Air conditioning system and control method of air conditioning system

Abstract

The application relates to the technical field of household appliance control, in particular to an air conditioning system and a control method of the air conditioning system. The control method of the air conditioning system comprises the following steps: inputting outdoor environment information, first operation information and load aggregation parameters to an upper-layer intelligent agent according to an upper-layer period to obtain an energy supply side decision variable of an energy supply side device; inputting indoor environment information, second operation information and penetration setting parameters to a lower-layer intelligent agent corresponding to each load side device according to a lower-layer period to obtain at least two load side demand variables corresponding to at least two load side devices; the time length of the upper-layer period is greater than the time length of the lower-layer period; and the at least two load side devices are controlled based on the at least two load side demand variables, and the energy supply side device is controlled based on the energy supply side decision variable. The application can realize accurate matching between an energy supply side scheduling strategy and actual cold and heat demands of a load side, and avoid energy waste caused by oversupply and comfort insufficiency caused by undersupply.

Description

Technical Field

[0001] This application relates to the field of air conditioning technology, and in particular to an air conditioning system and a control method for the air conditioning system. Background Technology

[0002] With the increasing popularity of residential and central air conditioning systems, especially in large building air conditioning systems, the power supply-side equipment (chilled water units, chilled / cooling water pumps, cooling towers) and the load-side equipment (fan coil units and air handling units in each floor / area) exhibit significant thermodynamic and hydraulic coupling characteristics during operation. The cooling output and flow distribution of the power supply-side equipment are difficult to precisely match with the real-time changing cooling and heating demands of the load-side equipment. Furthermore, the system has significant transmission delays and thermal inertia, which easily leads to problems such as control response lag, frequent valve and fan oscillations, and frequent fluctuations in operating conditions.

[0003] To ensure that all load-side equipment remains unaffected under the most extreme operating conditions, the water supply temperature parameters of the power supply equipment are typically set to extremely low levels. This approach leads to a mismatch between the power supply scheduling strategy and the actual heating and cooling demands of the load side, resulting in low overall energy efficiency. Summary of the Invention

[0004] In order to solve the above-mentioned technical problems, or at least partially solve the above-mentioned technical problems, this application provides an air conditioning system and a control method for the air conditioning system.

[0005] This application provides an air conditioning system, including: The power supply side equipment and the load side equipment are connected by a water supply pipeline. The controller communicates with the power supply and load side equipment. It contains upper-level and lower-level intelligent agents and is configured as follows: The system acquires outdoor environmental information, indoor environmental information, first operating information, second operating information, load aggregation parameters, and infiltration setting parameters. The first operating information includes the operating information of the power supply-side equipment; the second operating information includes the operating information of at least two load-side equipment; the load aggregation parameters are generated based on the aggregation of load-side demand from at least two load-measuring devices output by the lower-level intelligent agent; the infiltration setting parameters are generated based on the power supply-side decision quantities from the power supply-side equipment output by the upper-level intelligent agent; the load-side demand is used to characterize the cooling capacity adjustment needs of the corresponding load-side equipment; and the power supply-side decision quantities are used to characterize the power supply capacity and hydraulic operating condition targets of the power supply-side equipment. According to the upper-level cycle, outdoor environmental information, first operation information and load aggregation parameters are input to the upper-level intelligent agent to obtain the power supply side decision quantity of the power supply side equipment; And according to the lower-level cycle, indoor environmental information, second operation information and penetration setting parameters are input to the lower-level intelligent agent corresponding to each load-side device to obtain at least two load-side demands corresponding to at least two load-side devices; wherein, the duration of the upper-level cycle is longer than the duration of the lower-level cycle; Control at least two load-side devices based on at least two load-side demands, and control the power supply-side devices based on power supply-side decisions.

[0006] This application also provides a control method for an air conditioning system, applied to the air conditioning system in any of the above technical solutions. The control method includes: acquiring outdoor environmental information, indoor environmental information, first operating information, second operating information, load aggregation parameters, and permeability setting parameters; wherein, the first operating information includes operating information of the power supply side equipment, the second operating information includes operating information of at least two load side equipment, the load aggregation parameters are parameters generated based on the aggregation of load side demand from at least two load measurement devices output by the lower-level intelligent agent, and the permeability setting parameters are parameters generated based on the power supply side decision quantities of the power supply side equipment output by the upper-level intelligent agent; the load side demand is used to characterize the cooling capacity adjustment demand of the corresponding load side equipment, and the power supply side decision quantities are used to characterize the power supply capacity and hydraulic condition targets of the power supply side equipment; According to the upper-level cycle, outdoor environmental information, first operation information and load aggregation parameters are input to the upper-level intelligent agent to obtain the power supply side decision quantity of the power supply side equipment; And according to the lower-level cycle, indoor environmental information, second operation information and penetration setting parameters are input to the lower-level intelligent agent corresponding to each load-side device to obtain at least two load-side demands corresponding to at least two load-side devices; wherein, the duration of the upper-level cycle is longer than the duration of the lower-level cycle; Control at least two load-side devices based on at least two load-side demands, and control the power supply-side devices based on power supply-side decisions.

[0007] The air conditioning system and control method provided in this application embodiment can input outdoor environmental information, first operating information and load aggregation parameters to the upper-level intelligent agent according to the upper-level cycle to obtain the energy supply side decision quantity of the energy supply side equipment. According to the lower-level cycle, indoor environmental information, second operating information and permeability setting parameters are input to the lower-level intelligent agent corresponding to each load side equipment to obtain at least two load side demands corresponding to at least two load side equipment. The at least two load side equipment are controlled respectively based on the at least two load side demands, and the energy supply side equipment is controlled based on the energy supply side decision quantity.

[0008] Because the duration of the upper-level cycle is longer than that of the lower-level cycle, the upper-level agent completes global energy supply decisions in a slower cycle, while the lower-level agent responds to the real-time needs of the load-side terminal equipment in a faster cycle. This avoids increased energy consumption and equipment damage caused by frequent adjustments to the cooling source equipment, while ensuring the terminal load side's rapid response to environmental changes, thus achieving efficient coordination between the slow-changing cooling source and the fast-changing terminal.

[0009] Furthermore, the permeation setting parameters output by the upper-level intelligent agent are pre-distributed to each lower-level intelligent agent as input for their high-frequency decision-making, enabling the terminal controllers to anticipate upcoming changes in the operating conditions of the energy supply side. The upper-level intelligent agent bases its energy supply side decision-making on the load aggregation parameters generated by all load-side devices, rather than relying on fixed empirical values, achieving precise matching between the energy supply side scheduling strategy and the actual heating and cooling demands of the load side. By aggregating the real-time status and predicted cooling load of each terminal, the upper-level intelligent agent can dynamically optimize key parameters such as water supply temperature setpoints and flow distribution, avoiding energy waste due to oversupply and insufficient comfort due to undersupply, thus achieving physical coordination and optimal energy efficiency between the energy supply side and the load side. Controlling at least two load-side devices based on at least two load-side demands, and controlling the energy supply side devices based on energy supply side decision-making, can prevent over-cooling by air conditioning, improving overall comprehensive energy efficiency. Attached Figure Description

[0010] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 The topology diagram of an air conditioning system in some embodiments of this application is shown; Figure 2 The diagram shows a connection diagram between the power supply side equipment and the load side equipment in some embodiments of this application; Figure 3 A schematic diagram of the cooling water circulation of the power supply side equipment in some embodiments of this application is shown; Figure 4 A schematic diagram of the chilled water circulation of the power supply side equipment in some embodiments of this application is shown; Figure 5 The diagram shows connection schematics of load-side devices in some embodiments of this application; Figure 6 The diagram shows a schematic representation of the controller in some embodiments of this application; Figure 7 The diagram illustrates the information interaction between the upper-layer intelligent agent and the lower-layer intelligent agent in some embodiments of this application; Figure 8 This application shows schematic diagrams illustrating the process of training various intelligent agents using a cloud-based training platform in some embodiments; Figure 9 A flowchart of a control method for an air conditioning system according to some embodiments of this application is shown; Figure 10 A structural block diagram of the control device of an air conditioning system in some embodiments of this application is shown. Detailed Implementation

[0013] To make the objectives and implementation methods of this application clearer, the exemplary implementation methods of this application will be clearly and completely described below with reference to the accompanying drawings of the exemplary embodiments of this application. Obviously, the exemplary embodiments described are only some embodiments of this application, and not all embodiments.

