Neural network driven management optimization for system of chiller devices

Abstract

Systems and methods for optimized management of a system of multiple cooler devices for circulating a cooling medium within an indoor environment determine a current state of the environment, such as ambient temperature, temperature of the medium entering and exiting the environment, inlet flow rate of the cooled medium, target cooling load, and temperature of the medium entering and exiting the environment. And current configurations of the chiller system, such as the flow and outlet temperature setpoint per active device, the total energy consumption of the chiller system. Based on this information, the chiller system controller solves an optimal chiller system configuration offline or online and steps for implementing the configuration by adjusting the chiller device flow and outlet temperature setpoint for providing the required cooling load to the indoor environment while minimizing the total energy consumption of the entire chiller system. If an optimal solution is found, the controller monitors the implementation of the optimal solution.

Description

[0001] Cross - Reference to Related Applications

[0002] This application claims the benefit of priority to U.S. Provisional Patent Application No. 63 / 603,476, filed on November 28, 2023, the entire content of which is incorporated herein by reference. Technical Field

[0003] The present disclosure generally relates to the field of cooler systems for data centers and similar temperature - controlled environments, and more particularly, to control systems for the automatic management of such cooler systems. Background Art

[0004] A cooler device or cooler is a cooling delivery device that can circulate a cooling medium (e.g., water or some other working medium or fluid) through a data center, factory, or similar environment (e.g., a controlled indoor location or space to be cooled) to regulate, for example, the temperature, humidity, and / or air flow within the environment. Broadly speaking, the cooler cools the medium and supplies the cooled medium to the environment, where heat is transferred to the medium such that the medium leaves the environment and returns to the cooler at a higher temperature. The medium is cooled again and recirculated to the environment in a continuous cycle so that the environment can be maintained at a desired temperature. Thus, the goal of a cooler device (or a system of cooler devices) is to cool the hotter medium leaving the environment to a low enough temperature and supply it back to the environment at a sufficient flow rate to ensure that the desired temperature of the environment is maintained.

[0005] The amount of change or difference in temperature between the (cooler) medium entering the environment and the (hotter) medium leaving the environment defines the required cooling capacity of the cooler or cooler system (e.g., to maintain a desired or target temperature). For example, if the cooling demand increases (e.g., the ambient temperature at or near the environment increases relative to the target temperature), the cooler or cooler system can increase its cooling capacity, e.g., ramp the compressor or activate a previously inactive compressor in a closed multi - cooler system. Given such a closed multi - cooler system, the challenge is then to provide or maintain the required cooling capacity in the most efficient manner (e.g., at the lowest energy consumption rate). Summary of the Invention

[0006] In a first aspect, a cooler device system is disclosed that is configured to continuously maintain an environment or a factory within a target temperature range while dynamically minimizing total energy consumption. In an embodiment, the cooler device circulates a coolant fluid through a supply conduit to the factory in common and returns the coolant fluid to the cooler device via a return conduit. Each cooler device can be active or inactive (e.g., on or off) and has a range of possible outlet temperatures and flow rates (e.g., for the coolant fluid circulated to the factory). The supervisory control device of the cooler system monitors the environmental state of the environment or the factory, which includes the required cooling load based on the target temperature range. The control device evaluates the current configuration of the cooler system, such as the active or inactive state of each active cooler device and the current set points. In addition, the control device evaluates the total energy consumption based on the current configuration. The control device attempts to determine one or more optimal solutions or sequences of actions based on the current state, whereby the required cooling load can be maintained continuously while reducing the total energy consumption.

[0007] In some embodiments, the optimal solution includes adjusting the outlet temperature set point and / or the flow rate set point of at least one active cooler device.

[0008] In some embodiments, the optimal solution may further include enabling an inactive cooler device or disabling an active cooler device (e.g., changing the total number of active cooler devices).

[0009] In some embodiments, the control device changes the flow rate set point of an active cooler device by changing the current flow rate set point to a target flow rate set point associated with the optimal solution.

[0010] In some embodiments, the control device changes the outlet temperature set point of an active cooler device by changing the current outlet temperature set point to a target outlet temperature set point associated with the optimal solution.

[0011] In some embodiments, the control device models the possible configurations of the cooler system (e.g., the possible sets of active subsets of cooler devices and the outlet temperature / flow rate set points of each active cooler device). In addition, the cooler system includes a data storage device such as a memory in which the modeled possible cooler system configurations can be saved or stored.

[0012] In some embodiments, the cooler system determines the optimal solution by selecting a stored possible cooler system configuration from the memory (e.g., in the case where the possible configuration maintains the required cooling load associated with the current state).

[0013] In some embodiments, the control device models the possible cooler system configurations via a regression model such as a neural network.

[0014] In some embodiments, the control device performs actions for an optimized solution, where the optimized solution includes a set of actions or sequence of actions such as adjusting a setpoint, enabling a chiller device, disabling a chiller device. The control device collects data (e.g., setpoints, environmental conditions) from the overall chiller system to confirm steady-state operation of the chiller system for at least a threshold duration. Then, the control device (e.g., after establishing steady-state operation) evaluates the subsequent energy consumption level of the chiller system based on the actions performed. For example, if the subsequent energy consumption level is at least a threshold amount less than the current / previous energy consumption level, the control device proceeds to the next action in the sequence.

[0015] In some embodiments, the current state of the environment includes the ambient air temperature near the factory, the inlet temperature of the coolant fluid entering the factory via a supply conduit, the outlet temperature of the coolant fluid leaving the factory via a return conduit, and the flow rate of the coolant fluid entering the factory.

[0016] In another aspect, a method for dynamic optimization management of a chiller system is disclosed, by which the environment or the factory is maintained within a target temperature range. In an embodiment, the method includes: providing a chiller system for circulating a coolant fluid through one or more chiller devices of a factory, each chiller device being inactive or active, and each active chiller device having a series of possible outlet temperature setpoints and a series of possible flow rate setpoints. The method includes: determining the current state of the factory (including the required cooling load based on the target temperature range) via a supervisory control device of the chiller system. The method includes: determining the current configuration of the chiller system, such as the active or inactive state of each chiller device and the outlet temperature and flow rate setpoints. The method includes: determining the current energy consumption level of the chiller system based on the current configuration and state within the factory. The method includes: determining one or more optimized solutions for maintaining the required cooling load while reducing the total energy consumption, where each optimized solution includes a set of actions or sequence of actions.

