A computational collaborative control method and system for a multi-split air conditioning control system

By establishing a two-layer distributed computing architecture in a multi-split air conditioning system and dynamically scheduling tasks, the problems of wasted computing resources in indoor units and computing latency in outdoor units are solved, achieving more efficient utilization of computing resources and improved regulation efficiency of the air conditioning system.

CN123065084APending Publication Date: 2026-11-10GREE ELECTRIC APPLIANCE INC OF ZHUHAI +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611577367.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-13
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

In existing multi-split air conditioning systems, indoor unit computing resources are wasted and outdoor unit computing latency increases, resulting in lag in adjustment and an inability to effectively handle the increased data volume and control complexity.

Method used

By establishing a two-layer distributed computing architecture between the outdoor unit and multiple indoor units, the computing resource information and task execution information of each unit can be obtained in real time, and tasks can be dynamically scheduled and rationally allocated. This allows the indoor units to utilize idle computing power, reduce the load on the outdoor units, and reduce computing latency.

Benefits of technology

Effectively utilize the idle computing power of indoor units, reduce the load on outdoor units, reduce computing latency, improve regulation lag, and enhance the utilization rate and regulation efficiency of computing resources in air conditioning systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123065084A_ABST
    Figure CN123065084A_ABST
Patent Text Reader

Abstract

This invention discloses a computing power collaborative control method and system for a multi-split air conditioning control system. The method includes: acquiring computing power resource information and tasks to be executed for the outdoor unit; determining tasks to be scheduled from the tasks to be executed based on the computing power resource information of the outdoor unit; receiving computing power resource information and task execution information of each indoor unit, as well as the operating parameters of each indoor unit; acquiring the operating parameters of the outdoor unit; and performing computing power scheduling based on the operating parameters of each indoor unit, the operating parameters of the outdoor unit, the computing power resource information of each indoor unit, and the task execution information, allocating the tasks to be scheduled to the corresponding indoor units for execution. By dynamically determining the allocation of tasks to be scheduled based on the operating parameters and computing power resource information of both indoor and outdoor units, this method solves the problems of fixed task division in traditional air conditioning systems, idle computing power in indoor units, and lagging adjustment under high load in outdoor units, enabling dynamic collaboration of computing power resources between indoor and outdoor units.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of air conditioning collaborative control technology, and in particular to a computing power collaborative control method for a multi-split air conditioning control system, a multi-split air conditioning control system, a multi-split air conditioner, and a computer-readable storage medium. Background Technology

[0002] Existing air conditioning system control architectures typically employ a collaborative model between indoor and outdoor units. The indoor unit is responsible for collecting local environmental parameters and directly controlling the actuators, while the outdoor unit is responsible for adjusting global operating parameters such as compressor frequency and electronic expansion valve opening. As air conditioning systems evolve towards multi-split systems and independent temperature control for multiple rooms, the amount of data processed and the complexity of control have increased significantly. Indoor units need to perform lightweight calculations such as load forecasting and local fine-tuning, while outdoor units need to handle computationally intensive tasks such as multi-room heat load fusion, multi-objective optimization through compressor and electronic expansion valve collaboration, and iterative updates of reinforcement learning models.

[0003] Traditional air conditioners often divide the computing tasks between indoor and outdoor units according to fixed preset functions. The outdoor unit's main controller undertakes all the complex optimization calculations, while the indoor unit only performs simple temperature and humidity collection and air supply. This results in the indoor unit's computing power being idle for a long time, wasting computing resources, while the outdoor unit's computing latency increases and its adjustment lags when under high load. Summary of the Invention

[0004] In view of the above problems, embodiments of the present invention are proposed to provide a computing power collaborative control method for a multi-split air conditioning control system, a multi-split air conditioning control system, a multi-split air conditioner, and a computer-readable storage medium to overcome or at least partially solve the above problems.

[0005] To address the aforementioned problems, a first aspect of this invention provides a computational collaborative control method for a multi-split air conditioning control system, applied to an outdoor unit connected to multiple indoor units; the method includes: Obtain the computing power resource information and tasks to be executed of the outdoor unit, and determine the tasks to be scheduled from the tasks to be executed based on the computing power resource information of the outdoor unit; Receive computing resource information and task execution information of each indoor unit, as well as the operating parameters of each indoor unit; the operating parameters of each indoor unit include the temperature deviation and load change rate of the room where each indoor unit is located; Obtain the operating parameters of the outdoor unit; the operating parameters of the outdoor unit include at least one of the following: outdoor ambient temperature, compressor operating frequency, superheat, and condensing / evaporating pressure; Based on the operating parameters of each indoor unit, the operating parameters of the outdoor unit, the computing power resource information and task execution information of each indoor unit, computing power scheduling is performed, and the tasks to be scheduled are assigned to the corresponding indoor units for execution.

[0006] According to a second aspect of the present invention, a computational collaborative control method for a multi-split air conditioning control system is provided, applied to an indoor unit, wherein the indoor unit is connected to an outdoor unit, the method comprising: The system acquires the computing power resource information and task execution information of the indoor unit, as well as the operating parameters of the indoor unit, and reports the computing power resource information, the task execution information, and the operating parameters of each indoor unit to the outdoor unit. The task execution information includes the tasks currently undertaken by the indoor unit and the task queue status. The operating parameters include the temperature deviation and load change rate of the room where the indoor unit is located. Receive the scheduled tasks assigned by the outdoor unit and execute the scheduled tasks.

[0007] According to a third aspect of the present invention, a multi-split air conditioning control system is provided, the system comprising an outdoor unit and multiple indoor units, wherein the outdoor unit is connected to the multiple indoor units; The outdoor unit is used to obtain the computing power resource information and tasks to be executed from the outdoor unit, and to determine the tasks to be scheduled from the tasks to be executed based on the computing power resource information of the outdoor unit. The indoor unit is used to acquire its own computing power resource information and task execution information, as well as the operating parameters of the indoor unit, and to report the computing power resource information, the task execution information, and the operating parameters of each indoor unit to the outdoor unit; the task execution information includes the tasks currently undertaken by the indoor unit and the task queue status; the operating parameters include the temperature deviation and load change rate of the room where the indoor unit is located; The outdoor unit is used to receive computing resource information and task execution information from each indoor unit, as well as the operating parameters of each indoor unit; to acquire the operating parameters of the outdoor unit; the operating parameters of the outdoor unit include at least one of outdoor ambient temperature, compressor operating frequency, superheat, and condensing / evaporating pressure; and to perform computing power scheduling based on the operating parameters of each indoor unit, the operating parameters of the outdoor unit, the computing resource information and task execution information of each indoor unit, and to allocate the tasks to be scheduled to the corresponding indoor units for execution. The indoor unit is used to receive the scheduled tasks assigned by the outdoor unit and execute the scheduled tasks.

[0008] According to a fourth aspect of the present invention, a multi-split air conditioner is provided, comprising: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the air conditioner indoor and outdoor unit computing power coordinated control method as described in any of the preceding embodiments.

[0009] According to a fifth aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, wherein when executed by a processor, the computer program implements the steps of the air conditioner indoor and outdoor unit computing power coordinated control method as described in any of the preceding claims.