[0014] It should be noted that the brief descriptions of terms in this application are only for the convenience of understanding the embodiments described below, and are not intended to limit the embodiments of this application. Unless otherwise stated, these terms should be understood in their ordinary and common meaning.

[0015] The terms "first," "second," "third," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar or related objects or entities, and do not necessarily imply a specific order or sequence, unless otherwise specified. It should be understood that such terms are interchangeable where appropriate.

[0016] The terms “comprising” and “having”, and any variations thereof, are intended to cover but not exclude inclusion, for example, a product or device that includes a range of components is not necessarily limited to all of the components that are clearly listed, but may include other components that are not clearly listed or that are inherent to such product or device.

[0017] This application is applicable to air conditioning systems, especially air conditioning systems in large buildings.

[0018] In related technologies, the power supply-side equipment (chilled water units, chilled / cooling water pumps, cooling towers) and the load-side equipment (fan coil units and air handling units in each floor / area) exhibit significant thermodynamic and hydraulic coupling characteristics during operation. Specifically, the cooling output and flow distribution of the power supply-side equipment are difficult to precisely match with the real-time changing cooling and heating demands of the load-side equipment. In addition, the system has significant transmission delays and thermal inertia, which can easily lead to problems such as control response lag, frequent valve and fan oscillations, and frequent fluctuations in operating conditions.

[0019] To solve the above-mentioned technical problems, this application combines Figures 1 to 10 An air conditioning system and its control method are proposed.

[0020] Figure 1 The topology diagram of an air conditioning system in some embodiments of this application is shown. Figure 2 The diagram shows a connection diagram between the power supply side equipment and the load side equipment in some embodiments of this application. Figure 3 A schematic diagram of the cooling water circulation of the power supply-side equipment in some embodiments of this application is shown. Figure 4 A schematic diagram of the chilled water circulation in the power supply-side equipment of some embodiments of this application is shown. Figure 5 A connection diagram of the load-side equipment in some embodiments of this application is shown. For example... Figures 1 to 5 As shown, the air conditioning system 100 includes a controller 200, a power supply side device 110, a first load side device 120, and a second load side device 130.

[0021] The controller 200 is communicatively connected to the power supply device 110, the first load-side device 120, and the second load-side device 130. The controller 200 deploys upper-layer and lower-layer intelligent agents. The upper-layer intelligent agent manages the power supply device 110; the lower-layer intelligent agent manages the first load-side device 120. The communication protocol between the power supply device 110, the first load-side device 120, the second load-side device 130, and the controller 200 is not limited to building automation and control network protocols, but can also be replaced by industrial communication protocols or message queue telemetry transmission protocols, etc., industrial IoT communication protocols.

[0022] like Figure 2 As shown, Figure 2 The diagram illustrates an air conditioning system where the power supply side device 110 is connected to the first load side device 120 and the second load side device 130 via water supply lines. The power supply side device 110 includes, but is not limited to, a chiller, a cooling tower, and a cooling water pump. The load side devices include, but are not limited to, fan coil units and air handling units.

[0023] like Figure 3 and Figure 4As shown, the power supply side equipment 110 includes a first chiller 111, a second chiller 112, a cooling tower 113, and a cooling water pump 114. The cooling water pump 114 is connected to the condenser inlets of both the first chiller 111 and the second chiller 112; the condenser outlets of the first chiller 111 and the second chiller 112 are connected to the cooling tower 113. One end of the chilled water pump 115 is connected to the first load side equipment 120, and the other end is connected to the evaporator inlets of both the first chiller 111 and the second chiller 112. The evaporator outlets of the first chiller 111 and the second chiller 112 are connected to the first load side equipment 120. It is understood that the first load side equipment 120 can also be replaced by the second load side equipment 130.

[0024] The outlet of the cooling water pump 114 is connected to the condenser inlets of the first chiller 111 and the second chiller 112 respectively; the condenser outlets of the first chiller 111 and the second chiller 112 are connected to the cooling tower 113 together, thus forming a cooling water circulation loop.

[0025] like Figure 4 As shown, the power supply side device 110 also includes a chilled water pump 115. For example, the inlet of the chilled water pump 115 is connected to the return water end of the first load side device 120, and the outlet of the chilled water pump 115 is connected to the evaporator inlets of the first chiller 111 and the second chiller 112, respectively; the evaporator outlets of the first chiller 111 and the second chiller 112 are connected together to the water supply end of the first load side device 120, thereby forming a chilled water circulation loop. It is understood that the aforementioned first load side device 120 can also be replaced by the second load side device 130.

[0026] Optionally, the power supply equipment 110 may also include more chiller units, without limitation.

[0027] like Figure 5 As shown, the first load-side equipment 120 is the fan coil unit 121 of the first area, and the second load-side equipment 130 is the air handling unit 122 of the second area. The fan coil unit 121 and the air handling unit 122 are respectively connected to the power supply-side equipment 110. The fan coil unit 121 and the air handling unit 122 are arranged in parallel. The water inlet of each equipment is uniformly connected to the chilled water supply pipeline on the evaporator outlet side of the power supply side, and the water outlet of each equipment is uniformly connected to the chilled water return pipeline on the chilled water pump inlet side, so as to realize centralized cooling of multi-area terminal equipment (i.e., load-side equipment).

[0028] Understandably, the connection relationship of the second load-side device 130 is similar to that of the first load-side device 120, and will not be described in detail here.

[0029] Figure 6The following are schematic diagrams of the controller structure in some embodiments of this application, such as... Figure 6 As shown, the controller 200 includes a processor 201 and a memory 202 storing computer program instructions.

[0030] Specifically, the processor 201 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this disclosure.

[0031] Memory 202 may include a large-capacity storage device for information or instructions. For example, and not limitingly, memory 202 may include a hard disk drive (HDD), a floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 202 may include removable or non-removable (or fixed) media. Where appropriate, memory 202 may be internal or external to the integrated gateway device. In a particular embodiment, memory 202 is a non-volatile solid-state memory. In a particular embodiment, memory 202 includes read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (Electrically Programmable ROM, EPROM), an electrically erasable PROM (EEPROM), an electrically alterable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0032] The processor 201 executes the steps of the control method for the air conditioning system 100 provided in this embodiment of the present disclosure by reading and executing computer program instructions stored in the memory 202.

[0033] In one example, the controller 200 may also include a transceiver 203 and a bus 204. Wherein, as... Figure 6 As shown, the processor 201, memory 202 and transceiver 203 are connected via bus 204 and communicate with each other.

[0034] Bus 204 includes peripheral component interconnect (PCI) lines or extended industry standard architecture (EISA) buses. Bus 204 can be divided into address bus, data bus, control bus, etc.