[0017] In some embodiments, the optimized solution includes: adjusting the outlet temperature or flow rate setpoint of at least one active chiller device.

[0018] In some embodiments, the optimized solution includes: enabling an inactive chiller device or disabling an active chiller device, e.g., changing the number of active chiller devices.

[0019] In some embodiments, the method includes: adjusting the flow rate setpoint of an active chiller device by changing the current flow rate setpoint to a target flow rate setpoint (e.g., associated with the optimized solution).

[0020] In some embodiments, the method includes adjusting the outlet temperature setpoint of an active cooler device by changing a current outlet temperature setpoint to a target outlet temperature setpoint (e.g., associated with an optimized solution).

[0021] In some embodiments, the method includes mathematically modeling a set of possible configurations of a cooler system. Additionally, the method includes storing the modeled set of possible configurations in a memory of a supervisory control device or other data storage device.

[0022] In some embodiments, the method includes selecting a pre-modeled and stored configuration of the cooler system as an optimized solution (e.g., in a case where the modeled configuration achieves a desired cooling load while reducing total energy consumption).

[0023] In some embodiments, the method includes modeling possible cooler system configurations via a neural network or similarly suitable regression model.

[0024] In some embodiments, the method includes performing a first action of an optimized solution action sequence and verifying via data collection that a steady-state operation of the cooler system has occurred for at least a threshold duration. The method includes determining a subsequent energy consumption level of the cooler system based on the performed action. The method includes determining a change amount or difference between the subsequent energy consumption level and a previous energy consumption level. The method includes performing the next action of the optimized solution when the change amount meets or exceeds a threshold level.

[0025] In some embodiments, the current state of the environment includes the ambient air temperature near the plant, the inlet temperature of a coolant fluid entering the plant via a supply conduit, the outlet temperature of the coolant fluid leaving the plant via a return conduit, and the flow rate of the coolant fluid entering the plant.

[0026] This Summary is provided only as an introduction to the subject matter fully described in the Detailed Description and Drawings. The Summary should not be regarded as describing essential features nor used to determine the scope of the claims. Additionally, it should be understood that both the foregoing Summary and the following Detailed Description are merely exemplary and illustrative and are not necessarily restrictive of the claimed subject matter. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The detailed embodiments are described with reference to the drawings. The use of the same reference numerals in different instances in the specification and drawings may indicate similar or identical items. Various embodiments or examples of the present disclosure (“examples”) are disclosed in the following detailed description and drawings. The drawings are not necessarily to scale. Generally, unless otherwise provided in the claims, the operations of the disclosed processes may be performed in any order. In the drawings:

[0028] Figure 1Ais a block diagram of a multi-device cooler system (“coolant delivery system”) according to an example embodiment of the inventive concept disclosed herein;

[0029] Figure 1B is Figure 1A a block diagram of a single-device embodiment of the cooler system;

[0030] Figure 2 is a block diagram of a cooler system including a supervisory central control device; Figure 1A of the cooler system;

[0031] Figure 3 is Figure 2 a schematic diagram of an operation of mathematically modeling the cooler system;

[0032] Figure 4 is a schematic diagram of an operation of solution determination and evaluation of a reinforcement learning (RL, R / L) agent configured to optimize the cooler system; Figure 2 of the cooler system;

[0033] Figure 5A is Figure 2 a schematic diagram of data collection and modeling operations of the cooler system;

[0034] Figure 5B is Figure 2 a schematic diagram of initialization operations of the cooler system;

[0035] Figure 6A is Figure 2 a schematic diagram of online optimization operations of the cooler system;

[0036] Figure 6B is Figure 2 a schematic diagram of solution evaluation operations of the cooler system;

[0037] Figure 7 is Figure 2 a schematic diagram of offline optimization operations of the cooler system;

[0038] and Figure 8A and Figure 8B is a process flow block diagram showing a method for optimizing management of a cooler system according to an example embodiment of the inventive concept disclosed herein. Detailed Description

[0039] Before detailing one or more embodiments of the present disclosure, it should be understood that the embodiments are not limited in their application to the details of the construction and arrangement of components, steps, or methods set forth in the following description or illustrated in the drawings. In the detailed description of the following embodiments, numerous specific details may be set forth to provide a more thorough understanding of the present disclosure. However, it will be apparent to those of ordinary skill in the art who benefit from the present disclosure that the embodiments disclosed herein may be practiced without some of these specific details. In other instances, well-known features may not be described in detail to avoid unnecessarily complicating the present disclosure.

[0040] As used herein, the letters following a reference numeral are intended to refer to an embodiment of a feature or element that may be similar but not necessarily identical to a previously described element or feature having the same reference numeral (e.g., 1, 1a, 1b). Such shorthand notations are used only for convenience and should not be construed as limiting the present disclosure in any way unless expressly stated to the contrary.

[0041] Moreover, unless expressly stated to the contrary, "or" refers to an inclusive or rather than an exclusive or. For example, the condition A or B is satisfied by any of the following: A is true (or present) and B is false (or absent); A is false (or absent) and B is true (or present); and both A and B are true (or present).

[0042] In addition, the articles "a" or "an" may be used to describe elements and components of the embodiments disclosed herein. This is merely for convenience and, unless clearly otherwise indicated, "a" and "an" are intended to include "one" or "at least one," and the singular form also includes the plural.

[0043] Finally, as used herein, any reference to "one embodiment" or "some embodiments" means that a particular element, feature, structure, or characteristic described in connection with that embodiment is included in at least one embodiment disclosed herein. The phrase "in some embodiments" that appears in various places in this specification does not necessarily refer to the same embodiment, and the embodiments may include one or more of the features expressly described or inherently present herein, or any combination or sub-combination of two or more such features, as well as any other features that may not necessarily be expressly described or inherently present in the present disclosure.

[0044] Broadly speaking, embodiments of the inventive concept disclosed herein relate to methods and systems for optimizing the management of a closed system of multiple cooler devices to ensure that the cooler system delivers the required cooling capacity with optimal efficiency. For example, multi-cooler management optimization involves solving the problem of load division, as well as enabling and disabling the individual cooler devices of the system, to provide the required cooling capacity at minimum energy cost while the cooling load and ambient air conditions change.

