20a, 20b, 20c is fed through the cooling water pipe, and the refrigerant is cooled by the cooling water W1.

[0022] The evaporator 12 is connected to chilled water pipes 51 and 52, and chilled water W2 is supplied from the external load 50 via the chilled water pipes 51 and 52. The chilled water W2 and refrigerant are heated and evaporated. The chilled water W2 cooled in the evaporator 12 is supplied to the external load 50.

[0023] The external load 50 is, for example, an air conditioner (HVAC unit). The external load 50 dissipates heat to the chilled water W2 and then recirculates the chilled water W2 back to the chiller 10.

[0024] Each cooling tower 20a, 20b, and 20c cools the cooling water W1 by the rotation of a fan. Each cooling tower 20a, 20b, and 20c is connected to the chiller 10 via a supply pipe 41, to which the cooling water W1 used by the chiller 10 is supplied, and the cooling water W1 cooled in the cooling towers 20 is supplied to the chiller 10 via a return pipe 42.

[0025] A supply water header 43 is provided in the supply piping 41, and a return water header 44 is provided in the return piping 42. A cooling water pump 40 is provided downstream of the return water header 44 in the return piping 42. By adjusting the output of the cooling water pump 40, the cooling water W1 is circulated between the cooling towers 20a, 20b, 20c and the condenser 11. A bypass piping 45 is provided between the supply water header 43 and the return water header 44. A cooling water bypass valve (bypass valve) 46 is provided in the bypass piping 45. By adjusting the opening of the cooling water bypass valve 46, the flow rate bypassed from the supply water header 43 to the return water header 44 is adjusted.

[0026] In the return piping 42, a temperature sensor 22 is provided downstream of the return water header 44 in the cooling water flow to measure the temperature of the cooling water W1 (cooling water inlet temperature T1) that is cooled in the cooling towers 20a, 20b, and 20c and flows into the chiller 10. The information of the cooling water inlet temperature T1 measured by the temperature sensor 22 is output to the cooling tower control device 30.

[0027] Similarly, in the supply piping 41, a temperature sensor 23 is provided upstream of the water supply header 43 in the cooling water flow to measure the temperature of the cooling water W1 after heat exchange with the refrigerant in the chiller 10 (cooling water outlet temperature T2). The information of the cooling water outlet temperature T2 measured by the temperature sensor 23 is output to the cooling tower control device 30.

[0028] The heat source system 1 is equipped with wet-bulb thermometers 21a, 21b, and 21c corresponding to each cooling tower 20a, 20b, and 20c. The wet-bulb thermometers 21a, 21b, and 21c measure the wet-bulb temperature of the outside air and output the measured wet-bulb temperature information to the cooling tower control device 30.

[0029] A temperature sensor 24 is provided in the chilled water piping 52 that supplies chilled water W2 from the chiller 10 to the external load 50. The temperature sensor 24 measures the temperature T3 of the chilled water W2 flowing into the external load 50. The chilled water temperature T3 information measured by the temperature sensor 24 is output to the cooling tower control device 30.

[0030] Similarly, a temperature sensor 25 is provided in the chilled water piping 51 that returns chilled water W2 from the external load 50 to the chiller 10 to measure the temperature T4 of the chilled water W2 after heat exchange at the external load 50. The chilled water temperature T4 information measured by the temperature sensor 25 is output to the cooling tower control device 30.

[0031] A chilled water pump 47 is provided in the chilled water piping 52. By adjusting the output of the chilled water pump 47, chilled water W2 is circulated between the chiller 10 and the external load 50.

[0032] The cooling tower control device 30 calculates the target operating capacity of the cooling tower 20, calculates the number of cooling towers that need to be operated to achieve the target operating capacity, and controls the operation of the cooling towers according to this number.

[0033] Referring to Figure 5, we will now explain the reason for controlling the number of cooling towers 20 in the heat source system 1. Figure 5 shows an example of the simulation results of the system COP (Coefficient of Performance) of the heat source system 1. The system COP is the overall efficiency of the heat source system 1. In Figure 5, the horizontal axis shows the partial load factor of the chiller 10, and the vertical axis shows the system COP. In Figure 5, the characteristics of the system COP according to the partial load factor are plotted with the same mark for each ambient wet-bulb temperature. The solid line shows the characteristics when the cooling tower capacity is 100%, and the dashed line shows the characteristics when the cooling tower capacity is 300%.