[0035] The following describes the operation of the air conditioning system 100 using the controller 200 executing the control method of the air conditioning system 100 as an example: The controller 200 is configured to: acquire outdoor environmental information, indoor environmental information, first operating information, second operating information, load aggregation parameters, and permeability setting parameters; input the outdoor environmental information, first operating information, and load aggregation parameters to the upper-level intelligent agent according to the upper-level cycle to obtain the energy supply-side decision quantity of the energy supply-side device 110; and input the indoor environmental information, second operating information, and permeability setting parameters to the lower-level intelligent agent corresponding to each load-side device according to the lower-level cycle to obtain at least two load-side demands corresponding to at least two load-side devices; control at least two load-side devices respectively based on at least two load-side demands, and control the energy supply-side device 110 based on the energy supply-side decision quantity.

[0036] In this embodiment of the application, the outdoor environmental information includes outdoor temperature and humidity, solar irradiance, and weather forecast data for a future time period. For example, the future time period is 1 to 3 hours.

[0037] In this embodiment, the first operating information includes the operating information of the power supply side equipment 110. The first operating information includes the current chilled water supply and return temperatures, cooling water supply and return temperatures, the operating status and real-time load rate of each chiller, the operating frequency of the chilled water pumps and cooling water pumps, and the operating frequency of the cooling tower fans. The operating status of the chiller includes an on or off state.

[0038] In this embodiment of the application, the indoor environmental information includes the indoor temperature, indoor humidity, indoor carbon dioxide concentration, and measured values ​​of the cooling load of the service area of ​​the load-side equipment in the cycle of the floor above and below.

[0039] In this embodiment, the second operating information includes operating information of at least two load-side devices. The at least two load-side devices include a first load-side device 120 and a second load-side device 130; the second operating information includes the device status of each load-side device. For example, the device status includes the fan operating frequency and the current water valve opening.

[0040] In this embodiment, the load aggregation parameters are generated based on the aggregated load-side demand of at least two load-side devices output by the lower-level agent. The load aggregation parameters include the total predicted cooling load of the at least two load-side devices in the next upper-level cycle, the number of starved nodes, and the maximum water valve opening of the at least two load-side devices in the current upper-level cycle. Starved nodes indicate load-side devices with insufficient cooling capacity in the current upper-level cycle.

[0041] The load-side demand is used to characterize the cooling capacity adjustment requirements of the corresponding load-side equipment. The load-side demand includes the water valve opening adjustment and fan frequency adjustment for each load-side equipment in the next lower-level cycle, as well as the predicted cooling load for each load-side equipment in the next upper-level cycle. The water valve opening adjustment and fan frequency adjustment are both within preset ranges. These preset ranges can be -5% to 5% of the maximum water valve opening and -5% to 5% of the maximum fan frequency.

[0042] Understandably, limiting the water valve opening adjustment and fan frequency adjustment to within ±5% of the maximum water valve opening and maximum fan frequency preset range allows for small-amplitude, gradual fine-tuning. This avoids large abrupt changes in valve opening and fan frequency, effectively suppressing severe fluctuations in system hydraulic and wind pressure, as well as control overshoot and oscillation. It also reduces water hammer in pipelines, equipment impact, and component wear, extending equipment lifespan. Simultaneously, it can smoothly correct terminal load deviations, preventing drastic changes in indoor temperature, humidity, and air supply environment, ensuring regional environmental comfort and operational stability. It adapts to the gradual changes in cooling load under tiered control cycles, reserving reasonable adjustment margins for upper-level global coordinated control based on predicted cooling loads, and achieving stable, orderly, and low-disturbance refined control on the load side.

[0043] In this embodiment, the permeation setting parameters are parameters generated based on the power supply-side decision quantities of the power supply-side equipment output by the upper-level intelligent agent. The permeation setting parameters include the predicted water supply temperature setpoint and the actual water supply temperature value within the current upper-level cycle. The predicted water supply temperature setpoint is used to characterize the chilled water supply temperature setpoint of the power supply-side equipment in the next upper-level cycle. It is understood that the predicted water supply temperature setpoint is within an allowable range (e.g., 5℃~12℃).

[0044] Among them, the energy supply-side decision-making parameters are used to characterize the energy supply capacity and hydraulic operating target of the energy supply-side equipment. The energy supply-side decision-making parameters include the predicted water supply temperature setpoint, the differential pressure between the chilled water supply and return water in the chilled water network, and the chiller unit combination strategy.

[0045] The combination strategy is used to characterize the strategy for the combined operation of various chiller units. That is, the combination strategy is used to indicate the combination of chiller units that are turned on and off in the current time period, as well as the load distribution ratio of each chiller unit.

[0046] Optionally, the combination strategy may include the action mask corresponding to each power supply device 110. When the load rate of the target power supply device is less than the minimum safe load rate, or when the load rate is greater than the maximum rated load rate, the action mask corresponding to the target power supply device is a preset value.

[0047] The preset value is used to indicate that the target power supply device is shielded. For example, the preset value can be 1. That is, when the action mask corresponding to the target power supply device is 1, it indicates that the target power supply device is shielded and does not participate in cooling or heating.

[0048] The minimum safe load rate and the maximum rated load rate mentioned above can be set according to actual conditions and are not limited. For example, the minimum safe load rate can be 30% and the maximum rated load rate can be 100%.

[0049] Understandably, by using an action masking mechanism based on the host load range, unreasonable combinations of actions below the minimum safe load rate and exceeding the rated load limit are forcibly blocked from the source of strategy decision-making. This effectively avoids abnormal operating conditions such as low-load surge of the chiller host, frequent start-stop of the compressor, and overload, eliminating potential safety hazards in equipment operation and improving the overall operational stability and service life of the chiller unit.

[0050] In this embodiment, the duration of the upper-level cycle is longer than the duration of the lower-level cycle. The upper-level cycle is also called a low-frequency decision cycle, for example, the duration of the upper-level cycle is 15 to 30 minutes. The lower-level cycle is also called a high-frequency decision cycle, for example, the duration of the lower-level cycle is 30 to 60 seconds.

[0051] In this embodiment, an upper-layer intelligent agent manages all power supply-side devices. A lower-layer intelligent agent manages one load-side device (also known as an end device).

[0052] Specifically, when acquiring the permeation setting parameters, the controller is configured as follows: the controller 200 acquires the predicted water supply temperature setpoint and the actual water supply temperature; and determines the predicted water supply temperature setpoint and the actual water supply temperature as the permeation setting parameters.

[0053] The actual water supply temperature value is the temperature value collected by the water temperature sensor during the current upper-level cycle.

[0054] Understandably, sending the predicted water supply temperature setpoint from the energy supply side and the actual current water supply temperature together as infiltration setting parameters to the lower-level intelligent agent on the load side allows the lower-level intelligent agent to grasp the current operating condition and future trend of the cold source. Based on these two parameters, it can predict the magnitude and direction of changes in cooling capacity and plan the adjustment rhythm in advance. This effectively eliminates the control gap caused by the decoupling of fast and slow cycles, avoids the passive response mode of the terminal that relies solely on the current actual value and has no expectation of future changes in traditional control, significantly reduces temperature fluctuations and control oscillations during operating condition switching, avoids overshoot and repeated adjustments caused by the terminal passively chasing the temperature after the cold source is adjusted, reduces the adjustment frequency and energy consumption of the terminal equipment, and realizes pre-synchronous control of the cold source and the terminal and energy saving in a coordinated manner between supply and demand.