[0045] For example, the cooling load of a cooler system is proportional to the flow rate of the medium entering the environment to be cooled and the temperature difference between the supply medium and the return medium. Thus, the required cooling load can be achieved by various different combinations of the flow rate and the supply medium temperature within the cooler system, where different combinations are associated with different energy consumption levels or rates.

[0046] Referring to Figure 1A and Figure 1B , a cooler system 100 (e.g., a coolant delivery system) of n cooler devices 102a, 102b, ……, 102n (e.g., coolant delivery devices) is shown. For example, n can be an integer not less than 2, but the inventive concept disclosed herein can equally be applied to a cooler system 100a including a single cooler device 102 (e.g., n = 1). In an embodiment, where applicable, any reference to the cooler system 100 of cooler devices 102a to cooler devices 102n can also be considered to refer to the cooler system 100a of cooler device 102. Regarding the cooler system 100a shown by Figure 1B , it can be noted that since the cooler system 100a includes only one cooler device 102, the possible cooler system configurations may be limited to the flow rate set point and the outlet temperature set point of the cooler device 102. Although each flow rate and outlet temperature set point can each be associated with a range of possible values, at any particular time, a single cooler device 102 can only have one flow rate and one outlet temperature set point. Similarly, enabling of an inactive cooler device and disabling of an active cooler device may not be applicable to the cooler system 100a since the single cooler device 102 remains active.

[0047] In an embodiment, with particular reference to the cooler system 100 shown by Figure 1A , the cooler devices 102a to 102n can be connected in parallel such that a single conduit 104 supplies coolant fluid (e.g., coolant medium, working fluid) to the environment or plant 106 to maintain it within a target temperature range or below a threshold temperature. For example, the cooler system 100 can have a target (desired) inlet temperature T inlet and a target flow rate Coolant fluid is supplied to the plant 106 via the conduit 104. For example, the target inlet temperature T can be determined based on the desired temperature range or temperature threshold within the plant 106 inlet or the target flow rate In an embodiment, the cooler system 100 (e.g., where n>1) can achieve the target inlet temperature T via a mixture of a hotter fluid and a colder fluid at a specific ratio inlet or the target flow rate Regarding the single cooler system 100a (e.g., where n = 1), a specific inlet temperature may not provide room for optimization, but the acceptable range of the target inlet temperature T inlet (e.g., 18°C to 22°C) can provide an opportunity to optimize energy consumption within the acceptable inlet temperature range.

[0048] In an embodiment, after the coolant fluid has absorbed heat within the plant 106, it can return to the cooler system 100 or the cooler system 100a via the return conduit 108. For example, the return fluid can leave the plant 106 at a specific return temperature and can be cycled back to the cooler system 100 or the cooler system 100a via the return conduit 108 to be re-cooled and dissipate the absorbed heat.

[0049] Referring to Figure 2 , the cooler system 200 can be implemented and function similar to Figure 1A the cooler system 100, except that the cooler system 100 can include a supervisory control device 202 that includes a processor 204 and a memory 206 (or other suitable data storage device).

[0050] In an embodiment, the supervisory control device 202 can maintain the supply temperature T by managing each individual cooler device 102a to cooler device 102n inlet and the flow rate For example, each of the n cooler devices 102a to cooler device 102n can have an active / inactive setting (e.g., determining whether the compressor 208 of the cooler device is active or inactive; each cooler device having multiple compressor units can be based on the enabling, changing, deactivating, etc. of a single or group of local control component compressor units) and adjust the outlet temperature set point temperature T that regulates the temperature and flow rate of the coolant fluid from that specific cooler device (i.e., the cooling capacity of the cooler device) setpoint and the flow rate Thus, for any desired cooling load to be achieved by the cooler system 200 of cooler devices 102a through 102n, that load can be provided by a higher number of active cooler devices (e.g., where each active cooler device operates at a lower cooling capacity) or a lower number of active cooler devices (e.g., where each active cooler device operates at a higher cooling capacity). While any number of possible configurations of cooler devices 102a through 102n can achieve the desired cooling load, the question of which of these configurations or combinations is the most energy efficient, i.e., consumes the least amount of energy overall, is not a simple one.

[0051] In view of the foregoing, embodiments of the inventive concepts disclosed herein relate to a cooler system architecture and a control system for the cooler system architecture that is configured to determine, for a given desired cooling load, an optimal cooler system configuration for continuously achieving and maintaining the desired cooling load while minimizing energy consumption. For example, given a system 200 of n cooler devices 102a through 102n, any possible cooler system configuration (including the current cooler system configuration and the optimal cooler system configuration) will define which cooler devices are active and which are inactive, and a specific outlet temperature set point T setpoint and flow rate set point Moreover, implementing the determined optimal cooler system configuration that is different from the current optimal cooler system configuration may involve a series of actions or sequence of actions, i.e., an optimal solution, where the operation of the cooler system 200 is transitioned from the current configuration to the optimal configuration while maintaining an uninterrupted delivery of the desired cooling load. Further still, the supervisory control device 202 can evaluate whether a particular action recommended by the optimal solution should be implemented.

[0052] In an embodiment, now referring to Figure 3 , to determine whether a proposed modification (i.e., an optimal solution) to the cooler system control strategy can reduce energy consumption while maintaining an uninterrupted cooling capacity, the cooler system 200 (see Figure 2 ) can be mathematically modeled (300; via, e.g., white box, black box, gray box modeling). For example, black box modeling (e.g., based only on data and not physical rules) can incorporate a regression-based neural network 302 or other similar regression methods. Based on being trained to predict for a given cooler system 200 individual cooler devices 102a through 102n (see Figure 2a neural network model 302 of the power consumption, and neural network modeling of various configurations of a given cooler system under a variety of possible operating conditions (e.g., different inlet temperatures, cooling capacities, flow rates among component coolers), the cooler system model 300 based on the neural network can be trained to take environmental conditions representing possible states of the environment to be cooled by the cooler system as inputs:

[0053] · the ambient air temperature 304 of the environment (e.g., °C);

[0054] · the inlet fluid temperature 306 of the refrigerated coolant fluid entering the factory (106, FIG. 1);

[0055] · the outlet fluid temperature 308 of the coolant fluid returning from the factory 106 to the cooler system 100;

[0056] · the flow rate 310 of the refrigerated coolant fluid entering the factory 106 (e.g., m 3 / hr); and

[0057] · the required or demanded cooling load 312 (e.g., kW).