[0010] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a computing power collaborative control method and system for a multi-split air conditioning control system, applied to an outdoor unit connected to multiple indoor units. The method includes: acquiring computing power resource information and tasks to be executed from the outdoor unit; determining tasks to be scheduled from the tasks to be executed based on the computing power resource information of the outdoor unit; receiving computing power resource information and task execution information of each indoor unit, as well as operating parameters of each indoor unit; the operating parameters of each indoor unit include the temperature deviation and load change rate of the room where each indoor unit is located; acquiring the operating parameters of the outdoor unit; the operating parameters of the outdoor unit include at least one of outdoor ambient temperature, compressor operating frequency, superheat, and condensing / evaporating pressure; and performing computing power scheduling based on the operating parameters of each indoor unit, the operating parameters of the outdoor unit, the computing power resource information of each indoor unit, and the task execution information, allocating the tasks to be scheduled to the corresponding indoor units for execution. By acquiring the computing power resource information and pending tasks of the outdoor unit, the tasks to be scheduled are determined. The computing power resource information, task execution information, and indoor unit operating parameters of each indoor unit are received, along with the outdoor unit's operating parameters. Combining this information, computing power scheduling is performed, allocating tasks to the corresponding indoor units for execution. This utilizes the idle computing power of the indoor units, avoiding waste of computing resources. Some tasks are offloaded from the outdoor unit to the indoor units, reducing the outdoor unit's load, minimizing computational latency, and improving adjustment lag. Simultaneously, computing power scheduling incorporates the actual operating parameters of the multi-split air conditioner, matching task allocation with room load status and operational complexity. This avoids assigning tasks to busy indoor units or blindly scheduling tasks under complex conditions, thus ensuring that computing power scheduling truly serves the goals of air conditioning energy saving and comfort. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating the steps of a computational power collaborative control method for a multi-split air conditioning control system provided in an embodiment of the present invention. Figure 2 This is a flowchart of the steps of a computing power collaborative control method for a multi-split air conditioning control system provided in an embodiment of the present invention; Figure 3 This is a flowchart illustrating a computing power collaborative control method for a multi-split air conditioning control system provided in an embodiment of the present invention. Figure 4 This is a structural block diagram of a multi-split air conditioning control system provided in an embodiment of the present invention. Detailed Implementation

[0012] To make the above-mentioned objects, features and advantages of the present invention more readily understood, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0013] Traditional air conditioners often divide the computing tasks between indoor and outdoor units according to fixed preset functions. The outdoor unit's main controller undertakes all the complex optimization calculations, while the indoor unit only performs simple temperature and humidity collection and air supply. This results in the indoor unit's computing power being idle for a long time, wasting computing resources, while the outdoor unit's computing latency increases and its adjustment lags when under high load.

[0014] One of the core concepts of this invention is to determine the tasks to be scheduled by obtaining the computing power resource information of the outdoor unit and the tasks to be executed, receiving the computing power resource information, task execution information and indoor unit operating parameters of each indoor unit, and obtaining the outdoor unit operating parameters, and performing computing power scheduling by combining the above information, allocating the tasks to be scheduled to the corresponding indoor units for execution, making use of the idle computing power of the indoor units, avoiding the waste of computing power resources, diverting some tasks from the outdoor unit to the indoor units, reducing the load on the outdoor unit, reducing computing latency, and improving adjustment lag.

[0015] Reference Figure 1 This diagram illustrates a flowchart of a computational collaborative control method for a multi-split air conditioning control system according to an embodiment of the present invention. The method is applied to an outdoor unit connected to multiple indoor units, and specifically includes the following steps: Step 101: Obtain the computing power resource information and tasks to be executed of the outdoor unit; and determine the tasks to be scheduled from the tasks to be executed based on the computing power resource information of the outdoor unit. In this embodiment of the invention, the outdoor unit is connected to multiple indoor units, forming a two-layer distributed computing architecture of "outdoor unit collaborative computing hub + indoor unit edge computing nodes". The outdoor unit no longer undertakes all the computing alone, but instead, based on its own computing power assessment, schedules some tasks to the indoor units for execution.

[0016] Obtaining computing resource information from the outdoor unit refers to the outdoor unit's real-time assessment of its own computing capabilities. Specifically, the outdoor unit monitors its CPU utilization, memory load, and computation latency in real time to obtain information about its currently available computing resources. The outdoor unit must first know "how much computing power it has left" to determine which tasks it can complete independently and which tasks require scheduling indoor units for collaborative execution. If the outdoor unit has sufficient computing power, the tasks can be primarily completed by the outdoor unit itself, with lower scheduling requirements; if the outdoor unit's computing power is strained, more tasks need to be selected from the pending tasks as scheduling tasks to be allocated to the indoor units.

[0017] The tasks to be executed refer to all the tasks that the outdoor unit currently faces and needs to perform. In this embodiment of the invention, the tasks to be executed include, but are not limited to: multi-room total heat load fusion calculation, compressor frequency multi-objective optimization, electronic expansion valve refrigerant flow allocation, real-time evaluation of overall unit energy consumption, computing power load scheduling, and reinforcement learning model iterative update. Among these tasks, there are both computationally intensive tasks (such as compressor frequency multi-objective optimization and reinforcement learning model iterative update) and tasks related to global optimization (such as multi-room total heat load fusion calculation and real-time evaluation of overall unit energy consumption). Originally, these tasks were all performed by the outdoor unit, but in the dual-layer computing power architecture of this embodiment, some of them can be scheduled to be performed by the indoor unit.

[0018] Based on the computing power resource information of the outdoor unit, tasks to be scheduled are determined from the pending tasks. The outdoor unit compares its own computing power resource information with the computational load, deadline, and data dependencies of each pending task, and selects tasks that the outdoor unit alone cannot handle and that require the indoor unit to coordinate with for execution as pending tasks. If the computational load of a pending task exceeds the remaining computing power of the outdoor unit, it indicates that the outdoor unit has difficulty executing the task alone and should be identified as a pending task to be scheduled and considered for allocation to an indoor unit. If the deadline of a pending task is earlier than the estimated completion time of the outdoor unit, it indicates that the outdoor unit may not be able to complete the task alone and should be identified as a pending task to be scheduled and considered for allocation to an indoor unit. If the data dependencies of a pending task cannot be met by the outdoor unit alone, it indicates that the task requires data or collaborative computation from the indoor unit and should be identified as a pending task to be scheduled.

[0019] Tasks to be scheduled are those that are difficult for the outdoor unit to handle alone and require the indoor unit to be scheduled for collaborative execution. In this embodiment, the tasks to be scheduled are not determined by fixed preset functions, but are dynamically determined based on the real-time computing power resource information of the outdoor unit.

[0020] In some embodiments, step 101 may include the following steps: Sub-step S11: Obtain the computational cost, deadline, and data dependencies for each task to be executed; Sub-step S12: Based on the computational load, deadline, data dependency, and computing power resource information of each task to be executed, the tasks whose computational load exceeds the remaining computing power of the outdoor unit, whose deadline is earlier than the estimated completion time of the outdoor unit, and whose data dependency cannot be satisfied by the outdoor unit are identified as the tasks to be scheduled.

[0021] In this embodiment of the invention, the outdoor unit faces a variety of tasks to be executed. These tasks were originally all handled by the outdoor unit. However, under the two-layer distributed computing architecture of outdoor unit collaborative computing power hub plus indoor unit edge computing power nodes, it is necessary to determine which tasks are suitable for the outdoor unit to execute independently and which tasks need to be scheduled for collaborative execution by the indoor unit. To make this determination, the outdoor unit first needs to obtain the task attributes of each task to be executed, namely, the amount of computation, the deadline, and the data dependencies.

[0022] The computational load reflects the computing power required to execute the task. The tasks to be executed include multi-room total heat load fusion calculation, compressor frequency multi-objective optimization, electronic expansion valve refrigerant flow allocation, real-time overall energy consumption assessment, computing power load scheduling, and reinforcement learning model iterative updates. Among these, compressor frequency multi-objective optimization and reinforcement learning model iterative updates are computationally intensive tasks, requiring large-scale matrix operations or iterative optimization, resulting in a significant computational load; while tasks such as real-time overall energy consumption assessment and computing power load scheduling have relatively smaller computational loads.