[0055] Specifically, when acquiring load aggregation parameters, the controller is configured as follows: the controller 200 acquires the water valve opening of each load-side device in the current upper-level cycle, and the temperature deviation between the indoor temperature and the set temperature of the service area where each load-side device is located; calculates the sum of the predicted cooling loads of at least two load-side devices in the next upper-level cycle, to obtain the total predicted cooling load of at least two load-side devices in the next upper-level cycle; determines the number of starved nodes based on the water valve opening and temperature deviation; determines the maximum water valve opening of at least two load-side devices in the current upper-level cycle based on the water valve opening of each load-side device in the current upper-level cycle; and determines the total predicted cooling load, the number of starved nodes, and the maximum water valve opening as load aggregation parameters.

[0056] The number of starved nodes refers to the number of load-side devices whose water valve opening exceeds a first threshold and whose temperature deviation remains above a second threshold within a preset time period. The number of starved nodes is used to characterize the number of end devices (load-side devices) in a state of insufficient cooling, so as to influence the upper-level agent to determine whether to increase the cooling capacity.

[0057] Temperature deviation is the temperature difference between the indoor temperature of the service area where the load-side equipment is located and the set temperature.

[0058] The maximum water valve opening is the maximum value of the water valve openings of at least two load-side devices. The maximum water valve opening is used to characterize the terminal device with the least cooling capacity in the pipe network, so as to influence the upper-level agent to determine whether the current water supply temperature and water supply pressure difference are sufficient.

[0059] Understandably, defining the total predicted cooling load, the number of starved nodes, and the maximum water valve opening as load aggregation parameters and transmitting these parameters to the upper-level agent allows the agent to accurately reflect the overall cooling demand for the next upper-level cycle by summing the predicted cooling loads. It also effectively identifies the number of starved nodes with insufficient cooling and objectively characterizes the current adjustment margin and operational saturation state of the terminal equipment using the maximum water valve opening. This provides comprehensive and reliable input for the upper-level agent to formulate energy supply decisions, ensuring that the energy supply-side control strategy accurately matches the actual cooling and heating demands of the terminals. This effectively alleviates problems such as uneven local cooling, terminal adjustment saturation, and insufficient cooling, and improves the overall operational stability, comfort, and supply-demand coordination accuracy of the air conditioning control system.

[0060] Specifically, when acquiring the predicted cooling load of each load-side device output by the upper-level agent for the next upper-level cycle, the controller is configured as follows: the controller 200 acquires the current indoor temperature of the service area where each load-side device is located, the historical water valve opening at multiple historical moments, the historical indoor temperature of the service area, the historical indoor carbon dioxide concentration, and the historical outdoor meteorological data collected by the sensors; based on the historical indoor temperature, historical indoor carbon dioxide concentration, historical water valve opening, and historical outdoor meteorological data, a multi-dimensional time-series feature matrix is ​​generated; the multi-dimensional time-series feature matrix is ​​input into a preset model to output the predicted indoor temperature for the next upper-level cycle; the predicted indoor temperature, the air volume corresponding to the service area where each load-side device is located, the duration of the upper-level cycle, the preset target temperature, and the steady-state basic heat transfer load corresponding to the current indoor temperature are input into the lower-level agent corresponding to each load-side device to output the predicted cooling load of each load-side device for the next upper-level cycle.

[0061] The preset model is based on a lightweight gated recurrent unit (GRU).

[0062] Understandably, by comprehensively collecting various time-series data, including historical water valve opening, indoor temperature and humidity, carbon dioxide concentration, and outdoor weather data, a multi-dimensional time-series feature matrix is ​​constructed. This matrix fully explores the coupling relationship between the environment, equipment operation, and load changes, and accurately predicts the indoor temperature for the next cycle based on a preset model. Then, by combining key parameters such as regional air volume, cycle duration, target temperature, and steady-state heat transfer load, the lower-level intelligent agent calculates the predicted cooling load for each load-side device in a refined manner. This multi-factor fusion prediction method effectively improves the accuracy of cooling load prediction, avoids deviations caused by single-parameter predictions, and makes load demand predictions more closely aligned with actual operating conditions. This provides reliable data support for load aggregation and upper-level energy supply decisions, ensuring precise matching of supply and demand, and improving system control stability and overall energy-saving performance.

[0063] For example, taking the current time as t, the controller 200 acquires the current indoor temperature of the service area where each load-side device is located, collected by the sensor. At the current time, each terminal agent acquires historical water valve opening, historical indoor temperature, historical indoor carbon dioxide concentration, and historical outdoor meteorological data within a historical sliding window (e.g., multiple historical times over the past H time steps) through the local gateway. Millisecond-level downsampling and timestamp alignment are performed using a time-series database. Based on historical indoor temperature, historical indoor carbon dioxide concentration, historical water valve opening, and historical outdoor meteorological data, a multi-dimensional time-series feature matrix is ​​constructed. This multi-dimensional time-series feature matrix is ​​input into a GRU model. Under the baseline strategy of assuming "maintaining the current water valve opening unchanged for a period of time in the future," the GRU model performs autoregressive forward propagation, outputting the indoor temperature evolution trajectory for the next k time steps, thus obtaining the predicted indoor temperature for the next upper-level cycle (the kth time step).

[0064] The time-series database is used to store high-frequency time-series data from the end-sensor sensors. For example, the high-frequency time-series data includes valve opening degree and indoor temperature and humidity, with a sampling frequency of 1 minute.

[0065] Furthermore, after obtaining the indoor temperature evolution trajectory for the next k time steps, based on the indoor air thermodynamic mapping equation, the predicted indoor temperature, the air volume corresponding to the service area of ​​each load-side device, the duration of the upper-level cycle, the preset target temperature, and the steady-state basic heat transfer load corresponding to the current indoor temperature are input into the lower-level agent corresponding to each load-side device, and the predicted cooling load of each load-side device in the next upper-level cycle is output.

[0066] The air volume corresponding to the service area of ​​each load-side device is pre-stored in the database topology relation library.

[0067] The preset target temperature is the target comfort temperature set by the end user.

[0068] The steady-state basic heat transfer load corresponding to the current indoor temperature is the steady-state basic heat transfer load that needs to be overcome to maintain the current room temperature.

[0069] The predicted cooling load satisfies the following formula:

[0070] in, This represents the predicted cooling load of the i-th load-side device at time t+k. This indicates the specific heat capacity of air under constant pressure. Indicates indoor air density; This represents the air volume corresponding to the service area where the i-th load-side device is located; Indicates the preset target temperature; Indicates the duration of the upper-level cycle; This represents the steady-state basic heat transfer load corresponding to the current indoor temperature; This indicates the predicted indoor temperature.

[0071] Understandably, the lower-level intelligent agent can accurately quantify the differential cooling load required for temperature and humidity control in each room by combining the deviation between the predicted indoor temperature and the set target temperature with inherent physical parameters such as air specific heat capacity, air density, and effective volume of the area. It can then output the predicted cooling load of each load-side device in the next upper-level cycle, that is, convert the temperature prediction results into a quantified cooling load demand with physical meaning. This makes up for the shortcomings of a single data-driven model that lacks mechanistic constraints and is prone to deviating from actual thermal laws, thereby improving the scientificity and accuracy of cooling load calculation.

[0072] Furthermore, after obtaining the predicted cooling load of each load-side device in the next upper-level cycle, the controller 200 performs bidirectional data flow. For example, the predicted cooling load is used as an additional state input to the lower-level agents, causing the load-side devices (terminal water valves) to perform a small-scale cooling storage action before the physical room temperature rises. The predicted cooling loads of each lower-level agent are aggregated into the total predicted cooling load of the entire system, serving as an important feedforward basis for the upper-level cold source agent to adjust the global chilled water supply temperature.