[0058] In an embodiment, given the set of environmental conditions 304 to 312 as inputs, the cooler system model 300 can predict the total power consumption 314 (e.g., in kW) of a given cooler system configuration (e.g., for achieving and / or maintaining the required cooling load 312 from a given ambient air temperature 304, inlet fluid temperature 306, outlet fluid temperature 308, and flow rate 310). Additionally, based on the precise depth and / or architecture of the cooler system model 300, the total power consumption 314 can be predicted with a desired error bound.

[0059] In some embodiments, and as described in more detail below, computationally expensive modeling can be performed offline, thereby simplifying subsequent optimization operations online (trading off loss of operational flexibility such as the ability to finely adjust the cooler system for reduced computational requirements). For example, in the offline phase, a set of possible cooler system configurations can be modeled and stored in the memory 206 of the supervisory control device 202 (see, e.g., Figure 2 ). Thus, the processor 204 of the supervisory control device 202 can solve the optimization problem in the online phase by selecting the best cooler system configuration from the stored set of pre-modeled cooler system configurations.

[0060] In an embodiment, for a cooler system 200 having n cooler devices 102a to cooler device 102n (each cooler device having known properties and settings (active / inactive, T setpoint , )) The current state of the environment (e.g., Figure 2 factory 106 of

[0061] ; environmental conditions 304 - 308)) including the required cooling load 312, for determining the optimal cooler system configuration for maintaining the required cooling load while minimizing the total energy consumption 314, the number and nature of the n cooler devices 102a to 102n can be considered, as well as the need to maintain the required cooling load without interruption. Thus, any cooler system configuration that disrupts the required cooling load 312 (e.g., as a result of performing one or more actions to achieve the configuration) can be discarded. Additionally, any feasible cooler system configuration solution may need to consider other physical limitations or rules. For example, an optimization method may require that the number m of active cooler devices 102a to 102m (e.g., where m ≤ the total number n of cooler devices 102a to 102n) remains constant, thus prohibiting the enabling of inactive cooler devices or the disabling of active cooler devices, while another optimization method may make enabling or disabling individual cooler devices an actionable item.

[0062] In an embodiment, for a cooler system 200 with cooler devices 102a to 102n (each cooler device associated with specific physical properties and settings (e.g., a possible range of outlet fluid temperature setpoints T setpoint and a possible range of outlet flow rate setpoints associated with and corresponding to the mathematical cooler model 300), the supervisory control device 202 may attempt to establish and / or determine one or more possible optimization solutions. For example, possible optimization solutions may include any configuration of the cooler system 200, e.g., any set of possible (T setpoint , ) of each cooler device 102a through cooler device 102n, as well as the current set of environmental conditions 304 through 312 within the factory 106, where the required cooling load 312 is maintained continuously and the total energy consumption 314 is less than the current energy consumption based on the current configuration of the cooler system. It can be noted that for a given cooler system 200 in its current configuration and a factory 106 in a given environmental state, the number of possible optimization solutions can be zero, one, or more than one; in the latter case, the optimal solution can be the one that minimizes the energy consumption 314 to the greatest extent. In some embodiments, an optimization solution may be acceptable if the reduction in energy consumption 314 is within an acceptable range of a change threshold (e.g., 1%).

[0063] In an embodiment, optimization solutions can be defined in two basic ways. For example, a first type of optimization solution may provide that a number Z (e.g., where Z ≤ N) of active cooler devices 102a through cooler device 102n must remain unchanged, where no cooler device is enabled or disabled and any modification in the cooler system configuration is based only on the adjustment of the outlet fluid temperature setpoint T setpoint and / or the outlet flow rate setpoint of one or more cooler devices. If a greater reduction in the total energy consumption 314 can be achieved, a second type of optimization solution may additionally allow enabling or disabling cooler devices 102a through cooler device 102n.

[0064] In an embodiment, assuming a mass mathematical model 300 for each cooler device 102a through cooler device 102n of the cooler system 100 (e.g., a model of the cooler system 200 as a whole, or a set of models for each component cooler device 102a through cooler device 102n), each model may take the flow rate Q s (310, m 3 / h), the supply fluid temperature T s (306, °C), the ambient temperature T ambient (302, °C), the outlet fluid temperature T outlet (308, °C), and the required cooling capacity Req load (312, kW) as inputs. As a non-limiting example, the achievement of the first type of optimization solution can be expressed according to the following formula:

[0065]

[0066]

[0067]

[0068] where the result of the optimized solution includes the outlet fluid temperature setpoints T of one or more of the cooler devices 102a to 102n of the cooler system 200 setpoint and / or the outlet flow rate setpoint

[0069]

[0070] Similarly, as a non-limiting example, the implementation of the second type of optimized solution can be represented by the following formula - where the cooler devices 102a to 102n of the cooler system 200 are allowed to be enabled or disabled:

[0071]

[0072]

[0073] where, for example, A i refers to the active / inactive state (e.g., 0 = inactive, 1 = active) of a given cooler device 1 ≤ i ≤ n of the cooler system 200 having n cooler devices. Thus, this second type of optimized solution can additionally output a list of the active cooler devices i selected from the entire set of n cooler devices 102a to 102n, where each active cooler device i represents a cooler device that is already active (e.g., at a specific flow rate setpoint and outlet temperature setpoint) or must be enabled.

[0074] In an embodiment, with reference to Figure 4 , the simulated cooler system 400 can be implemented and operate similar to Figure 1A , Figure 1B and Figure 2 's cooler systems 100, 100a, 200, except that the simulated cooler system 400 can incorporate the modeling of possible cooler system configurations, determining the total energy consumption level 314 associated with each cooler system configuration as discussed above (e.g., offline modeling and storage of possible cooler system configurations), and the formulation of the optimized solution. For example, the optimized solution can include, for example, based on the active / inactive, flow rate setpoint, and outlet temperature setpoint of each cooler device 102a to 102n (see Figure 2 ), the current configured cooler system 200 (see Figure 2)The sequence of actions to convert to the optimal chiller system configuration, e.g., where the total energy consumption 314 of the chiller system is minimized via a trained reinforcement learning (RL) agent 402.