[0023] Deadlines reflect the time constraints that tasks must be completed. For control loop tasks such as compressor frequency regulation and electronic expansion valve opening control, end-to-end delay requirements are typically in the millisecond range; while for tasks that allow for longer calculation cycles, such as multi-room total heat load fusion calculation and overall unit energy consumption assessment, the cycle can be in the second range.

[0024] Data dependencies reflect the source of the input data required for a task and the destination of the output data. If the input data required for a task mainly comes from the local sensors and the output directly acts on the local actuator, the data dependency of the task is relatively simple. If the task needs to integrate data from two or more indoor units, or the output needs to be broadcast to multiple nodes, the data dependency of the task is more complex and may require the indoor units to provide data or perform collaborative calculations.

[0025] The outdoor unit compares the computational load, deadline, and data dependencies of each pending task with its own computing power information. Tasks that the outdoor unit cannot handle alone and require coordination with indoor units are selected as pending scheduling tasks. If the computational load of a pending task exceeds the outdoor unit's remaining computing power, it indicates that the outdoor unit has difficulty executing the task alone and should be designated as a pending scheduling task, to be assigned to an indoor unit. If the deadline of a pending task is earlier than the outdoor unit's estimated completion time, it indicates that the outdoor unit may not be able to complete the task alone and should be designated as a pending scheduling task, to be assigned to an indoor unit. If the data dependencies of a pending task cannot be met by the outdoor unit alone, it indicates that the task requires data or collaborative computation from the indoor unit and should be designated as a pending scheduling task.

[0026] The judgment criteria used are: "the computational load exceeds the remaining computing power of the outdoor unit, the deadline is earlier than the estimated completion time of the outdoor unit, and the data dependency cannot be satisfied by the outdoor unit." In this implementation, tasks that meet all three conditions can be identified as tasks awaiting scheduling. That is, only those tasks with excessive computational load, insufficient deadline, and data dependency that cannot be satisfied by the outdoor unit will be included in the list of tasks awaiting scheduling. Alternatively, tasks that meet only one of the three conditions can be identified as tasks awaiting scheduling. That is, if the computational load exceeds the limit, the deadline is insufficient, or the data dependency cannot be satisfied by the outdoor unit, it indicates that the outdoor unit has difficulty processing the task alone, and the task should be included in the list of tasks awaiting scheduling. The specific criteria can be adjusted according to the actual situation, and this embodiment of the invention does not limit the specific criteria.

[0027] Tasks awaiting scheduling are not the same as those ultimately assigned to indoor units. They include both tasks ultimately assigned to indoor units and tasks ultimately executed by outdoor units. The specific task assigned depends on the computing power resources of each indoor unit, task execution information, temperature deviations and load change rates in the rooms where each indoor unit is located, and the operating parameters of the outdoor unit, determined through computing power scheduling. The computational load, deadlines, and data dependencies of each task awaiting execution are obtained. These task attributes are compared with the computing power resources of the outdoor unit itself, and tasks that the outdoor unit cannot handle alone are selected as tasks awaiting scheduling. This allows the outdoor unit to proactively select tasks that need scheduling when its computing power is insufficient, rather than blindly undertaking all the computation, thereby reducing the outdoor unit load, decreasing computational latency, and improving adjustment lag.

[0028] Step 102: Receive computing resource information and task execution information of each indoor unit, as well as the operating parameters of each indoor unit; the operating parameters of each indoor unit include the temperature deviation and load change rate of the room where each indoor unit is located; In this embodiment of the invention, before performing computing power scheduling, the outdoor unit must first understand the computing power resource information and task execution information of each indoor unit; otherwise, it cannot determine which indoor unit is suitable to undertake the scheduling task. Simultaneously, the temperature deviation and load change rate of the room where the indoor unit is located reflect the load status of that room, and the outdoor unit also needs to determine based on this whether the indoor unit's local control tasks are heavy and whether it is suitable to undertake further external tasks.

[0029] The outdoor unit receives computing resource information from each indoor unit via a bidirectional computing communication bus. This information includes metrics such as CPU utilization, memory load, and computational latency, reflecting the available computing power of each indoor unit. Based on its own computing resource information and the task attributes of the pending task, the indoor unit determines the local computing tasks it can execute and reports its own computing resource information and task execution information to the outdoor unit. Therefore, the outdoor unit receives this information reported by each indoor unit. After obtaining the computing resource information from each indoor unit, the outdoor unit can determine which indoor units have sufficient remaining computing power and are suitable for handling scheduled tasks, and which indoor units have limited computing power and should not be further burdened.

[0030] The outdoor unit receives the current task execution status reported by each indoor unit. Task execution information includes the tasks currently assigned to the indoor unit and the task queue status, reflecting how many tasks the indoor unit has, whether its task queue is full, and whether it can accept new tasks. Indoor units report their computing resource information and task execution information to the outdoor unit, which receives it. After obtaining the task execution information from each indoor unit, the outdoor unit can determine which indoor units have shorter task queues and still have capacity to handle scheduled tasks, and which indoor units have full task queues and are not suitable for assigning new tasks.

[0031] Computing resource information reflects "how much computing power" an indoor unit has, while task execution information reflects "how much computing power the indoor unit has already used and whether it can still be used." Only by combining these two information can the outdoor unit accurately determine whether a particular indoor unit is suitable for handling scheduling tasks. Looking only at computing resource information might misjudge an indoor unit with ample remaining computing power but a full task queue as a suitable candidate; similarly, looking only at task execution information might misjudge an indoor unit with a short task queue but near-full computing power as a suitable candidate.

[0032] The outdoor unit receives operating parameters from each indoor unit, reporting the conditions of the rooms in which they are located. These parameters include the temperature deviation and load change rate of the room in which the indoor unit is located. Temperature deviation refers to the difference between the actual temperature of the room and the user-set temperature, reflecting the degree of deviation from the room's current comfort level. A larger absolute value of the temperature deviation indicates that the room requires more immediate adjustment, the more urgent the local control task of the indoor unit, and the less suitable it is to undertake external tasks. The load change rate refers to the degree of change in the heating and cooling load of the room in which the indoor unit is located over time, reflecting the stability of the room's load. A larger load change rate indicates more drastic load fluctuations in the room, the heavier the local load prediction and fine-tuning tasks of the indoor unit, and the less suitable it is to undertake external tasks.

[0033] Step 103: Obtain the operating parameters of the outdoor unit; the operating parameters of the outdoor unit include at least one of the following: outdoor ambient temperature, compressor operating frequency, superheat, and condensing / evaporating pressure; In this embodiment of the invention, the outdoor unit collects parameters reflecting its own operating conditions through its own sensors and operating status monitoring. The outdoor unit's operating parameters include at least one of the following: outdoor ambient temperature, compressor operating frequency, superheat, and condensing / evaporating pressure. These parameters reflect the current operating status of the outdoor unit from different perspectives.

[0034] The outdoor ambient temperature is collected by the outdoor unit's ambient temperature sensor, reflecting the outdoor climate conditions. The more extreme the outdoor ambient temperature, whether it is extreme high or low temperature, the greater the load that the air conditioning system needs to handle, the greater the computational load of global optimization tasks, and the higher the demand for computing power scheduling.