[0073] like Figure 7 As shown, Figure 7 The diagram illustrates information interaction between upper-layer and lower-layer intelligent agents in some embodiments of this application. Figure 7 In this system, for the upper-layer intelligent agent and at least two lower-layer intelligent agents deployed by the controller 200, a two-way information communication channel is established between the upper-layer intelligent agents and the lower-layer intelligent agents. At the end of each low-frequency decision cycle of the upper-layer intelligent agent, the controller 200 broadcasts the penetration setting parameters output by the upper-layer intelligent agent to all lower-layer intelligent agents through the edge gateway. Similarly, within each upper-layer decision cycle, the controller 200 aggregates the states of all lower-layer intelligent agents and the predicted cold loads predicted by each lower-layer intelligent agent, and reports the load aggregation parameters to the upper-layer intelligent agents.

[0074] Based on the above-mentioned air conditioning control system, the controller can input outdoor environmental information, first operating information and load aggregation parameters to the upper-level intelligent agent according to the upper-level cycle to obtain the energy supply side decision quantity of the energy supply side equipment. According to the lower-level cycle, the controller can input indoor environmental information, second operating information and permeability setting parameters to the lower-level intelligent agent corresponding to each load side equipment to obtain at least two load side demands corresponding to at least two load side equipment. Based on the at least two load side demands, the controller can control the at least two load side equipment respectively, and control the energy supply side equipment based on the energy supply side decision quantity.

[0075] Because the duration of the upper-level cycle is longer than that of the lower-level cycle, the upper-level agent completes global energy supply decisions in a slower cycle, while the lower-level agent responds to the real-time needs of the load-side terminal equipment in a faster cycle. This avoids increased energy consumption and equipment damage caused by frequent adjustments to the cooling source equipment, while ensuring the terminal load side's rapid response to environmental changes, thus achieving efficient coordination between the slow-changing cooling source and the fast-changing terminal.

[0076] Furthermore, the permeation setting parameters (such as predicted water supply temperature and differential pressure setpoints) output by the upper-level intelligent agents are pre-distributed to each lower-level intelligent agent as input for their high-frequency decision-making, enabling the terminal controller to anticipate upcoming changes in the operating conditions on the power supply side. For example, when the water supply temperature setpoint is increased from 7°C to 9°C, the terminal can proactively adjust the water valve opening before the new setpoint takes effect, adapting to changes in cooling capacity in advance. This completely eliminates the control delay caused by the lag in power supply side adjustment and the passive response of the terminal in traditional control, significantly improving the dynamic stability of the system.

[0077] The upper-level intelligent agent bases its energy supply decisions on aggregated load parameters generated from all load-side devices, rather than relying on fixed empirical values. This achieves a precise match between the energy supply scheduling strategy and the actual heating and cooling demands of the load side. By aggregating the real-time status and predicted cooling load of each terminal, the upper-level intelligent agent can dynamically optimize key parameters such as water supply temperature setpoints and flow distribution. This avoids energy waste due to oversupply and insufficient comfort due to undersupply, achieving physical coordination and optimal energy efficiency between the energy supply and load sides.

[0078] Furthermore, according to the upper-level cycle, after inputting outdoor environmental information, first operating information, and load aggregation parameters to the upper-level agent to obtain the energy supply-side decision quantity of the energy supply-side equipment, the controller is configured as follows: the controller 200 obtains the current cooling performance coefficient of the air conditioning system 100 and the number of state transitions of the energy supply-side equipment 110 in the current upper-level cycle; determines the reward score of the upper-level agent in the current upper-level cycle based on the current cooling performance coefficient, the number of hungry nodes, and the number of state transitions; and updates the upper-level agent based on the reward score of the upper-level agent, outdoor environmental information, first operating information, load aggregation parameters, and energy supply-side decision quantity.

[0079] The number of state flips is used to characterize the start-stop change state of the power supply side equipment in the current upper-level cycle.

[0080] In this embodiment, the current cooling performance coefficient is positively correlated with the reward score of the upper-layer agent. The number of hungry nodes and the number of state transitions are negatively correlated with the reward score of the upper-layer agent.

[0081] Understandably, the reward score for the upper-level agent is determined based on the system's cooling performance coefficient, the number of state transitions of the power supply-side equipment, and the number of hungry nodes. The cooling performance coefficient is positively correlated with the reward, while the number of hungry nodes and the number of state transitions are negatively correlated with the reward. This is then combined with outdoor environment, power supply-side operational information, load aggregation parameters, and power supply decision-making parameters to update the agent. Furthermore, based on system energy efficiency, end-point cooling demand, and power supply-side equipment losses, the agent can be guided to continuously optimize power supply decisions. This improves system cooling efficiency, reduces end-point cooling insufficiency, and lowers start-up and shutdown losses of power supply-side equipment. Simultaneously, relying on multi-dimensional data updates, the agent continuously adapts to changing operating conditions, improving decision-making accuracy and the system's long-term operational stability and energy efficiency.

[0082] Specifically, when determining the reward score of the upper-layer agent in the current upper-layer cycle based on the current cooling performance coefficient, the number of hungry nodes, and the number of state transitions, the controller is configured to: calculate the ratio between the current cooling performance coefficient and the historical maximum cooling performance coefficient based on the first algorithm to obtain the coefficient ratio; calculate the product between the coefficient ratio and the corresponding weight value to obtain the cooling performance coefficient reward score; calculate the product between the number of hungry nodes and the corresponding weight value to obtain the cooling adaptability reward score; calculate the product between the number of state transitions and the corresponding weight value to obtain the cooling stability reward score; and determine the sum of the cooling performance coefficient reward score, the cooling adaptability reward score, and the cooling stability reward score as the reward score of the upper-layer agent in the current upper-layer cycle.

[0083] In this embodiment of the application, the first algorithm is: R= × + × + × Where R represents the reward score of the upper-level agent in the current upper-level cycle; This represents the weight value corresponding to the coefficient ratio; Indicates the ratio of coefficients; The weight value represents the number of starved nodes; Indicates the number of starving nodes; This represents the weight value corresponding to the number of state transitions; Indicates the number of state transitions.

[0084] In this embodiment, the weight value corresponding to the number of state transitions can be set as an incremental penalty weight. That is, when the number of state transitions is less than or equal to a set threshold, the weight value corresponding to the number of state transitions is the first weight; when the number of state transitions is greater than the set threshold, an incremental penalty weight is applied, and the weight value corresponding to the number of state transitions is the second weight, which is greater than the first weight. This can prevent mechanical damage to the compressor caused by frequent start-stop cycles.

[0085] In this embodiment of the application, the weights corresponding to the cooling performance coefficient, the number of hungry nodes, and the number of state transitions can be set according to the building owner's operation and maintenance preferences (e.g., whether to focus more on energy saving or more on comfort), without any restrictions.

[0086] Understandably, the cooling performance coefficient reward score can guide the upper-level intelligent agent to optimize scheduling strategies such as chiller combination and water supply temperature, prioritizing the improvement of the overall cooling performance coefficient of the system, effectively reducing the overall energy consumption on the air conditioning cold source side, and achieving energy-saving operation. The cooling adaptability reward score guides the upper-level intelligent agent to rationally allocate cold source output, reduce the problem of supply and demand imbalance at the terminal, and ensure balanced coverage of cooling demand in various areas. The cooling adaptability reward score can guide the upper-level intelligent agent to directly perceive the actual cooling demand of the load-side equipment, avoiding the situation where some load-side equipment is in a state of insufficient cooling for a long time due to the preset water supply temperature setting value being too high.