[0034] As shown in Figure 5, the system COP tends to increase when the ambient wet-bulb temperature is 8°C or lower, by increasing the cooling tower capacity from 100% to 300% at all chiller partial load rates. On the other hand, when the ambient wet-bulb temperature is 12°C or higher, the system COP with a 300% cooling tower capacity is higher than that with a 100% capacity in the region of high chiller partial load rates, but the system COP with a 100% cooling tower capacity is higher than that with a 300% capacity in the region of low chiller partial load rates. Thus, the chiller partial load rate at which the relationship between the system COP at a 300% cooling tower capacity and the system COP at a 100% cooling tower capacity is reversed tends to shift towards the higher load side as the ambient wet-bulb temperature increases.

[0035] Therefore, in order to improve the operating efficiency of the heat source system 1, the optimal operating capacity of the cooling tower 20 must be obtained by taking into account parameters such as operating conditions like the outside air wet-bulb temperature and operating capacity like the chiller load factor.

[0036] In conventional technologies such as those described in Patent Document 1, a table was created to determine the COP of the entire heat source system based on the relationship between the outside wet-bulb temperature and the chiller load factor, as shown in Figure 5. From this table, parameters used in a calculation formula that maximizes the COP of the entire heat source system were determined, and the number of cooling towers in operation was controlled based on the calculation results. This presented problems such as requiring storage capacity to prepare and maintain the table in the system beforehand, and incurring computational costs due to the complex calculation process.

[0037] In contrast, in this embodiment, the cooling tower control device 30 uses a learning system to estimate the appropriate number of operating cooling towers 20 based on parameters such as operating conditions and operating capacity acquired from each position of the heat source system 1. This reduces memory capacity and computation costs, and easily improves operating efficiency.

[0038] The hardware configuration of the cooling tower control device 30 according to the embodiment will be described below. Figure 2 is a diagram showing an example of the hardware configuration of the cooling tower control device 30.

[0039] The cooling tower control device 30 has a CPU (Central Processing Unit) 101, a ROM (Read Only Memory) 102, and a RAM (Random Access Memory) 103. The CPU 101, ROM 102, and RAM 103 form a so-called computer. The cooling tower control device 30 also has an auxiliary storage device 104, an output device 105, an input device 106, an I / F (Interface) device 107, and a drive device 108. Each piece of hardware in the cooling tower control device 30 is interconnected via a bus 109.

[0040] The CPU 101 is a computing device that executes various programs (for example, a target operating capacity estimation program) installed in the auxiliary storage device 104. The ROM 102 is non-volatile memory. The ROM 102 functions as the main memory device and stores various programs and data necessary for the CPU 101 to execute the various programs installed in the auxiliary storage device 104. Specifically, the ROM 102 stores boot programs such as the BIOS (Basic Input / Output System) and EFI (Extensible Firmware Interface).

[0041] RAM103 is a volatile memory such as DRAM (Dynamic Random Access Memory) or SRAM (Static Random Access Memory). RAM103 functions as the main memory device and provides a work area that is expanded when various programs installed in the auxiliary storage device 104 are executed by the CPU 101.

[0042] The auxiliary storage device 104 stores various programs and information used when these programs are executed.

[0043] The output device 105 is an output device that outputs information acquired by the cooling tower control device 30. The output device 105 may be, for example, a display device. The input device 106 is, for example, an input device for performing various operations on the cooling tower control device 30. The I / F device 107 is a communication device for the cooling tower control device 30 to communicate with other equipment.

[0044] The drive device 108 is a device for setting the recording medium 110. The recording medium 110 here includes media that record information optically, electrically, or magnetically, such as CD-ROMs, flexible disks, and magneto-optical disks. The recording medium 110 may also include semiconductor memory that records information electrically, such as ROMs and flash memory.