[0035] The compressor's operating frequency reflects its current workload. A higher compressor operating frequency indicates that the system is operating under high load, the operating conditions that global optimization tasks need to handle are more complex, and the computational workload of tasks such as multi-objective optimization of the compressor frequency is greater. At the same time, the degree to which the compressor operating frequency deviates from the reference frequency also affects task priority; the greater the deviation, the more urgently the system needs to be adjusted, and the more urgent the relevant global optimization tasks are.

[0036] Superheat reflects the difference between the refrigerant vapor temperature at the evaporator outlet and the saturation temperature, and is an important indicator of the safe operation of the refrigeration system. The further the superheat deviates from the preset range, the more unstable the system operation is, requiring more frequent global optimization and electronic expansion valve regulation, and increasing the computational load and urgency of related tasks.

[0037] Condensing / evaporating pressures reflect the pressure status of the high-pressure and low-pressure sides of the refrigeration system. A rapid increase in condensing pressure or a low evaporating pressure indicates a deterioration in system conditions, necessitating immediate adjustments to compressor frequency and electronic expansion valve opening. The more urgent the related global optimization tasks, the more critical they become.

[0038] The system receives operating parameters from each indoor unit, including temperature deviation and load change rate in the room where each indoor unit is located, reflecting the room load status on the indoor unit side. It also acquires operating parameters from the outdoor unit, including at least one of outdoor ambient temperature, compressor operating frequency, superheat, and condensing / evaporating pressure, reflecting the system operating status on the outdoor unit side. Only by combining these two parameters can the outdoor unit fully determine how to perform computing power scheduling. If only indoor unit parameters are considered, the outdoor unit only knows which rooms have large load fluctuations and which have stable loads, but not the overall complexity of the system's operating conditions, and cannot determine whether the global optimization task requires more computing power. Similarly, if only outdoor unit parameters are considered, the outdoor unit only knows the complexity of the system's operating conditions, but not the load status of each room, and cannot determine which indoor unit is suitable to handle external tasks.

[0039] Step 104: Perform computing power scheduling based on the operating parameters of each indoor unit, the operating parameters of the outdoor unit, the computing power resource information and task execution information of each indoor unit, and allocate the task to be scheduled to the corresponding indoor unit for execution.

[0040] In this embodiment of the invention, the operating parameters of each indoor unit include the temperature deviation and load change rate of the room where each indoor unit is located. The temperature deviation reflects the degree of deviation from the current comfort level of the room; the larger the absolute value of the temperature deviation, the more urgently the room needs adjustment, the more urgent the local control task of the indoor unit, and the less suitable it is to undertake external tasks. The load change rate reflects the stability of the room's load; the larger the load change rate, the more drastic the load fluctuation in the room, the heavier the local load prediction and local fine-tuning task of the indoor unit, and the less suitable it is to undertake external tasks. Therefore, the operating parameters of each indoor unit are used to determine which indoor units are located in rooms with stable loads, light local control tasks, and are suitable for undertaking dispatching tasks, and which indoor units are located in rooms with drastic load fluctuations, heavy local control tasks, and are not suitable for undertaking external tasks.

[0041] The operating parameters of the outdoor unit include at least one of the following: outdoor ambient temperature, compressor operating frequency, superheat, and condensing / evaporating pressure. These parameters reflect the complexity of the system's current operating conditions. Higher compressor operating frequencies, greater condensing / evaporating pressures, greater deviations of superheat from the preset range, and more extreme outdoor ambient temperatures indicate more complex system conditions. This means that global optimization tasks require a greater computational load, necessitating the allocation of some scheduled tasks to the indoor unit to reduce the load on the outdoor unit. Therefore, the outdoor unit's operating parameters are used to determine the computational load level of global optimization tasks among the scheduled tasks, thereby affecting task allocation priority and the determination of task recipients.

[0042] The computing resource information for each indoor unit, including its own CPU utilization, memory load, and computation latency, reflects how much available computing power that indoor unit currently has. This information is used to determine which indoor units have ample remaining computing power and are suitable for handling scheduled tasks, and which have limited computing power and should not be burdened with additional tasks. If an indoor unit has low CPU utilization, low memory load, and short computation latency, it indicates that it has ample remaining computing power and can be used to handle tasks; conversely, if an indoor unit has high CPU utilization, high memory load, and long computation latency, it indicates that its local computing tasks are already heavy and it is not suitable to handle further external tasks.

[0043] The task execution information for each indoor unit includes the tasks currently assigned to that unit and the status of its task queue. This information reflects how many tasks the unit has on hand, whether its task queue is full, and whether it can accept new tasks. The task execution information is used to determine which indoor units have shorter task queues and still have capacity to handle scheduling tasks, and which indoor units have full task queues and are not suitable for assigning new tasks.

[0044] The process of executing computing power scheduling includes determining task allocation priorities, determining task recipients, and allocating the tasks to be scheduled to the corresponding indoor units for execution according to priorities and recipients.

[0045] When determining task allocation priorities, the outdoor unit determines the load fluctuation level of each indoor unit's corresponding room based on the temperature deviation and load change rate of the room where each indoor unit is located. It then determines the computational workload level of globally optimized tasks among the tasks to be scheduled based on the compressor operating frequency, condensing / evaporating pressure, superheat, and outdoor ambient temperature collected by the outdoor unit itself. Finally, based on the load fluctuation level and the computational workload level of the globally optimized tasks, the task allocation priority is determined. Specifically, the temperature deviation impact value is determined based on the temperature deviation of the room corresponding to each task to be assigned; the larger the absolute value of the temperature deviation, the higher the impact value. The remaining time impact value is determined based on the ratio of the remaining execution time of each task to the total allowed execution time; the smaller the ratio, the higher the remaining time impact value. The operating condition impact value is determined based on the degree to which the compressor operating frequency deviates from the reference frequency, the rate of increase of the condensing pressure, and the degree to which the outdoor ambient temperature deviates from the set range; the greater the deviation, the higher the operating condition impact value. The temperature deviation impact value, remaining time impact value, and operating condition impact value are multiplied by their respective weights and summed to obtain the priority score for each task to be assigned. The task allocation priority is then determined according to the priority scores from highest to lowest.

[0046] When determining the task recipient, the outdoor unit selects indoor units with load fluctuation levels lower than a preset level as candidate recipients based on the load fluctuation level of the rooms where each indoor unit is located. It then obtains the hardware capability tags and current remaining computing resources for each candidate recipient. The hardware capability tags include whether a neural network processing unit is integrated, memory size, and instruction set support. The task attributes of the task to be assigned are matched with the hardware capability tags of each candidate recipient to select those with matching hardware capabilities. Among the candidate recipients with matching hardware capabilities, the indoor unit with the largest remaining computing resources and the lowest load fluctuation level in its room is selected as the task recipient.

[0047] When assigning tasks to be scheduled to the corresponding indoor units, the outdoor unit sorts the tasks according to the task assignment priority. Based on the sorted order, the target indoor unit is selected as the task receiving object corresponding to each task, and the assignment operation is performed on each task.

[0048] In some embodiments, step 104 may include the following steps: Sub-step S21: Determine the load fluctuation level of each indoor unit in the corresponding room based on the temperature deviation of the room where each indoor unit is located and the load change rate. Sub-step S22: Determine the computational load level of the task to be scheduled based on the compressor operating frequency, the condensing / evaporating pressure, the superheat and the outdoor ambient temperature; Sub-step S23: Based on the load fluctuation level and the computational load level of the task to be scheduled, determine the task allocation priority of the task to be scheduled and the indoor unit to take over the task to be scheduled. In this embodiment of the invention, combining temperature deviation and load change rate can determine the load fluctuation level of a room. A large temperature deviation and a drastic load change rate indicate a high load fluctuation level in the room, meaning the indoor unit has a heavy local control workload and is unsuitable for handling external tasks. Conversely, a small temperature deviation and a stable load change rate indicate a low load fluctuation level in the room, meaning the indoor unit has a light local control workload and is suitable for handling dispatch tasks. Only indoor units with low load fluctuation levels are suitable candidates to accept tasks assigned by the outdoor unit.