[0087] Similarly, according to the lower-level cycle, after inputting indoor environmental information, second operating information, and permeation setting parameters to the lower-level intelligent agent corresponding to each load-side device to obtain at least two load-side demands corresponding to at least two load-side devices, the controller is further configured to: acquire the indoor temperature deviation between the current indoor temperature and the preset target temperature, and acquire the change difference of the control quantity corresponding to each load-side device between the current lower-level cycle and the previous lower-level cycle; determine the reward score of the lower-level intelligent agent corresponding to each load-side device in the current lower-level cycle based on the indoor temperature deviation and the change difference; and update each lower-level intelligent agent based on the reward score of each lower-level intelligent agent, indoor environmental information, second operating information, permeation setting parameters, and load-side demand.

[0088] Among them, the indoor temperature deviation and the difference in change were negatively correlated with the reward scores of the lower-level agents.

[0089] Understandably, the reward score is determined based on the difference between the indoor temperature deviation and the cross-cycle change in load demand. The smaller the temperature deviation and the smoother the change in control quantity, the higher the reward score. Then, each lower-level intelligent agent is updated based on indoor environmental information and secondary operational information. This approach guides the lower-level intelligent agents to accurately adapt to the terminal cooling demand, quickly eliminate indoor temperature deviation, ensure environmental comfort, suppress drastic fluctuations in control quantity, reduce adjustment losses of terminal equipment, improve adjustment stability, enable the lower-level intelligent agents to continuously adapt to changes in operating conditions, optimize high-frequency adjustment strategies, further strengthen the coordination with the upper-level intelligent agents, and improve the control accuracy and operational efficiency of the entire air conditioning control system.

[0090] Specifically, when determining the reward score of the lower-level agent corresponding to each load-side device in the current lower-level cycle based on the indoor temperature deviation and the change difference, the controller is also configured to: calculate the product between the indoor temperature deviation and the corresponding weight value based on the second algorithm to obtain the comfort reward score; calculate the product between the change difference and the corresponding weight value to obtain the motion smoothness reward score; and determine the sum of the comfort reward score and the motion smoothness reward score as the reward score of the lower-level agent corresponding to each load-side device in the current lower-level cycle.

[0091] Specifically, when the absolute value of the indoor temperature deviation is greater than the comfort tolerance threshold (e.g., 1℃), the weight value corresponding to the temperature deviation is set as a step-enhanced negative penalty weight value.

[0092] In this embodiment of the application, the second algorithm is: Q= × + ×

[0093] Where Q represents the reward score of the lower-level agent corresponding to each load-side device in the current lower-level cycle; This represents the weight value corresponding to the indoor temperature deviation; Indicates indoor temperature deviation; This represents the weight value corresponding to the change difference; This represents the difference in change.

[0094] Understandably, comfort reward scores can be used to increase penalties for severe deviations from operating conditions such as excessive temperature differences or exceeding temperature limits, guiding lower-level agents to prioritize reducing room temperature deviations and suppressing issues like temperature overshoot and sudden temperature changes. Action smoothness reward scores can be used to impose penalty constraints on the fluctuation amplitude of water valve adjustment increments between different time points, effectively limiting continuous large-scale changes in water valve opening, suppressing high-frequency reciprocating oscillations and frequent adjustments of the valve, and reducing mechanical wear on the actuator.

[0095] In this embodiment, both the upper-layer agent and at least two lower-layer agents are trained on a cloud-based training platform. The cloud-based training platform is deployed on a remote server or private cloud and is responsible for offline training, hyperparameter tuning, and policy updates of the reinforcement learning model. After training of the upper-layer agent and at least two lower-layer agents is completed, the lightweight inference network model parameters corresponding to the upper-layer agent and at least two lower-layer agents are distributed to the edge gateway.

[0096] Optionally, each lower-level agent runs on an edge gateway. When a lower-level agent loses communication with the cloud training platform or the upper-level agent, the controller 200 enters a degraded operation mode. In degraded operation mode, the controller 200 uses the last successfully received permeation setting parameters as input to the lower-level agent, and performs independent comfort adjustments based on indoor environmental information, secondary operating information, and permeation setting parameters to maintain basic room temperature control functions and prevent regional temperature runaway due to a single point of failure.

[0097] Optionally, the controller 200 can set a set of hard constraint rules based on HVAC engineering experience as the final safety boundary. These hard constraint rules include, but are not limited to: chilled water supply temperature setpoint greater than or equal to the lower safety limit temperature; chilled water return temperature greater than or equal to the dew point temperature; continuous operation time of any chiller unit must not exceed the rated upper limit of the equipment; chiller unit start-stop interval greater than or equal to the protection time threshold. When the energy supply-side decision output by the upper-level intelligent agent or the two load-side demand outputs by the lower-level intelligent agent trigger the above hard constraint rules, the energy supply-side decision or load-side demand is automatically intercepted and replaced with the corresponding safe default value to ensure that the equipment is not damaged or the system's operational safety is not compromised under any circumstances.

[0098] The training methods for the upper-level agent and at least two lower-level agents are introduced below.

[0099] In this embodiment, during the training phase, the cloud-based training platform can build a high-fidelity virtual environment based on building energy consumption simulation tools. This environment simulates the thermal characteristics of the target building's envelope, the hydraulic characteristics of the air conditioning water system, the performance curve of the chiller, and typical schedules for heat dissipation from indoor personnel and equipment. The training dataset covers 8760 hours of typical annual meteorological data to ensure that the agent can be adequately trained under various operating conditions.

[0100] Specifically, cloud-based training platforms can perform phased training on upper-layer and lower-layer intelligent agents. For example... Figure 8 As shown, Figure 8 The diagram illustrates a process flow for training various intelligent agents using a cloud-based training platform, as shown in some embodiments of this application. Figure 8 The training process includes a lower-level agent training phase, an upper-level agent training phase, and a joint fine-tuning training phase.

[0101] Lower-level agent training phase: First, the strategy of the upper-level agent is fixed (e.g., a fixed water supply temperature setpoint of 7℃), and the lower-level agent is trained independently. In this phase, the lower-level agent learns how to optimally adjust water valves and fans to maintain comfort in each area and ensure smooth operation under given cooling conditions. The training algorithm employs Multi-Agent Proximal Policy Optimization (MAPPO), utilizing parameter sharing to accelerate convergence.

[0102] Upper-layer agent training phase: After the lower-layer agent's policy converges, the lower-layer agent's policy is fixed, and the upper-layer agent is trained. The upper-layer agent learns, through interaction with the pre-trained lower-layer agent, how to optimally adjust the supply water temperature setpoint, supply and return water pressure difference, and chiller unit combination strategy to maximize the overall system COP while satisfying global comfort constraints. The training algorithm uses the Soft Actor-Critic (SAC) algorithm.

[0103] Understandably, the training algorithm for the upper-layer agent, besides SAC, can be replaced with policy gradient algorithms in continuous action spaces, such as Proximal Policy Optimization (PPO) or Deep Deterministic Policy Gradient (DDPG). Similarly, the training algorithm for the lower-layer agent, besides MAPPO, can be replaced with multi-agent cooperative reinforcement learning algorithms such as value-mixing algorithms or multi-agent deep deterministic policy gradient algorithms.

[0104] Joint Fine-Tuning Training Phase: All policy network parameters for both the upper and lower layer agents are unfrozen, and end-to-end joint fine-tuning training is performed in a complete simulation environment. The learning rate in this phase is set to 1 / 10 to 1 / 5 of that in the previous two phases to prevent the learned policies from being violated. Joint fine-tuning enables the upper and lower layer policies to adapt to each other collaboratively, further improving overall performance.