[0075] In an embodiment, the RL agent 402 can map the current state 404 of the factory 106 (e.g., ambient air temperature 304, inlet fluid temperature 306, outlet fluid temperature 308, incoming fluid flow rate 310, required cooling load 312, as shown by Figure 3 ) to possible actions applicable to the current state to minimize one or more cost functions (e.g., power consumption level 314). For example, the component actions for the optimal solution and any other candidate solutions can include:

[0076] · Adjust (e.g., raise or lower) the outlet fluid temperature set point T of the active chiller devices 102a to 102n setpoint ;

[0077] · Adjust the flow rate set point of the active chiller devices

[0078] · Enable currently inactive chiller devices (if the control scheme allows); or

[0079] · Disable currently active chiller devices (if the control scheme allows).

[0080] In an embodiment, the RL agent 402 can be trained on a simulated factory 406 environment corresponding to the factory 106 of the chiller system 200 (see Figure 2 ). For example, given any possible current state 404 of the factory 106 (including the corresponding current configuration of the component chiller devices 102a to 102n of the chiller system 200), the RL agent 402 can be trained to learn the best possible actions in that state to minimize the total energy consumption 314 while maintaining the required cooling load 312. In an embodiment, the development of the RL agent 402 can start with data acquisition (e.g., chiller modeling via rating software during the steady-state operation of the chiller system 200), followed by training a neural network 302 (see Figure 3 ) within each chiller model 300a to 300n corresponding to the chiller devices 102a to 102n of the chiller system (see Figure 3 ). The development can continue with training the RL agent 402 and ultimately result in a trained RL agent. One advantage of this approach is that the optimal solution can be determined independent of the actual environmental conditions of the chiller system 200 and / or the factory 106, because the simulated chiller system 400 can reproduce the full range of possible environmental conditions 304 to 312 (seeFigure 3 )。

[0081] In an embodiment, the neural network 302 representing the trained RL agent 402 can be implemented via the supervisory control device 202 (see Figure 2 ) to provide real-time direct control of the chiller system 200 and its components, the chiller devices 102a through 102n. For example, the uploaded neural network 302 can map real-time operating parameters (e.g., conditions within the plant 106, the configuration of each chiller device 102a through 102n) to determine an optimal chiller system configuration and implement any actions needed to achieve the optimal refrigerant system configuration from the current chiller system configuration. Conditions within the plant 106 can include the current state 404 of the plant (e.g., at time t) and a subsequent state 408 of the plant (e.g., at time t+1), the subsequent state indicating a change in environmental conditions within the plant due to the implemented actions.

[0082] Generally referring to Figures 5A to 7 , an example architecture for optimization management can provide: offline data collection and modeling of possible configurations of the chiller system 200 (e.g., as shown in Figure 5A ), online initialization and operation of the chiller system 200 (e.g., as shown in Figure 5B ), a supervised evaluation of feasible solutions based on a given current state 404 of the plant 106 (see Figure 6A ) (see Figure 4 ), by changing the actions associated with a possible solution or the implementation of changes to specific chiller devices 102a through 102n (see Figure 6B ), and / or an offline precomputation and preevaluation of possible solutions including the optimal solution (e.g., as shown in Figure 7 ).

[0083] Specifically referring to Figure 5A , the chiller system 200 is shown.

[0084] In an embodiment, the offline data collection 502 can allow the supervisory control device 202 (see Figure 2 ) to maintain the chiller system 200 in a steady-state operation 504 for at least a threshold duration, such that information regarding chiller device set points (e.g., inlet temperature, flow rate) can be observed, for example, via rating software. For example, the steady-state data can allow for the modeling 300 of the chiller system 200, as shown above by Figure 3 and Figure 4 , such that the chiller models 300a through 300n correspond to each chiller device 102a through 102n.

[0085] Also referring to Figure 5B, showing the cooler system 200.

[0086] In an embodiment, the plant 106 (see Figure 2 ) and the respective cooler devices 102a to 102n of the cooler system 200 (see Figure 2 ) can be initialized 506 based on any necessary properties such as cooler data 508 (e.g., the number and type of cooler devices 102a to 102n), the operating range 510 of each cooler device, and / or operating parameters 512 (e.g., the frequency at which the allowable setpoint change is permitted, whether cooler device enabling and disabling is permitted). For example, once the cooler system 200 is initialized 506 and operating 514 (as Figure 6A shown in more detail), the collection 502 of operating data can continue (e.g., at a predetermined frequency, such as every 10 seconds or 15 seconds), and the collection of the operating data can include, for example, the flow rate and outlet temperature setpoint of each cooler device 102a to 102n, the ambient conditions 304 to 312 within the plant 106 (see Figure 3 ) and / or the measurement or collection of other inputs for the optimization solution. In an embodiment, the collected data set can be stored and / or saved for evaluation and / or retraining of the RL agent 402 (see Figure 4 ).

[0087] In an embodiment, returning to reference Figure 5A , the collection 502 of operating data can ensure that the cooler system 200 operates in a steady state 504 for a sufficient duration (e.g., 15 minutes) such that optimization can occur. For example, the steady state operation 504 can be defined as a duration in which the cooler devices 102a to 102n are not enabled or disabled, and in which the total inlet temperature T inlet and the flow rate entering the plant 106 (see FIG. 1; e.g., via the input conduit 104) fluctuate by no more than a predetermined threshold.

[0088] Now referring to Figure 6A , an optimization process 600 for the initialized and operating cooler system 200 (see Figure 2 ) is shown.

[0089] In an embodiment, once the data collection 502 verifies that the steady state operation 504 of the cooler system 200 has been established for at least a threshold duration (e.g., 15 minutes), the supervisory control device 202 (see Figure 2) It is possible to evaluate the total cooling capacity (602) delivered by the chiller system 200 in its current configuration. For example, based on the desired type of the selected optimization solution (604; for example, a less conservative method (second type) that allows the chiller device to be enabled / disabled, or a more conservative method (first type) that prohibits enabling / disabling), the supervisory control device 202 may attempt to solve (606) one or more optimization solutions. In some embodiments, to prevent unwanted fluctuations within the chiller system 200 (e.g., due to frequent setpoint changes), the timing or frequency of perturbing the chiller system to solve (606) and / or implement the optimization solution may be restricted (e.g., not more frequently than every 30 minutes).