[0045] Furthermore, the various programs to be installed on the auxiliary storage device 104 are installed, for example, when the distributed recording medium 110 is set in the drive device 108 and the various programs recorded on the recording medium 110 are read by the drive device 108. Alternatively, the various programs to be installed on the auxiliary storage device 104 may be installed by downloading them from the network.

[0046] Next, the functional configuration of the cooling tower control device 30 will be described with reference to Figure 3. Figure 3 is a diagram illustrating the functional configuration of the cooling tower control device 30 according to the first embodiment.

[0047] In this embodiment, the cooling tower control device 30 achieves the functions shown in Figure 3 by, for example, the CPU 101 reading and executing a target operating capacity estimation program installed in the auxiliary storage device 104.

[0048] The cooling tower control device 30 comprises a state observation unit 31, an evaluation data acquisition unit 32, a reward calculation unit 33, a reinforcement learning unit 34, and a cooling tower control unit 35.

[0049] The state observation unit 31 acquires state variables of the heat source system 1 and outputs them to the reinforcement learning unit 34. The state variables are data that indicates the current state of the environment in which the learning target exists, and in this embodiment, they include at least the operating conditions, operating capacity, and operating capacity of the cooling tower 20.

[0050] The operating conditions of the cooling tower 20 include, for example, information on the current ambient wet-bulb temperature (outside air temperature) of the cooling tower 20, and can be obtained from the measurements of the wet-bulb thermometers 21a, 21b, and 21c.

[0051] The operating capacity of the cooling tower 20 is the current load of the cooling tower 20, and is the amount of cooling heat that the cooling tower 20 exchanges with the chiller 10. The operating capacity can be calculated, for example, by the product of the supply and return water temperature difference (the temperature difference of the cooling water at the chiller inlet and outlet) and the cooling water flow rate (for example, calculated from the voltage and frequency of the inverter of the cooling water pump 40, or measured by a flow meter). The supply and return water temperature difference can be calculated, for example, using the difference between the cooling water inlet temperature T1 and the cooling water outlet temperature T2 measured by temperature sensors 22 and 23.

[0052] The operating capacity of the cooling towers 20 can be calculated, for example, by multiplying the number of cooling towers currently in operation by the capacity of each cooling tower. The number of operating towers can be obtained, for example, from the information on the number of cooling towers 20 currently in operation by the cooling tower control unit 35 based on the output of the target operating capacity model 36.

[0053] The reward calculation unit 33 calculates the reward that the reinforcement learning unit 34 uses when performing reinforcement learning. The reward calculation unit 33 calculates the reward based on the energy consumed. In this embodiment, the reward calculation unit 33 calculates a higher reward the lower the energy consumed.

[0054] The evaluation data acquisition unit 32 acquires the energy consumption of the cooling tower 20 and the cooling water pump 40 as evaluation data to evaluate the control results of the cooling tower 20, and outputs it to the reward calculation unit 33. Here, "energy consumption" includes at least one of power consumption, energy efficiency (COP; Coefficient of Performance), CO2 emissions, and energy cost. This allows the reinforcement learning unit 34 to reflect the energy consumption in the reward used when performing reinforcement learning.

[0055] The reinforcement learning unit 34 performs reinforcement learning on a model for outputting a target operating capacity, which is a target value of the operating capacity, based on state variables input from the state observation unit 31. The reinforcement learning unit 34 learns the relationship between the input of the operating conditions and operating capacity of the cooling tower 20 and the output of the target operating capacity. The reinforcement learning unit 34 also transmits the target operating capacity obtained through reinforcement learning to the cooling tower control unit 35.

[0056] The reinforcement learning unit 34 has a target operating capacity model 36. Using the reward calculated by the reward calculation unit 33, the reinforcement learning unit 34 modifies the model parameters of the target operating capacity model 36 so as to maximize the reward, that is, so as to minimize energy consumption under the operating conditions and operating load conditions, thereby acquiring the relationship between the input operating conditions and operating load of the cooling tower 20 and the output operating capacity through reinforcement learning. The target operating capacity model 36 is, for example, a multilayer neural network.