[0105] The trained Actor inference network corresponding to the agent is distributed to the edge gateway in a containerized manner. After deployment in the actual building, the cloud training platform adopts an online fine-tuning strategy, continuously collects real-world operating data, and incrementally retrains the Critic and Actor networks in the cloud. Every preset time period (2-4 weeks), the updated Actor network parameters are pushed to the edge gateway, thereby realizing the gradual migration and continuous adaptation from the simulation environment to the real environment.

[0106] The Actor and Critic networks can employ multilayer perceptrons, long short-term memory networks, or Transformer architectures. When capturing longer temporal dependencies, long short-term memory networks or Transformers can be selected in the upper-layer agents to enhance temporal modeling capabilities.

[0107] In some embodiments, Figure 9 A flowchart of a control method for an air conditioning system according to some embodiments of this application is shown, such as... Figure 9 As shown, the control method for the air conditioning system is applied to the air conditioning system in any of the above embodiments, and the control method for the air conditioning system includes: S901 acquires outdoor environmental information, indoor environmental information, first operating information, second operating information, load aggregation parameters, and permeability setting parameters.

[0108] S902, according to the upper-level cycle, inputs outdoor environmental information, first operating information and load aggregation parameters to the upper-level intelligent agent to obtain the power supply side decision quantity of the power supply side equipment.

[0109] S903, and according to the lower-level cycle, inputs indoor environmental information, second operation information and permeation setting parameters to the lower-level intelligent agent corresponding to each load-side device, so as to obtain at least two load-side demands corresponding to at least two load-side devices.

[0110] S904 controls at least two load-side devices based on at least two load-side demands, and controls the power supply-side devices based on power supply-side decisions.

[0111] It should be noted that when the air conditioning system is controlled by the control device of the air conditioning system in the embodiments of this application, the specific implementation method is similar to the specific implementation method of the air conditioning system in any of the above embodiments of the present invention. Therefore, for a detailed exemplary description of the control process of the air conditioning system, please refer to the relevant description of the air conditioning system mentioned above. To reduce redundancy, it will not be repeated here.

[0112] In some embodiments, Figure 10 Structural block diagrams of the control device of an air conditioning system in some embodiments of this application are shown, such as... Figure 10 As shown, the control device for the air conditioning system is applied to the air conditioning system in any of the above embodiments, and the control device for the air conditioner includes: The acquisition module 1001 is used to acquire outdoor environmental information, indoor environmental information, first operating information, second operating information, load aggregation parameters, and permeation setting parameters. The load aggregation parameters are generated based on the aggregation of load-side demand from at least two load measuring devices output by the lower-level intelligent agent. The permeation setting parameters are generated based on the energy supply-side decision quantities from the energy supply-side devices output by the upper-level intelligent agent. The load-side demand is used to characterize the cooling capacity adjustment requirements of the corresponding load-side devices. The energy supply-side decision quantities are used to characterize the energy supply capacity and hydraulic condition targets of the energy supply-side devices.

[0113] The processing module 1002 is used to input outdoor environmental information, first operating information, and load aggregation parameters to the upper-level intelligent agent according to the upper-level cycle, so as to obtain the energy supply-side decision quantity of the energy supply-side equipment. It is also used to input indoor environmental information, second operating information, and penetration setting parameters to the lower-level intelligent agent corresponding to each load-side equipment according to the lower-level cycle, so as to obtain at least two load-side demands corresponding to at least two load-side equipment, wherein the duration of the upper-level cycle is longer than the duration of the lower-level cycle.

[0114] The control module 1003 is used to control at least two load-side devices based on at least two load-side demands, and to control the power supply-side devices based on power supply-side decision quantities.

[0115] It should be noted that when the air conditioning system is controlled by the control device of the air conditioning system in the embodiments of this application, the specific implementation method is similar to the specific implementation method of the air conditioning system in any of the above embodiments of the present invention. Therefore, for a detailed exemplary description of the control process of the air conditioning system, please refer to the relevant description of the air conditioning system mentioned above. To reduce redundancy, it will not be repeated here.

[0116] This application discloses a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the methods described in the above embodiments.

[0117] This application discloses a computer program product, including a computer program, which, when executed by a processor, implements the methods described in the above embodiments.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0119] For ease of explanation, the above description has been provided in conjunction with specific embodiments. However, the exemplary discussion above is not intended to be exhaustive or to limit the embodiments to the specific forms disclosed above. Various modifications and variations can be obtained based on the above teachings. The embodiments were chosen and described to better explain the principles and practical applications, thereby enabling those skilled in the art to better utilize the embodiments and various different variations suitable for specific application considerations.

Claims

1. An air conditioning system, characterized in that, include: The power supply side equipment and the load side equipment are connected by a water supply pipeline; The controller is communicatively connected to the power supply-side equipment and the load-side equipment. The controller contains upper-layer and lower-layer intelligent agents and is configured as follows: Acquire outdoor environmental information, indoor environmental information, first operating information, second operating information, load aggregation parameters, and infiltration setting parameters; wherein, the first operating information includes the operating information of the energy supply-side equipment, the second operating information includes the operating information of at least two load-side equipment, and the load aggregation parameters are parameters generated based on the aggregation of the load-side demand of at least two load-side equipment output by the lower-level agent; the load aggregation parameters include total predicted cooling load, number of starved nodes, and maximum water valve opening; the total predicted cooling load is used to characterize the cooling load demand of at least two load-side equipment in the next upper-level cycle; the number of starved nodes is the number of load-side equipment whose water valve opening is greater than a first threshold and whose temperature deviation is continuously greater than a second threshold within a preset time period, and the temperature deviation is... The difference is the temperature deviation between the indoor temperature of the service area where the load-side equipment is located and the set temperature; the maximum water valve opening is the maximum value of the water valve openings of at least two load-side equipments; the permeation setting parameter is a parameter generated based on the energy supply-side decision quantity of the energy supply-side equipment output by the upper-level agent; the permeation setting parameter includes a predicted water supply temperature setpoint and an actual water supply temperature; the predicted water supply temperature setpoint is used to characterize the water supply temperature setpoint of the energy supply-side equipment in the next upper-level cycle; the actual water supply temperature is the temperature value collected by the water temperature sensor in the current upper-level cycle; the load-side demand is used to characterize the cooling capacity adjustment demand of the corresponding load-side equipment; the energy supply-side decision quantity is used to characterize the energy supply capacity and hydraulic condition target of the energy supply-side equipment. According to the upper-level cycle, the outdoor environmental information, the first operating information, and the load aggregation parameters are input to the upper-level intelligent agent to obtain the energy supply side decision quantity of the energy supply side equipment; And according to the lower-level cycle, the indoor environment information, the second operation information and the permeation setting parameters are input to the lower-level intelligent agent corresponding to each load-side device to obtain at least two load-side demands corresponding to at least two load-side devices; wherein, the duration of the upper-level cycle is greater than the duration of the lower-level cycle; The at least two load-side devices are controlled based on the at least two load-side demands, and the energy-side devices are controlled based on the energy-side decision.

2. The system according to claim 1, characterized in that, The energy supply side decision quantity includes the predicted water supply temperature setpoint; When the permeation setting parameters are obtained, the controller is configured as follows: Obtain the predicted water supply temperature setpoint and the actual water supply temperature; The predicted water supply temperature setpoint and the actual water supply temperature are determined as the permeability setting parameter.