[0090] In an embodiment, based on the data collection 502 and the determined current cooling capacity (602) of the chiller system 200 in its current configuration, the supervisory control device 202 may determine (608) that there may be one or more optimization solutions or there may be no solution. For example, if no optimization solution is found, the data collection 502 may continue, and the steady state operation 504 may be maintained until a subsequent attempt to solve 606 (e.g., after confirming a sufficient duration of the steady state operation via data collection and after re-determining the total cooling capacity 602). In an embodiment, if a single optimization solution is found, the supervisory control device 202 may determine (610) the total energy consumption 314 associated with the optimization solution (e.g., with the chiller system configuration proposed by the optimization solution) (see Figure 3 ) whether it is sufficiently lower than the energy consumption level associated with the current configuration of the chiller system 200. For example, if the change amount between the proposed energy consumption and the current energy consumption or the amount of energy saved by implementing the proposed optimization solution meets or exceeds a threshold level, the proposed solution (or, for example, the next action or sequence of actions of the proposed solution) may be evaluated by the supervisory control device 202 for implementation (612; as Figure 6B shown).

[0091] In an embodiment, if multiple optimal solutions are found, the supervisory control device 202 may test each solution, rank or sort the optimal solutions in order of decreasing energy consumption (614), and select the optimal solution of the first order (616) as the optimal solution, for example, for the comparison of energy consumption 610. If, for example, the optimal solution of the first order sufficiently reduces energy consumption (e.g., at least reduced to a threshold level), the solution of the first order may be evaluated for implementation 612. Additionally, for example, if the solution of the first order is not implemented, the supervisory control device 202 may evaluate any remaining optimal solutions in order. For example, if enabling or disabling the cooler devices 102a to 102n is allowed to implement the optimal solution of the first order, the required energy savings threshold may be higher than the case where enabling or disabling is not allowed. However, if the optimization type that explicitly prohibits enabling or disabling the cooler devices 102a to 102n is selected (604), the solution of the first order may be discarded in favor of an optimal solution that achieves at least a threshold reduction in energy consumption without enabling or disabling the cooler devices.

[0092] In an embodiment, the supervisory control device 202 may save (618) any collected data set to, for example, the memory 206 (see Figure 2 ), for the evaluation and / or further refinement of the cooler system model 300 (see Figure 3 ) and / or the RL agent 402 (see Figure 4 ).

[0093] Also referring to Figure 6B , an evaluation process 612 of the proposed optimal solution by the supervisory control device 202 (see Figure 2 ) is shown.

[0094] In an embodiment, the evaluation 612 of a proposed optimal solution incorporating an adjustment to the inlet temperature or flow rate setpoint of one or more components of the cooler system 100, the cooler devices 102a to 102n, may include an evaluation 618 of each action or step of the proposed optimal solution. For example, if the proposed optimal solution only includes a change to the inlet temperature and / or flow rate setpoint (e.g., as opposed to enabling or disabling the cooler devices 102a to 102n), implementation may be accomplished by gradually changing 620 each current device setpoint (e.g., inlet temperature and / or flow rate) to the corresponding solution setpoint to reduce the pressure on the cooler system 200 and its component cooler devices. Additionally, referring back to Figure 6A , once the solution setpoint has been achieved and once steady state operation 504 for a sufficient duration has been confirmed via data collection 502, the overall impact of the proposed solution on energy consumption may be evaluated (610) relative to a suitable change amount threshold, and the total cooling capacity under the solution conditions may be evaluated (602).

[0095] In some embodiments, when the proposed optimized solution includes enabling and / or disabling the cooler devices 102a through 102n, the threshold change in the reduced energy consumption (e.g., where the supervisory control device 202 compares the energy consumption under the proposed solution to the current energy consumption (624)) can be higher than the threshold change where only a setpoint change is required. For example, the higher threshold change can account for the increased pressure of the cooler system 200 and its components, the cooler devices 102a through 102n, associated with the enabling / disabling of the cooler devices, and thus may require a more significant reduction in energy consumption to adjust for the additional pressure.

[0096] In an embodiment, when the reduction in energy consumption associated with a second type of optimized solution meets or exceeds the higher threshold change (e.g., in kW), the enabling or disabling of the cooler devices 102a through 102n can be performed first. For example, the supervisory control device 202 can first evaluate (606) the current number of active cooler devices i within a set of n cooler devices 102a through 102n. If the number of current active cooler devices 102a through 102n is equal to the target number Z of active cooler devices proposed by the optimized solution, the supervisory control device 202 can then continue to implement any setpoint changes (628) proposed by the optimized solution, e.g., by changing the flow rate and outlet temperature setpoints (620) to the proposed targets.

[0097] In an embodiment, if the current number of active cooler devices i is less than or greater than the target number Z of active cooler devices proposed by the optimized solution, the supervisory control device 202 can first enable and / or disable cooler devices (630) until the target number Z of active cooler devices is reached.

[0098] In some embodiments, the calculation and evaluation of the optimized solution can be performed offline rather than online (e.g., as shown above in Figure 6A and Figure 6B ). For example, referring to Figure 7 , an optimization process 700 implemented by an RL agent 402 (see Figure 2 ) of the optimized solution evaluated and stored offline by the supervisory control device 202 (see Figure 4 ) is shown.

[0099] In an embodiment, the optimization process 700 can provide an offline evaluation in advance of all possible optimized solutions (e.g., as shown in Figure 6B ) and store them in, for example, the memory 206 (see Figure 2 ). For example, the optimization process 700 can first assume a factory 106 (see Figure 4) the current state 404 within (see Figure 4 ; including environmental conditions 304 to 312), and data collection 502 can be performed to verify (504) the steady-state operation of the chiller system 200 (see Figure 2 ) for at least a threshold duration. In an embodiment, the collected data set can be saved (618) for model refinement.