[0057] Furthermore, the reinforcement learning unit 34 outputs the target operating capacity by inputting state variables acquired by the state observation unit 31 into the target operating capacity model 36, whose model parameters have been modified. The reinforcement learning unit 34 transmits the target operating capacity output from the target operating capacity model 36 to the cooling tower control unit 35. This allows the reinforcement learning unit 34 to derive an appropriate target operating capacity.

[0058] Furthermore, in the above explanation, the reinforcement learning unit 34 was assumed to input all state variables received from the state observation unit 31 into the target operating capacity model 36. However, the reinforcement learning unit 34 may be configured to input only some of the state variables into the target operating capacity model 36. For example, the information on the operating capacity of the currently operating cooling tower 20 may not be included in the input to the target operating capacity model 36.

[0059] In controlling the cooling tower 20, it is common to fix the cooling water temperature to a predetermined value, but there are also cases where the cooling water temperature is varied. In this case, the state variables acquired by the state observation unit 31 may also include information on the supply water temperature (cooling water temperature setpoint). In this case, the supply water temperature is also added to the input of the target operating capacity model 36. This allows the model to be trained while considering the cooling water temperature even when performing control that varies the cooling water temperature, thereby improving the training accuracy. For example, the cooling water inlet temperature T1 measured by the temperature sensor 22 can be used as the supply water temperature.

[0060] "Cooling water W1" and "Chilled water W2" may be replaced with a heat transfer medium other than water.

[0061] The cooling tower control unit 35 calculates the number of cooling towers 20 required to achieve the target operating capacity output by the target operating capacity model 36, and controls the operation of the cooling towers 20 based on the calculated number of operating towers.

[0062] Next, the flow of the target operating capacity estimation process by the cooling tower control device 30 will be described. Figure 4 is a flowchart showing the flow of the target operating capacity estimation process according to the embodiment.

[0063] In step S1, the state observation unit 31 acquires the operating conditions, operating capacity, and operating capacity of the currently operating cooling tower as state variables of the heat source system 1 based on the current target operating capacity, and outputs them to the reinforcement learning unit 34.

[0064] In step S2, the evaluation data acquisition unit 32 acquires the energy consumption of the cooling tower 20 and cooling water pump 40 currently in operation based on the current target operating capacity and outputs it to the reward calculation unit 33.

[0065] The reward calculation unit 33 calculates a reward related to the current operating status of the cooling tower 20 based on the energy consumption acquired in step S2. First, in step S3, the reward calculation unit 33 determines whether the energy consumption acquired in step S2 has decreased compared to the energy consumption acquired previously.

[0066] If energy consumption decreases (Yes in step S3), the process proceeds to step S4, and the reward calculation unit 33 increases the reward from its current value. On the other hand, if energy consumption increases (No in step S3), the process proceeds to step S5, and the reward calculation unit 33 decreases the reward from its current value.

[0067] In step S6, the reinforcement learning unit 34 performs machine learning on the target operating capacity model 36 based on the reward calculated in step S5 or S6. The reinforcement learning unit 34 updates the model parameters of the target operating capacity model 36, for example, so that learning progresses in the direction that maximizes the reward.

[0068] In step S7, the reinforcement learning unit 34 uses the target operating capacity model 36, which was learned in step S6, to infer the target operating capacity. The reinforcement learning unit 34 inputs the state variables acquired in step S1 into the target operating capacity model 36, executes the target operating capacity model 36 to output the target operating capacity, and performs inference of the target operating capacity. The reinforcement learning unit 34 outputs the inferred target operating capacity to the cooling tower control unit 35.

[0069] In step S8, the cooling tower control unit 35 determines the number of cooling towers 20 to be operated in order to achieve the inferred target operating capacity, and activates the determined number of cooling towers 20.

[0070] In step S9, the reinforcement learning unit 34 determines whether to continue learning the target operating capacity model 36. If learning is to be continued (Yes in step S9), the process returns to step S1. If learning is to be terminated (No in step S9), this control flow is terminated. As a criterion for determining the end of learning, for example, learning is terminated when the reward value calculated by the reward calculation unit 33 exceeds a predetermined threshold.