3. The system according to claim 1, characterized in that, The load-side demand includes the predicted cooling load, which is used to characterize the cooling load demand of each load-side device in the next upper-level cycle. When acquiring the load aggregation parameters, the controller is configured as follows: The water valve opening degree of each load-side device in the current upper-level cycle and the temperature deviation between the indoor temperature and the set temperature in the service area of ​​each load-side device are obtained. The sum of the predicted cooling loads of at least two of the load-side devices in the next upper-level cycle is calculated to obtain the total predicted cooling load of the at least two of the load-side devices in the next upper-level cycle. The number of starved nodes is determined based on the water valve opening degree and the temperature deviation. Based on the water valve opening of each load-side device in the current upper-level cycle, determine the maximum water valve opening of at least two of the load-side devices in the current upper-level cycle; The total predicted cooling load, the number of starved nodes, and the maximum water valve opening are determined as the load aggregation parameters.

4. The system according to claim 3, characterized in that, When acquiring the predicted cooling load for each load-side device in the next upper-level cycle from the upper-level agent, the controller is configured to: Acquire the current indoor temperature, historical water valve opening at multiple historical moments, historical indoor temperature, historical indoor carbon dioxide concentration, and historical outdoor meteorological data of the service area where each load-side device is located, collected by the sensor. Based on the historical indoor temperature, the historical indoor carbon dioxide concentration, the historical water valve opening degree, and the historical outdoor meteorological data, a multidimensional time-series feature matrix is ​​generated; The multidimensional time-series feature matrix is ​​input into a preset model, and the predicted indoor temperature for the next upper-level cycle is output. The predicted indoor temperature, the air volume corresponding to the service area of ​​each load-side device, the duration of the upper-level cycle, the preset target temperature, and the steady-state basic heat transfer load corresponding to the current indoor temperature are input into the lower-level agent corresponding to each load-side device, and the predicted cooling load of each load-side device in the next upper-level cycle is output.

5. The system according to any one of claims 1 to 4, characterized in that, The load aggregation parameters include the number of starved nodes. After inputting the outdoor environmental information, the first operational information, and the load aggregation parameters to the upper-layer agent according to the upper-layer cycle to obtain the power supply-side decision quantity of the power supply-side equipment, the controller is configured as follows: The current cooling performance coefficient of the air conditioning system and the number of state flips of the energy supply side equipment are obtained in the current upper-level cycle. The number of state flips is used to characterize the start-stop change state of the energy supply side equipment in the current upper-level cycle. Based on the current cooling performance coefficient, the number of hungry nodes, and the number of state transitions, the reward score of the upper-layer agent in the current upper-layer cycle is determined. The current cooling performance coefficient is positively correlated with the reward score of the upper-layer agent, while the number of hungry nodes and the number of state transitions are negatively correlated with the reward score of the upper-layer agent. The upper-layer agent is updated based on its reward score, outdoor environmental information, the first operational information, load aggregation parameters, and energy supply-side decision quantity.

6. The system according to claim 5, characterized in that, When determining the reward score of the upper-layer agent in the current upper-layer cycle based on the current cooling performance coefficient, the number of hungry nodes, and the number of state transitions, the controller is configured as follows: Calculate the ratio between the current coefficient of performance (COP) and the historical maximum COP to obtain the COP ratio. The product of the coefficient ratio and the corresponding weight value is calculated to obtain the cooling performance coefficient bonus score; Calculate the product between the number of hungry nodes and their corresponding weight values ​​to obtain the cooling adaptation reward score; The product of the number of state transitions and the corresponding weight value is calculated to obtain the cooling stability reward score; The sum of the cooling performance coefficient reward score, the cooling adaptability reward score, and the cooling stability reward score is determined as the reward score of the upper-layer agent in the current upper-layer cycle.

7. The system according to any one of claims 1 to 4, characterized in that, After inputting the indoor environmental information, the second operating information, and the permeability setting parameters to the lower-level intelligent agent corresponding to each load-side device according to the lower-level cycle, so as to obtain at least two load-side demands corresponding to at least two load-side devices, the controller is further configured to: The indoor temperature deviation between the current indoor temperature and the preset target temperature is obtained, and the change difference of the control quantity corresponding to each load-side device between the current lower layer cycle and the previous lower layer cycle is obtained. Based on the indoor temperature deviation and the change difference, determine the reward score of the lower-level agent corresponding to each load-side device in the current lower-level cycle; Each lower-level agent is updated based on its reward score, indoor environment information, second operational information, penetration setting parameters, and load-side demand.

8. The system according to claim 7, characterized in that, When determining the reward score of the lower-level agent corresponding to each load-side device in the current lower-level cycle based on the indoor temperature deviation and the change difference, the controller is further configured to: The comfort bonus score is obtained by multiplying the indoor temperature deviation by the corresponding weight value. The product of the change difference and the corresponding weight value is calculated to obtain the action smoothness reward score; The sum of the comfort reward score and the motion smoothness reward score is determined as the reward score of the lower-level agent corresponding to each load-side device in the current lower-level cycle.

9. The system according to any one of claims 1 to 4, characterized in that, The power supply side equipment includes multiple chillers; the power supply side decision quantity includes a combination strategy, which is used to characterize the strategy for the combined operation of each chiller; the combination strategy includes an action mask corresponding to each power supply side equipment. When the load rate of the target power supply device is less than the minimum safe load rate, or when the load rate is greater than the maximum rated load rate, the action mask corresponding to the target power supply device is a preset value, which is used to indicate that the target power supply device is shielded.

10. A control method for an air conditioning system, characterized in that, The control method, applied to any one of claims 1 to 9, comprises: Acquire outdoor environmental information, indoor environmental information, first operating information, second operating information, load aggregation parameters, and infiltration setting parameters; wherein, the first operating information includes the operating information of the energy supply-side equipment, the second operating information includes the operating information of the at least two load-side equipment, and the load aggregation parameters are parameters generated based on the aggregation of the load-side demand of at least two load-side equipment output by the lower-level agent; the load aggregation parameters include total predicted cooling load, number of starved nodes, and maximum water valve opening; the total predicted cooling load is used to characterize the cooling load demand of at least two load-side equipment in the next upper-level cycle; the number of starved nodes is the number of load-side equipment whose water valve opening is greater than a first threshold and whose temperature deviation is continuously greater than a second threshold within a preset time period, and the temperature deviation is... The difference is the temperature deviation between the indoor temperature of the service area where the load-side equipment is located and the set temperature; the maximum water valve opening is the maximum value of the water valve openings of at least two load-side equipments; the permeation setting parameter is a parameter generated based on the energy supply-side decision quantity of the energy supply-side equipment output by the upper-level agent; the permeation setting parameter includes a predicted water supply temperature setpoint and an actual water supply temperature; the predicted water supply temperature setpoint is used to characterize the water supply temperature setpoint of the energy supply-side equipment in the next upper-level cycle; the actual water supply temperature is the temperature value collected by the water temperature sensor in the current upper-level cycle; the load-side demand is used to characterize the cooling capacity adjustment demand of the corresponding load-side equipment; the energy supply-side decision quantity is used to characterize the energy supply capacity and hydraulic condition target of the energy supply-side equipment. According to the upper-level cycle, the outdoor environmental information, the first operating information, and the load aggregation parameters are input to the upper-level intelligent agent to obtain the energy supply side decision quantity of the energy supply side equipment; And according to the lower-level cycle, the indoor environment information, the second operation information and the permeation setting parameters are input to the lower-level intelligent agent corresponding to each load-side device to obtain at least two load-side demands corresponding to at least two load-side devices; wherein, the duration of the upper-level cycle is greater than the duration of the lower-level cycle; The at least two load-side devices are controlled based on the at least two load-side demands, and the energy-side devices are controlled based on the energy-side decision.

Images