[0100] In an embodiment, the RL agent 402 does not online solve (606, Figure 6A ) one or more possible optimization solutions, but instead can alternatively look up or retrieve (702) from the memory 206 the optimization solution corresponding to or most closely matching the current state 404 within the factory 106. For example, once the corresponding optimization solution is retrieved (702) by the RL agent 402 and implemented (704) by the supervisory control device 202, a safety check of the optimization solution can be introduced (706) to verify that the real-world effects and consequences of the proposed optimization solution match its previously offline-computed effects. For example, the retrieved RL optimization solution can be observed first for a minimum duration (e.g., 30 minutes) to ensure that the result determined and retrieved offline by the RL agent 402 corresponds to the actual result of the proposed action when implemented by the chiller system 200 (e.g., a sufficiently reduced energy consumption relative to the current configuration). If this check fails, for example, the supervisory control device 202 can return to data collection 502 to verify 504 the steady-state operation of the chiller system 200. In an embodiment, once the steady-state operation is continued and verified, the RL agent 402 can be directed to retrieve (702) a subsequent optimization solution. Additionally, an invalid optimization solution may lead to further refinement of the chiller system model 300 (see Figure 3 ). In an embodiment, when the retrieved RL optimization solution is checked 706 and verified to be executed in accordance with the previously determined offline result, the effective retrieved RL optimization solution can be evaluated 610, for example, as described above regarding Figure 6A 's online optimization process 600, to determine whether the optimization solution sufficiently reduces energy consumption to allow implementation 612 of the optimization solution, as Figure 6B illustrated.

[0101] Now referring to Figure 8A , a method 800 for optimizing the management of a chiller system to minimize total energy consumption while maintaining the desired cooling capacity of a factory or other target environment can include the following steps.

[0102] Step 802 includes: providing a chiller system of n chiller devices, where each chiller device has an active or inactive state, a coolant fluid flow rate set point, and an inlet medium temperature set point.

[0103] Step 804 includes: determining a current state within a plant or target environment (e.g., via a supervisory control device of a chiller system). For example, the current state includes the ambient air temperature near or within the plant, the inlet temperature of the chilled coolant fluid entering the plant, and the outlet temperature of the warmer coolant fluid returning from the plant to the chiller system, the flow rate of the coolant fluid entering the plant, and a target cooling load (e.g., desired cooling capacity) to be maintained within the plant.

[0104] Step 806 includes: determining a current configuration of the chiller system based on the current state of the plant. For example, the current configuration may provide a plurality of active chiller devices and current flow rate set points and temperature set points for each active chiller device. Additionally, based on the number of active chiller devices and their corresponding set points, step 806 includes determining a current total energy consumption level of the chiller system.

[0105] Step 808 includes: determining an optimized solution by which the target cooling load of the plant can be maintained without interruption, but the energy consumption level can be minimized. For example, the optimized solution may involve adjusting the inlet temperature and / or flow rate set points of one or more active chiller devices. In some embodiments, the optimized solution may also include enabling an inactive chiller device or disabling an active chiller device.

[0106] Still referring Figure 8B , method 800 may include additional steps 810 to 818.

[0107] Step 810 includes: performing a first action of the determined optimized solution. For example, if permitted by the supervisory control device, the first action may be enabling or disabling a chiller device. Alternatively or additionally, the first action may be adjusting the flow rate set point or outlet temperature set point of an active chiller device of the chiller system.

[0108] Step 812 includes: confirming, for example via collection of data from the chiller device and / or from within the plant, a steady state operation of the chiller system for at least a threshold duration after performing the first action.

[0109] Step 814 includes: determining an energy consumption associated with the chiller system (e.g., whether and to what extent the execution of the first action has reduced the total energy consumption) based on subsequent steady state operation of the chiller system for at least the threshold duration.

[0110] Step 816 includes: determining a change amount or difference between the subsequent energy consumption of the chiller system in its current configuration and the total energy consumption (e.g., before the execution / realization of the optimized solution's first action).

[0111] Step 818 includes: when the change amount or difference meets or exceeds a threshold level, performing subsequent actions to achieve a solution. For example, when an optimization solution of the second type provides a cooler device, the threshold for reducing energy consumption may be higher than that of the first type of optimization solution that prohibits enabling or disabling. The higher threshold of the second type of optimization solution requires adjusting the additional pressure placed on the cooler device due to enabling and / or disabling.

[0112] The devices and methods described in this application can be partially or fully implemented by a special-purpose computer created by configuring a general-purpose computer to execute one or more specific functions embodied in a computer program. The above functional blocks, flowchart components, and other elements serve as software specifications, which can be compiled into a computer program through the routine work of an experienced technician or programmer.

[0113] A computer program includes processor-executable instructions stored on at least one non-transitory, tangible computer-readable medium. The computer program may also include or rely on the stored data. The computer program may include a basic input / output system (BIOS) that interacts with the hardware of the special-purpose computer, device drivers that interact with specific devices of the special-purpose computer, one or more operating systems, user applications, background services, background applications, etc.

[0114] A computer program may include: (i) descriptive text to be parsed, such as HTML (HyperText Markup Language), XML (eXtensible Markup Language), or JSON (JavaScript Object Notation), (ii) assembly code, (iii) object code generated from source code by a compiler, (iv) source code executed by an interpreter, (v) source code compiled and executed by a just-in-time compiler, etc. By way of example only, the source code may be written using the syntax of languages including C, C++, C#, Objective-C, Swift, Haskell, Go, SQL, R, Lisp, Fortran, Peri, Pascal, Curl, OCaml, HTML5 (HyperText Markup Language 5th Revision), Ada, ASP (Active Server Pages), PHP (PHP: Hypertext Preprocessor), Scala, Eiffel, Smalltalk, Erlang, Ruby, Visual Lua, MATLAB, SIMULINK, and the syntax of the language.

[0115] The process flows discussed herein illustrate possible operations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart may represent a module, segment, or portion of code, which includes one or more executable instructions for implementing a particular logical function. It should also be noted that, in some alternative implementations, the functions noted in the blocks may occur out of the order depicted in the figures. For example, the blocks shown in succession may be executed substantially concurrently. It will also be noted that each block of the flowchart illustration may be implemented by a system based on dedicated hardware for performing a particular function or action, or by a combination of dedicated hardware and computer instructions.