[0071] After the learning process is complete, as long as the current operating conditions and load of the cooling towers 20 are maintained, the number of cooling towers 20 will continue to be controlled based on the current target operating capacity output by the target operating capacity model 36.

[0072] Note that the process of acquiring state variables in step S1 may be performed before step S7. Also, the training of the target operating capacity model 36 in step S6 may be performed only when the reward increases.

[0073] As described above, the heat source system 1 according to the first embodiment includes a chiller 10, a cooling water pump 40 that supplies cooling water to cool the refrigerant by performing heat exchange within the chiller 10, and a cooling tower 20 that cools the cooling water led from the chiller 10 by the cooling water pump 40 by bringing it into contact with the outside air. Multiple cooling towers 20 are provided so that they have an operating capacity corresponding to the maximum load of the chiller 10. The cooling tower control device 30 of the heat source system 1 includes a reinforcement learning unit 34 that learns the relationship between the input of operating conditions and operating capacity of the cooling tower 20 and the output of the operating capacity, a state observation unit 31 that acquires state variables including at least operating conditions, operating capacity, and operating capacity, an evaluation data acquisition unit 32 that acquires the energy consumption of the cooling tower 20 and the cooling water pump 40 to evaluate the control results of the cooling tower 20, and a reward calculation unit 33 that calculates a reward based on the energy consumption. The reinforcement learning unit 34 uses the reward calculated by the reward calculation unit 33 to acquire the relationship between the input and output of the target operating capacity model 36 by reinforcement learning so that the energy consumption is minimized under the operating conditions and operating capacity conditions.

[0074] This configuration allows the number of cooling towers 20 to be started in the heat source system 1 to be determined in order to minimize energy consumption, depending on the operating conditions and operating load of the cooling towers 20. As a result, the heat source system 1 of the first embodiment can improve its operating efficiency.

[0075] Furthermore, as explained with reference to Figure 5, conventional cooling tower control methods such as those described in Patent Document 1 required the creation of a table that could grasp the COP of the entire heat source system based on the relationship between the outside air wet-bulb temperature and the chiller load factor, as shown in the relationship in Figure 5, determining the parameters to be used in the calculation formula that maximizes the COP of the entire heat source system from this table, and controlling the number of chillers in operation based on the calculation result. In contrast, the heat source system 1 of the first embodiment uses a target operating capacity model 36 learned by the reinforcement learning unit 34 to estimate the operating capacity (number of operating units) of the cooling tower 20 based on information such as the operating conditions of the cooling tower 20. This reduces computational costs and memory capacity, and makes it easy to determine the appropriate number of operating units for the cooling tower 20.

[0076] In the above embodiment, a configuration is shown in which a refrigerator 10 is connected to a plurality of cooling towers 20a, 20b, and 20c and supplied with cooling water cooled by the cooling towers 20a, 20b, and 20c. However, devices other than the refrigerator 10 included in the cooled section may also be used. The cooled section is a general term for devices that circulate cooling water to cool and control the temperature of a target sample, or a device or a part of a device. It includes configurations in which a refrigerant is cooled by the cooling water and the chilled water cooled by this refrigerant is supplied to an external load 50, and configurations in which the device itself is cooled by the cooling water without being connected to an external load 50. Examples of devices other than the refrigerator 10 include cooling devices for molds in a manufacturing line, condensers for steam turbines, and water-cooled servers.

[0077] In this way, when the chiller 10 is expanded to a water-cooled chiller, a configuration in which the heat source system 1 does not have an external load 50 is also conceivable. In the first embodiment, since the evaluation data is limited to the energy consumption of the cooling tower 20 and the cooling water pump 40, it can be applied to configurations without an external load 50, thereby improving versatility.

[0078] Furthermore, in the heat source system 1 according to the first embodiment, the reward calculation unit 33 calculates a higher reward the lower the energy consumption. As a result, the reinforcement learning unit 34 can obtain a higher reward as the energy consumption decreases, which allows it to quickly and appropriately learn the relationship between the input of the operating conditions and operating capacity of the cooling tower 20 and the output of the operating capacity.