[0049] The compressor operating frequency, condensing / evaporating pressure, superheat, and outdoor ambient temperature are all collected by the outdoor unit itself, reflecting the complexity of the system's current operating conditions. A higher compressor operating frequency indicates that the system is operating under high load, and the operating conditions that global optimization tasks need to handle are more complex. Higher condensing / evaporating pressure indicates higher high-pressure side pressure, and a greater computational load for tasks such as multi-objective compressor frequency optimization. The further the superheat deviates from the preset range, the more unstable the system operation, requiring more frequent execution of global optimization and electronic expansion valve regulation. More extreme outdoor ambient temperatures indicate a greater load that the air conditioning system needs to handle, and a greater computational load for global optimization tasks. Combining these parameters allows us to determine the computational load level of global optimization tasks among the tasks to be scheduled. The more complex the system operating conditions and the higher the computational load level, the greater the difficulty for the outdoor unit to handle these tasks, and the more necessary it is to allocate some of the tasks to the indoor unit. Conversely, the more stable the system operating conditions and the lower the computational load level, the more tasks the outdoor unit can handle independently.

[0050] When determining task allocation priorities, the outdoor unit determines the temperature deviation impact value based on the temperature deviation of the room corresponding to each scheduled task; the larger the absolute value of the temperature deviation, the higher the temperature deviation impact value. The remaining time impact value is determined based on the ratio of the remaining execution time to the total allowed execution time for each scheduled task; the smaller the ratio, the higher the remaining time impact value. The operating condition impact value is determined based on the degree to which the compressor operating frequency deviates from the reference frequency, the rate of increase of the condensing pressure, and the degree to which the outdoor ambient temperature deviates from the set range; the greater the deviation, the higher the operating condition impact value. The temperature deviation impact value, remaining time impact value, and operating condition impact value are multiplied by their respective weights and then summed to obtain the priority score for each scheduled task. The task allocation priority is determined according to the priority scores from highest to lowest. In this way, tasks with large temperature deviations, close deadlines, and complex system operating conditions are prioritized for scheduling.

[0051] When determining the indoor unit to undertake the scheduling task, the outdoor unit filters indoor units with load fluctuation levels lower than a preset level as candidate targets based on the load fluctuation level of the room where each indoor unit is located. It then obtains the hardware capability tags and current remaining computing resources of each candidate target. The hardware capability tags include whether a neural network processing unit is integrated, memory size, and instruction set support. The task attributes of the task to be assigned are matched with the hardware capability tags of each candidate target to filter out candidates with matching hardware capabilities. Among the candidates with matching hardware capabilities, the indoor unit with the largest remaining computing resources and the lowest load fluctuation level in its room is selected as the indoor unit to undertake the scheduling task. This ensures that the task matches the indoor unit's hardware capabilities while avoiding assigning the task to an indoor unit in a room with drastic load fluctuations.

[0052] The outdoor unit sorts the tasks to be scheduled according to their priority. Based on this sorting order, the corresponding indoor unit for each task is designated as the target indoor unit, and an allocation operation is performed on each task. The allocation operation includes cost assessment: estimating the execution time of the task on the outdoor unit; estimating the execution time on the target indoor unit based on its hardware capability tags and current computing resources; obtaining the bus communication latency between the outdoor unit and the target indoor unit; and calculating the total remote execution time, which is the sum of the target indoor unit's execution time and the bus communication latency. If the total remote execution time is less than the outdoor unit's local execution time, a task allocation instruction is sent to the target indoor unit to allocate the task for execution. If the total remote execution time is greater than or equal to the outdoor unit's local execution time, the task is either kept on the outdoor unit for execution or downgraded to a simplified algorithm for execution on the outdoor unit. Based on the system operating conditions, determine the computational load level of the tasks to be scheduled, determine the task allocation priority and the indoor unit to be assigned, and assign the tasks to the corresponding indoor units for execution, so that the outdoor unit does not have to bear all the calculations alone, thereby reducing the load on the outdoor unit, reducing calculation latency, and improving adjustment lag.

[0053] In some embodiments, step S23 may include the following steps: Sub-step S231: Determine whether the task to be scheduled is related to the room where the indoor unit is located or related to the global optimization of the outdoor unit; Sub-step S232: For the tasks to be scheduled related to the room where the indoor unit is located, determine the task allocation priority of the tasks to be scheduled based on the load fluctuation level of the room where the indoor unit is located; the higher the load fluctuation level, the higher the task allocation priority. Sub-step S233: For the scheduled tasks related to the global optimization of the outdoor unit, determine the task allocation priority of the scheduled tasks according to the computational load level of the scheduled tasks; the higher the computational load level, the higher the task allocation priority.

[0054] In this embodiment of the invention, the tasks to be scheduled are those selected from the tasks to be executed, which are difficult for the outdoor unit to handle alone and require the coordinated execution of the indoor units. These tasks differ in nature and their association with the air conditioning system also differs. One type of task is directly related to the room where the indoor unit is located, such as room temperature and humidity preprocessing, room cooling and heating load prediction, indoor air supply ratio integral and differential local fine-tuning, and local anomaly sensor verification. The service object of this type of task is a specific room, and its execution effect directly affects the temperature control quality of that room. The urgency of the task is mainly determined by the load status of that room. Another type of task is related to the global optimization of the outdoor unit, such as multi-room total heat load fusion calculation, compressor frequency multi-objective optimization, electronic expansion valve refrigerant flow distribution, real-time evaluation of overall unit energy consumption, and reinforcement learning model iterative update. The service object of this type of task is the entire multi-split system, and its execution effect affects the overall energy efficiency and operational safety of the system. The urgency of the task is mainly determined by the complexity of the system operating conditions.

[0055] The two types of tasks are different in nature, and the criteria for determining task allocation priorities should also be different. If the same set of criteria is used to determine priorities, tasks related to rooms may be underestimated due to stable system conditions, while tasks related to global optimization may be overestimated due to large temperature deviations in a particular room.

[0056] The load fluctuation level is determined based on the temperature deviation and load change rate of each indoor unit's room. A large temperature deviation and a drastic load change rate indicate a high load fluctuation level for that room, meaning the room requires immediate adjustment and the related scheduling tasks are more urgent. For example, if a room has a temperature deviation of three degrees Celsius and a drastic load change rate, its temperature control needs are urgent, and the priority of its heating / cooling load prediction and local fine-tuning tasks should be increased to expedite calculations, generate execution instructions, and improve temperature control. Conversely, if a room has a small temperature deviation and a stable load change rate, its load fluctuation level is low, and the priority of its related tasks can be appropriately reduced, allocating computing power to more urgent tasks. Using the room load fluctuation level as a basis for task priority ensures that tasks related to that room prioritize those with high load fluctuations, preventing rooms with high load fluctuations from being left unadjusted due to low task priority.