[0116] For purposes of illustration, the foregoing description uses specific nomenclature to provide a thorough understanding of the described embodiments. However, it will be apparent to those skilled in the art that specific details are not required in order to practice the described embodiments. Thus, for purposes of illustration and description, the foregoing description of the specific embodiments described herein is presented. They are not intended to be exhaustive or to limit the embodiments to the precise form disclosed. It will be apparent to those of ordinary skill in the art that, given the above teachings, many modifications and variations are possible.

Claims

1. An optimized cooler system, comprising: one or more cooler devices fluidly coupled to the environment, the one or more cooler devices collectively configured to circulate a coolant fluid through the environment; as well as a controller operatively coupled to each of the one or more chiller devices, the controller comprising at least one processor and configured to: determining a current state of the environment including a target cooling load; Determining a chiller system configuration corresponding to a current state of the environment, the chiller system configuration corresponding to: the active or inactive status, flow rate set point, and outlet temperature set point of each chiller unit of the chiller system; as well as The energy consumption level of the chiller system; and At least one optimization solution configured to minimize an energy consumption level of the chiller system is determined based on the target cooling load associated with a current state of the environment, each optimization solution comprising at least one action.

2. The chiller system according to claim 1, wherein: The at least one action includes at least one of the following: changing at least one flow rate set point of an active chiller device of the chiller system, or At least one outlet temperature set point of an active chiller device of the chiller system is changed.

3. The chiller system of claim 2, wherein: The at least one action also includes at least one of the following: activating at least one inactive chiller device of the chiller system, or At least one active chiller unit of the chiller system is deactivated.

4. The chiller system of claim 2, wherein: The controller is configured to change the flow rate set point of the active cooler device by changing a current flow rate set point of the active cooler device to a target flow rate set point.

5. The chiller system of claim 2, wherein: The controller is configured to change an outlet temperature set point of the active chiller device by changing a current outlet temperature set point to a target outlet temperature set point.

6. The chiller system of claim 1, wherein: The controller is configured to model a plurality of possible chiller system configurations of the chiller system, and Therein, each possible chiller system configuration corresponds to an active or inactive state, a flow rate set point, and an outlet temperature set point for each chiller device of the chiller system.

7. The chiller system of claim 6, further comprising: a memory coupled to the at least one processor, the memory configured to store the plurality of possible chiller system configurations; Wherein the controller is configured to determine the at least one optimization solution by selecting a target chiller system configuration from a plurality of stored possible chiller system configurations based on the target cooling load associated with a current state of the environment.

8. The chiller system of claim 6, wherein: The controller is configured to model the plurality of possible chiller system configurations via at least one regression-based neural network.

9. The chiller system of claim 1, wherein: The controller is also configured to: executing at least one first action of the optimization solution, confirming steady state operation of the chiller system for at least a threshold duration after at least one first action was performed; determining, via the controller, a subsequent energy consumption level of the chiller system, the subsequent energy consumption level being based on the steady state operation of the chiller system for at least the threshold duration, determining a difference between the subsequent energy consumption level and the current energy consumption level, and When the determined difference exceeds a threshold level, at least one second action of the optimization solution is performed.

10. The chiller system of claim 1, wherein: The current state of the environment includes: Ambient air temperature; an inlet temperature of the coolant fluid entering the environment; an inlet flow rate of coolant into the environment; and The outlet temperature of the coolant fluid leaving the environment.

11. A method for optimized management of a chiller system, the method comprising: Providing a chiller system comprising one or more chiller devices, wherein the chiller system is configured to circulate a coolant fluid through an environment, and wherein each chiller device of the chiller system is associated with an active or inactive state, a flow rate set point, and an outlet temperature set point; determining a current state of the environment including a target cooling load; determining, via a controller of the chiller system, a chiller system configuration corresponding to a current state of the environment, the chiller system configuration corresponding to: the active or inactive status, flow rate set point, and outlet temperature set point of each chiller unit of the chiller system; and the energy consumption level of the chiller system; being associated with an energy consumption level of the chiller system; and At least one optimization solution configured to minimize an energy consumption level of the chiller system is determined via the controller based on the target cooling load, each optimization solution including at least one action.

12. The method according to claim 11, wherein: The at least one action includes at least one of the following: changing a flow rate set point of at least one active chiller device of the chiller system; or An outlet temperature set point of at least one active chiller device of the chiller system is changed.

13. The method according to claim 12, wherein: The at least one action also includes at least one of the following: activating at least one inactive chiller device of the chiller system; or At least one active chiller unit of the chiller system is deactivated.

14. The method according to claim 12, wherein: Changing the flow rate set point of an active chiller device of the chiller system includes changing a current flow rate set point of the active chiller device to a target flow rate set point.

15. The method according to claim 12, wherein: Changing an outlet temperature set point of an active chiller device of the chiller system includes changing a current outlet temperature set point to a target outlet temperature set point.

16. The method according to claim 11, wherein: Determining, via the controller, a current state of the environment includes: modeling a plurality of possible chiller system configurations of the chiller system, wherein each possible chiller system configuration corresponds to an active or inactive state, a flow rate set point, and an outlet temperature set point for each chiller device of the chiller system; and The plurality of possible chiller system configurations are stored in a memory of the controller.

17. The method according to claim 16, wherein: Determining, via the controller, at least one optimization solution configured to minimize an energy consumption level of the chiller system includes selecting a target chiller system configuration from the plurality of possible chiller system configurations based on the target cooling load.

18. The method according to claim 16, wherein: Determining, via the controller, a current chiller system configuration corresponding to a current state of the environment includes modeling, via at least one regression-based neural network, the plurality of possible chiller system configurations based on the current state of the environment.

19. The method according to claim 11, further comprising: executing at least one first action of the optimization solution; confirming, via the controller, steady state operation of the chiller system for at least a threshold duration after performing the at least one first action; determining, via the controller, a subsequent energy consumption level based on the steady state operation of the chiller system for at least the threshold duration; determining a difference between the subsequent energy consumption level and the current energy consumption level; as well as When the determined difference exceeds a threshold level, at least one second action of the optimized solution is performed.

20. The method according to claim 11, wherein: Determining the current state of the environment includes determining: Ambient air temperature, the inlet temperature of the coolant fluid entering the environment, the inlet flow rate of the coolant fluid into the environment, and The outlet temperature of the coolant fluid leaving the environment.

Images