[0079] [Second Embodiment] A second embodiment will be described with reference to Figure 6. Figure 6 is a diagram illustrating the functional configuration of the cooling tower control device 30A according to the second embodiment.

[0080] As shown in Figure 6, the state variables (input information to the model) acquired by the state observation unit 31 also include the load of the chiller 10. The load of the chiller 10 can be calculated, for example, from the product of the supply and return water temperature difference (temperature difference of chilled water at the inlet and outlet of the external load 50) and the chilled water flow rate (for example, calculated from the voltage and frequency of the inverter of the chilled water pump 47, or measured by a flow meter), or from the load factor of the chiller 10.

[0081] The evaluation data acquisition unit 32 acquires the energy consumption of the chiller 10 and chilled water pump 47 in addition to the energy consumption of the cooling tower 20 and cooling water pump 40. In this case, the reward calculation unit 33 calculates the reward based on the energy consumption of the cooling tower 20 and cooling water pump 40 and the energy consumption of the chiller 10 and chilled water pump 47. In other words, the reward is calculated based on the energy consumption of the entire heat source system 1. This allows the energy consumption of the entire heat source system 1 to be reflected in the reward calculation, enabling the reinforcement learning unit 34 to proceed with learning in a more appropriate direction and enabling the reinforcement learning unit 34 to estimate the operating capacity with higher accuracy.

[0082] The target operating capacity model 36 may include the cooling water temperature in its output. In this case, the output items of the state variables acquired by the state observation unit 31 also include information on the cooling water temperature. Alternatively, similar to the first embodiment, the cooling water temperature may be acquired as the operating target value of the state variables and input into the target operating capacity model 36.

[0083] In the second embodiment, it is preferable to perform the learning of the target operating capacity model 36 after the operating state of the current heat source system 1 has transitioned to steady-state operation where "load = operating capacity" after startup. This improves the accuracy of the load information included in the state variables acquired by the state observation unit 31, and thus improves the learning accuracy of the target operating capacity model 36.

[0084] The embodiments have been described above with reference to specific examples. However, this disclosure is not limited to these specific examples. Modifications made to these specific examples by those skilled in the art are also included within the scope of this disclosure, as long as they retain the features of this disclosure. The elements, their arrangement, conditions, shapes, etc., of each of the aforementioned specific examples are not limited to those illustrated and can be modified as appropriate. The elements of each of the aforementioned specific examples can be combined in different ways as appropriate, as long as no technical inconsistencies arise.

[0085] In the above embodiment, a configuration was illustrated in which the learning unit's target operating capacity model outputs a target operating capacity, and the cooling tower control unit determines the number of cooling towers to operate according to the target operating capacity. However, a configuration in which the learning unit's target operating capacity model directly outputs the number of operating cooling towers is also possible. [Explanation of Symbols]

[0086] 1. Heat source system 10 Refrigerator (cooled part) 20, 20a, 20b, 20c cooling tower 30, 30A Cooling Tower Control System 31 State Observation Unit 32 Evaluation Data Acquisition Unit 33 Compensation Calculation Department 34 Reinforcement Learning Department 40 Cooling water pump 47. Chilled water pump 50 External load

Claims

1. A heat source system comprising a section to be cooled, a cooling water pump that supplies cooling water to the section to be cooled, and a cooling tower that cools the cooling water introduced from the section to be cooled by the cooling water pump by bringing it into contact with the outside air, wherein multiple cooling towers are provided so as to have an operating capacity corresponding to the maximum load of the section to be cooled, A reinforcement learning unit that learns the relationship between the input of the operating conditions and operating capacity of the cooling tower and the output of the operating capacity, A state observation unit that acquires state variables including at least the operating conditions, the operating capacity, and the operating capacity, An evaluation data acquisition unit that acquires the energy consumption of the cooling tower and the cooling water pump to evaluate the control results of the cooling tower, The system includes a reward calculation unit that calculates a reward based on the energy consumed, The reinforcement learning unit uses the reward calculated by the reward calculation unit to acquire the relationship between the input and the output through reinforcement learning so that the energy consumption is minimized under the operating conditions and operating capacity conditions. The state variable and the output include the temperature of the cooling water. The operating capacity of the cooling tower included in the inputs that the reinforcement learning unit learns the input-output relationship of is the amount of heat that the cooling tower exchanges with the part being cooled. The heat source system is External load and The system includes a chilled water pump that supplies chilled water, cooled by heat exchange within the cooled section, to the external load. The state variables and the inputs include the load on the part being cooled. The evaluation data acquisition unit acquires, in addition to the energy consumption of the cooling tower and the cooling water pump, the energy consumption of the cooled part and the chilled water pump. The reward calculation unit calculates the reward based on the energy consumed by the cooling tower and the cooling water pump, and the energy consumed by the cooled part and the chilled water pump. Heat source system.