[0057] The computational workload level of the tasks to be scheduled is determined based on the compressor operating frequency, condensing / evaporating pressure, superheat, and outdoor ambient temperature. Higher compressor operating frequencies, greater condensing / evaporating pressures, greater deviations from the preset superheat range, and more extreme outdoor ambient temperatures indicate more complex system conditions. Global optimization tasks require a larger computational workload, resulting in a higher computational workload level. A higher computational workload level also means a greater difficulty for the outdoor unit to handle these tasks, necessitating faster scheduling to prevent task backlog leading to increased computational delays and lag in adjustments. For example, when the compressor operating frequency is close to its upper limit and the condensing pressure rises too quickly, the computational workload level of the compressor frequency multi-objective optimization task is high, and its priority should be increased to expedite optimization, generate globally optimal control baseline parameters, and distribute them to each indoor unit for execution, preventing the system from deviating from safety boundaries. Conversely, when system conditions are stable, the computational workload level of global optimization tasks is low, and their priority can be appropriately reduced to allocate computing power to more urgent tasks. By using the computational load level of the scheduled tasks as the basis for task priority, tasks related to global optimization are given priority to serve scenarios with complex system conditions. This avoids global optimization tasks not being executed in a timely manner due to low priority when the system conditions are complex, thus ensuring system operation safety and energy efficiency.

[0058] First, distinguish whether the tasks to be scheduled are related to rooms or to global optimization, and then use different criteria to determine their priorities to better reflect the nature of the tasks. Tasks related to rooms are prioritized based on room load fluctuation levels, while tasks related to global optimization are prioritized based on computational load levels, avoiding the use of the same standard to measure tasks of different natures.

[0059] In some embodiments, step S23 may include the following steps: Sub-step S234: Determine the indoor unit screening level based on the computational load level of the task to be scheduled; the higher the computational load level of the task to be scheduled, the higher the indoor unit screening level. Sub-step S235: Based on the load fluctuation level of the room where each indoor unit is located, select indoor units whose load fluctuation level of the room is lower than the indoor unit selection level as candidate indoor units. Sub-step S236: Obtain the hardware capability tags and current remaining computing resources for each candidate indoor unit; Sub-step S237: Match the computational load, deadline, and data dependencies of the task to be scheduled with the hardware capability tags of each candidate indoor unit, and select the indoor unit with matching hardware capabilities. Sub-step S238: Among the indoor units with matching hardware capabilities, select the indoor unit with the largest remaining computing power resources and the lowest load fluctuation level in the room as the indoor unit to undertake the task to be scheduled.

[0060] In this embodiment of the invention, the computational load level of the task to be scheduled is determined based on the compressor operating frequency, condensing / evaporating pressure, superheat, and outdoor ambient temperature, reflecting the complexity of the current system operating conditions. The more complex the system operating conditions, the higher the computational load level of the task to be scheduled, indicating that the task is more demanding and requires a more powerful indoor unit to handle it. Therefore, the selection threshold for indoor units should also be raised accordingly.

[0061] First, candidate indoor units are screened based on load fluctuation levels, including those in rooms with load fluctuation levels lower than the initial screening level. A low load fluctuation level indicates that the room where the indoor unit is located has a stable load and light local control tasks, allowing it to handle external tasks. A high load fluctuation level indicates that the indoor unit already has heavy local control tasks and is not suitable for handling external tasks. Next, the hardware capability tags and remaining computing resources of each candidate indoor unit are obtained. Hardware capability tags include whether a neural network processing unit is integrated, memory size, and instruction set support. Remaining computing resources reflect how much usable computing power the indoor unit has.

[0062] The computational load, deadline, and data dependencies of the tasks to be scheduled are matched with the hardware capability tags of each candidate indoor unit. Indoor units with matching hardware capabilities are selected to avoid assigning tasks with inference attributes to older indoor units without integrated neural network processing units, or assigning tasks with high memory requirements to indoor units with insufficient memory. Among the indoor units with matching hardware capabilities, the one with the largest remaining computing resources and the lowest load fluctuation level in the room is selected as the receiving indoor unit, balancing the conditions of sufficient computing power and light local control tasks.

[0063] The indoor unit screening level is dynamically determined based on the computational load level of the task to be scheduled. The higher the computational load level, the higher the screening level, so as to avoid assigning heavy tasks to indoor units with insufficient computing power, and also to avoid assigning simple tasks to indoor units with excessive computing power, thus avoiding waste.

[0064] Sub-step S24: Based on the task allocation priority and the indoor unit that will receive the task to be scheduled, the task to be scheduled is allocated to the corresponding indoor unit for execution.

[0065] In some embodiments, step S24 may include the following steps: Sub-step S241: Sort each task to be scheduled according to the task allocation priority of the task to be scheduled; Sub-step S242: Estimate the execution time of the task to be scheduled on the outdoor unit. Sub-step S243: Based on the hardware capability tag of the indoor unit that undertakes the task to be scheduled and the current remaining computing power resources, estimate the execution time of the task to be scheduled on the indoor unit that undertakes the task to be scheduled. Sub-step S244: Obtain the bus communication delay between the outdoor unit and the indoor unit that undertakes the task to be scheduled; Sub-step S245: Determine the total execution time based on the execution time and the bus communication delay; Sub-step S246: When the total execution time is less than the local execution time of the outdoor unit, the tasks to be scheduled in the sorting results are sequentially assigned to the corresponding indoor units for execution.

[0066] Hardware capability tags are used to describe the hardware configuration and computing capabilities of indoor units. When selecting indoor units to handle a task, the outdoor unit needs to know the hardware capabilities of each indoor unit to determine whether a particular indoor unit is suitable for a given task. Hardware capability tags serve as the basis for this determination. These tags include whether a neural network processing unit is integrated, memory size, and instruction set support.

[0067] Whether a neural network processing unit is integrated refers to whether the indoor unit has a dedicated hardware unit for accelerating neural network inference. Indoor units with this unit perform inference tasks such as load forecasting faster; indoor units without this unit perform similar tasks slower. Therefore, scheduling tasks with inference attributes should be preferentially matched to indoor units with the neural network processing unit acceleration tag.

[0068] Memory size refers to the amount of memory that the indoor unit can use to run tasks. Indoor units with large memory are suitable for tasks with large amounts of data and the need to cache intermediate results; indoor units with small memory may fail to execute such tasks due to insufficient memory, or experience frequent page swapping and slow execution.

[0069] Instruction set support refers to the types of instruction sets supported by the indoor unit's CPU. Different instruction sets have varying acceleration capabilities for specific operations. If the task to be scheduled involves a specific type of operation, and the target indoor unit's instruction set does not support the corresponding accelerated instructions, the execution efficiency will decrease.

[0070] In this embodiment of the invention, the tasks to be scheduled are sorted according to their task allocation priority, so that tasks with higher priority enter the allocation process first, ensuring that urgent tasks are scheduled first. The execution time of the task to be scheduled on the outdoor unit is estimated; if the task is to be executed on the outdoor unit, how long would it take? Based on the hardware capability tags and current remaining computing resources of the indoor unit receiving the task, the execution time of the task on the receiving indoor unit is estimated. This time is combined with the indoor unit's hardware capability profile; for example, older indoor units without integrated neural network processing units are given a higher time penalty coefficient when performing inference tasks.

[0071] The bus communication latency between the outdoor unit and the receiving indoor unit is obtained. Since tasks are offloaded to the indoor unit for execution, both task data and execution results must be transmitted via the bidirectional computing communication bus; this communication latency is part of the total remote execution time. The total execution time is determined based on the indoor unit's execution time and the bus communication latency; that is, the total remote execution time is the sum of the indoor unit's execution time and the bus communication latency. When the total execution time is less than the outdoor unit's local execution time, the tasks in the sorting results are sequentially assigned to the corresponding indoor units for execution. If the total execution time is greater than or equal to the outdoor unit's local execution time, the task is retained for execution on the outdoor unit, or downgraded to a simplified algorithm for local execution. Tasks are sorted by allocation priority, with higher-priority tasks entering the allocation process first to avoid urgent tasks being delayed due to queuing. Tasks are only allocated when the total remote execution time is less than the local execution time to avoid assigning tasks to indoor units with incompatible hardware capabilities or excessive communication latency, preventing an increase in total execution time due to offloading.