2. The part to be cooled is a refrigerator. The heat source system according to claim 1.

3. The reward calculation unit calculates a higher reward the lower the energy consumed. The heat source system according to claim 1 or 2.

4. In a heat source system comprising a section to be cooled, a cooling water pump that supplies cooling water to the section to be cooled, and a cooling tower that cools the cooling water introduced from the section to be cooled by the cooling water pump by bringing it into contact with the outside air, wherein multiple cooling towers are provided, each having an operating capacity corresponding to the maximum load of the section to be cooled, The cooling tower control system The steps include obtaining state variables, including at least the operating conditions, operating capacity, and operating capacity of the cooling tower, A step of obtaining the energy consumption of the cooling tower and the cooling water pump to evaluate the control result of the cooling tower, A step of calculating a reward based on the aforementioned energy consumption, Using the calculated reward, the relationship between the input of the operating conditions and operating capacity of the cooling tower and the output of the operating capacity is obtained by reinforcement learning such that the energy consumption is minimized under the operating conditions and operating capacity conditions. Execute, The state variable and the output include the temperature of the cooling water. Among the input-output relationships obtained by the reinforcement learning described above, the operating capacity of the cooling tower included in the input is the amount of heat that the cooling tower exchanges with the part being cooled. The aforementioned heat source system is External load and The system includes a chilled water pump that supplies chilled water, cooled by heat exchange within the cooled section, to the external load. The state variables and the inputs include the load on the part being cooled. The step of acquiring the aforementioned energy consumption includes acquiring the energy consumption of the cooling tower and the cooling water pump, as well as the energy consumption of the part being cooled and the chilled water pump. The step of calculating the reward involves calculating the reward based on the energy consumed by the cooling tower and the cooling water pump, and the energy consumed by the cooled part and the chilled water pump. Method for estimating target operating capacity.

5. In a heat source system comprising a section to be cooled, a cooling water pump that supplies cooling water to the section to be cooled, and a cooling tower that cools the cooling water introduced from the section to be cooled by the cooling water pump by bringing it into contact with the outside air, wherein multiple cooling towers are provided, each having an operating capacity corresponding to the maximum load of the section to be cooled, In the cooling tower control system, The steps include obtaining state variables, including at least the operating conditions, operating capacity, and operating capacity of the cooling tower, A step of obtaining the energy consumption of the cooling tower and the cooling water pump to evaluate the control result of the cooling tower, A step of calculating a reward based on the aforementioned energy consumption, Using the calculated reward, the relationship between the input of the operating conditions and operating capacity of the cooling tower and the output of the operating capacity is obtained by reinforcement learning such that the energy consumption is minimized under the operating conditions and operating capacity conditions. Make it run, The state variable and the output include the temperature of the cooling water. Among the input-output relationships obtained by the reinforcement learning described above, the operating capacity of the cooling tower included in the input is the amount of heat that the cooling tower exchanges with the part being cooled. The aforementioned heat source system is External load and The system includes a chilled water pump that supplies chilled water, cooled by heat exchange within the cooled section, to the external load. The state variables and the inputs include the load on the part being cooled. The step of acquiring the aforementioned energy consumption includes acquiring the energy consumption of the cooling tower and the cooling water pump, as well as the energy consumption of the part being cooled and the chilled water pump. The step of calculating the reward involves calculating the reward based on the energy consumed by the cooling tower and the cooling water pump, and the energy consumed by the cooled part and the chilled water pump. Target operating capacity estimation program.

Images