[0072] In some embodiments, the method further includes the following steps: Step 11: Receive a computing power support request sent by the indoor unit. The computing power support request carries the task deadline and the temperature deviation of the room where the indoor unit is located. Step 12: Determine whether the outdoor unit meets the preset computing resource allocation conditions; Step 13: If the outdoor unit does not meet the computing power resource allocation conditions, it replies with a rejection message to the indoor unit. Step 14: If the outdoor unit meets the computing power resource allocation conditions, determine whether the idle computing power of the outdoor unit can handle the pending tasks of the indoor units corresponding to all computing power support requests. Step 15: If the outdoor unit's idle computing power can handle all the pending tasks of the indoor units corresponding to the computing power support requests, then handle all the pending tasks of the indoor units, add the pending tasks to the task queue of the outdoor unit, and send support confirmation information to the indoor units.

[0073] In this embodiment of the invention, when the indoor unit lacks sufficient local computing power, it proactively initiates a computing power support request to the outdoor unit. The outdoor unit determines whether it can help and to what extent, and then decides whether to accept all, partially, or refuse the request. The outdoor unit first receives the request, then determines whether it meets the allocation conditions. If not, it directly refuses; if it does, it further determines whether its idle computing power can handle all the pending tasks corresponding to the requests. If it can handle all, it accepts all and replies with a support confirmation; if it cannot handle all, it enters the subsequent arbitration process.

[0074] The system receives computing power support requests from indoor units. These requests include the task deadline and the temperature deviation of the room where the indoor unit is located. When an indoor unit lacks sufficient local computing power or its local prediction latency exceeds the limit, it will send a computing power support request to the outdoor unit, requesting the outdoor unit to help complete tasks that it cannot handle locally. The request includes the task deadline, which the outdoor unit uses to determine if it can complete the task in time; it also includes the temperature deviation of the room where the indoor unit is located, which the outdoor unit uses to determine which room is more urgent when arbitrating multiple requests in the future.

[0075] Determine whether the outdoor unit meets the preset computing resource allocation conditions. These conditions include: the outdoor unit's current CPU load is below a preset load threshold; the estimated task completion time is less than the task deadline; the most recent heartbeat round-trip latency is less than a preset round-trip latency threshold; the remaining task time is greater than the sum of the estimated outdoor unit processing time and twice the round-trip latency; and the energy consumption increment of the outdoor unit's auxiliary computing is less than the energy consumption of the task executed locally on the indoor unit. These conditions assess the outdoor unit's suitability for handling the task from five perspectives: outdoor unit load, time feasibility, communication quality, time margin, and energy efficiency. If any condition is not met, the outdoor unit is currently unsuitable for handling the task. Forcing it to do so may further increase its own load, prevent the task from being completed before the deadline, cause communication overhead to offset computing power gains, or increase the total system energy consumption.

[0076] If the outdoor unit does not meet the computing power resource allocation conditions, it will send a rejection message to the indoor unit. After receiving the rejection message, the indoor unit will automatically downgrade to a simplified algorithm or extend the calculation cycle. For example, it may switch the load forecasting model to a linear regression approximation model or extend the local calculation cycle to maintain basic control functions even when local computing power is insufficient.

[0077] If the outdoor unit meets the computing power resource allocation conditions, it is further determined whether the outdoor unit's idle computing power can handle the pending tasks of all indoor units corresponding to the computing power support requests. This is a total quantity judgment for the scenario of "multiple indoor units initiating requests at the same time". Even if the outdoor unit itself meets the allocation conditions, if too many requests come in at the same time, the idle computing power may not be enough to handle them all. Therefore, it is necessary to first determine whether the total amount of idle computing power is sufficient.

[0078] If the outdoor unit's idle computing power is sufficient to handle all pending tasks from the corresponding indoor units, then it will accept all pending tasks from the indoor units, add them to the outdoor unit's task queue, and send a support confirmation message to the indoor units. Upon receiving the support confirmation message, the indoor unit will transfer its local pending tasks to the outdoor unit for execution, avoiding recalculation locally. The suitability of the outdoor unit for handling tasks is determined based on computing power allocation conditions; if these conditions are not met, the task will be rejected directly to prevent the outdoor unit from being forced to accept tasks, which could further increase its own load and prevent tasks from being completed before the deadline, thus ensuring the stable operation of the outdoor unit. When the outdoor unit meets the allocation conditions and has sufficient idle computing power, all tasks will be accepted, preventing some indoor units from being forced to degrade due to lack of support and ensuring the temperature control effect in each room.

[0079] In some embodiments, the method further includes: If the idle computing power of the outdoor unit is insufficient to handle all pending tasks corresponding to computing power support requests, then based on the task deadlines of the computing power support requests of each indoor unit, the pending tasks of the indoor unit corresponding to the computing power support request with the closest task deadline will be handled first. If multiple indoor units have the same deadline for their computing power support requests, the absolute value of the temperature deviation of the room where the corresponding indoor unit is located is compared, and the task to be executed for the indoor unit corresponding to the computing power support request with the larger absolute value of the temperature deviation is given priority. If the absolute value of the temperature deviation is still the same, the computing power mutual assistance contribution of the indoor unit corresponding to each computing power support request in the past preset time period is obtained, and the pending tasks of the indoor unit corresponding to the computing power support request with the higher computing power mutual assistance contribution are given priority; the computing power mutual assistance contribution is the ratio of the number of times the indoor unit has accepted and successfully completed the outdoor unit's assigned tasks in the past preset time period to the total number of times the indoor unit has accepted the outdoor unit's assigned tasks. For computing power support requests that are not accepted, a message indicating insufficient computing power is returned to the corresponding indoor unit.

[0080] In this embodiment of the invention, the scenario described is where "the outdoor unit meets the computing power resource allocation conditions, but its idle computing power is insufficient to handle all pending tasks corresponding to computing power support requests." In this situation, the outdoor unit cannot handle all tasks, and it is necessary to determine which requests to prioritize and which cannot be handled temporarily. First, tasks are sorted by their deadlines from earliest to latest, with the earliest deadline taking priority. If deadlines are the same, they are then sorted by the absolute value of the temperature deviation in the room where the indoor unit is located, from largest to smallest, with the largest temperature deviation taking priority. If the temperature deviations are still the same, they are then sorted by the contribution of computing power assistance from highest to lowest, with the highest contribution taking priority. Unhandled requests return a insufficient computing power warning to the corresponding indoor unit.

[0081] The task deadline reflects the urgency of the task corresponding to the computing power support request. The closer the deadline, the more urgent the task needs to be completed and the higher its priority should be. The outdoor units sort the tasks in each computing power support request from earliest to latest according to their deadlines, and prioritize the pending tasks corresponding to the requests with the closest deadlines. In this way, urgent tasks receive priority support, avoiding situations where tasks cannot be completed before the deadline due to being ranked lower in the queue.

[0082] If multiple requests have the same deadline, it indicates that these tasks are of similar urgency. In this case, the absolute value of the temperature deviation in the room where the corresponding indoor unit is located is compared. The larger the absolute value of the temperature deviation, the more serious the deviation in comfort level in that room, and the more urgently it needs adjustment. This request should be prioritized. The outdoor unit then prioritizes the pending tasks corresponding to requests with larger absolute values ​​of temperature deviation, ensuring that the room with the worst comfort level receives support first.

[0083] If the absolute values ​​of the temperature deviations remain the same, it indicates that these requests are comparable in terms of urgency and comfort requirements. In this case, the computing power contribution of the indoor unit corresponding to each request over a preset period of time is taken into account. The computing power contribution is defined as the ratio of the number of times the indoor unit has accepted and successfully completed tasks assigned by the outdoor unit over the preset period to the total number of tasks accepted by the indoor unit. This reflects the indoor unit's past willingness and reliability in cooperating with the outdoor unit's computing power scheduling. A higher contribution indicates that the indoor unit has been more willing to undertake tasks assigned by the outdoor unit and has performed better in the past, and should therefore receive priority support. The outdoor unit will prioritize undertaking the pending tasks corresponding to requests with high computing power contributions, reflecting the reciprocity principle in computing power collaboration and encouraging indoor units to actively participate in computing power collaboration.

[0084] For unaccepted computing power support requests, the outdoor unit sends a computing power shortage warning to the corresponding indoor unit. Upon receiving this warning, the indoor unit automatically downgrades to a simplified algorithm or extends the calculation cycle. For example, it switches the load forecasting model to a linear regression approximation model or extends the local calculation cycle to maintain basic control functions even without support. Requests with the closest deadlines are prioritized to prevent urgent tasks from being delayed due to queuing. Requests with large temperature deviations are prioritized to ensure that rooms with significant comfort deviations are adjusted first, guaranteeing effective temperature control in all rooms. Indoor units with high contribution rates receive priority support, encouraging active participation in computing power collaboration and improving the overall system's computing power collaboration efficiency.

[0085] Furthermore, in this embodiment, the outdoor unit determines the tasks to be scheduled from the pending tasks based on its own computing power resource information. In addition to the aforementioned three judgment dimensions—computational load, deadline, and data dependency—the outdoor unit can also introduce a rolling time window to statistically analyze the actual task completion rate, average response latency, and cache hit rate of each indoor unit, and calculate the available computing power score of each indoor unit in real time. The available computing power score is equal to the calibration score multiplied by the health coefficient and then multiplied by the current load margin. Simultaneously, the outdoor unit can establish computational load feature tags for different types of tasks and record the actual execution time of each type of task on each model of indoor unit during actual operation, constructing a task-equipment model-execution time mapping table, enabling accurate estimation of task completion time when making subsequent task assignment decisions. Furthermore, the outdoor unit can also use the future load forecast values ​​output by the indoor unit load forecast model to reverse-calculate the required increase in computing power resources, pre-adjusting task allocation before actual load changes occur, achieving predictive scheduling rather than reactive scheduling.

[0086] Each indoor unit reports computing resource information, task execution information, and operating parameters. In addition to the aforementioned CPU utilization, memory load, computation latency, currently assigned tasks and task queue status, room temperature deviation, and load change rate, the indoor unit can also report its own hardware capability tags, including whether it integrates a neural network processing unit, memory size, and instruction set support. This information is used by the outdoor unit for task and hardware matching verification during subsequent computing power scheduling. Communication between indoor and outdoor units is via an enhanced CAN FD bus. The CAN FD frame header can introduce a 2-bit urgency flag, corresponding to four priority levels from P0 to P3, enabling key control frames to gain priority access to the bus during the bus arbitration phase, linking computing task priority with communication bus priority. The outdoor unit can also construct a link quality prediction model using the round-trip delay sequence of historical heartbeat frames. Before an indoor unit initiates a computing power support request, it can predict the communication quality within a preset timeframe. If the predicted round-trip delay or packet loss rate exceeds a preset threshold, the outdoor unit will proactively relinquish its offload and initiate a local degradation mode, bringing communication status awareness forward to the scheduling decision stage. In addition, the communication interruption, packet loss, and delay jitter handling mechanisms include: the outdoor unit broadcasts a heartbeat packet every 100ms; if the indoor unit does not receive the heartbeat within 500ms, it determines that the communication is interrupted and automatically switches to local backup control; for task offloading frames and baseline transmission frames marked as requiring acknowledgment, the receiver must reply with an ACK within 10ms; if the sender does not receive an ACK within the timeout, it initiates backoff retransmission, with the retransmission interval increasing sequentially, up to a maximum of a preset number of retransmissions; after failure, the link is marked as abnormal and the task offloading of the node is suspended; all frames carry a transmission timestamp, the receiver calculates the one-way delay, and if the delay jitter exceeds a preset threshold for a consecutive preset number of frames, the outdoor unit dynamically reduces the task offloading frequency of the node or switches to a feature upload mode with a lower data volume.

[0087] The outdoor unit performs computing power scheduling based on the operating parameters of each indoor unit, the operating parameters of the outdoor unit, the computing power resource information of each indoor unit, and the task execution information. In addition to determining task allocation priorities, identifying the receiving indoor unit, and allocating tasks after cost assessment, a task priority management method combining four levels of fixed priorities and dynamic weight correction can be introduced. The four levels of fixed priorities, from highest to lowest, are: P0 fault tolerance, P1 real-time control, P2 lightweight computing, and P3 background iteration. P0 includes sensor anomaly verification, communication interruption degradation, and hardware fault protection; P1 includes compressor frequency adjustment, electronic expansion valve opening control, and indoor fan speed fine-tuning; P2 includes indoor unit load prediction, feature extraction, and proportional-integral-derivative local fine-tuning; and P3 includes reinforcement learning model weight updates, computing power load log caching, and strategy optimization. Based on the fixed levels, the outdoor unit calculates a dynamic priority score for each task. The dynamic score equals the sum of the base priority score, urgency correction, and deadline urgency. The urgency correction is determined based on the temperature deviation of the room corresponding to the task, and the deadline urgency is determined based on the ratio of the remaining execution time of the task to the total allowed execution time. Tasks of the same priority employ a hybrid strategy of priority queues and first-come, first-served (FFS). The queues are dynamically sorted by score; if scores are the same, they are sorted by arrival time. The preemption strategy is as follows: P0 can preempt all tasks from P1 to P3; P1 can preempt tasks from P2 to P3 but not P0; P2 cannot preempt P1 but can preempt P3; P3 is a background task and does not preempt any tasks. If a higher-priority task arrives during its execution, it is immediately suspended and its context is saved. Each task's maximum continuous execution time does not exceed a preset time slice; after timeout, it is forcibly relinquished the CPU and moved to the tail of the ready queue.

[0088] Reference Figure 2 This diagram illustrates a flowchart of another computing-powered collaborative control method for a multi-split air conditioning control system provided by an embodiment of the present invention. The method is applied to an indoor unit connected to an outdoor unit and may include the following steps: Step 201: Obtain the computing power resource information and task execution information of the indoor unit itself, as well as the operating parameters of the indoor unit, and report the computing power resource information, the task execution information, and the operating parameters of each indoor unit to the outdoor unit; the task execution information includes the tasks currently undertaken by the indoor unit and the task queue status; the operating parameters include the temperature deviation and load change rate of the room where the indoor unit is located; Step 202: Receive the task to be scheduled assigned by the outdoor unit and execute the task to be scheduled.

[0089] In this embodiment of the invention, the outdoor unit is responsible for determining the tasks to be scheduled, receiving information, executing computing power scheduling, and allocating tasks to the indoor units; the indoor unit is responsible for reporting its own information, receiving tasks allocated by the outdoor unit, and executing them. Together, they constitute a two-layer distributed computing power architecture of "outdoor unit collaborative computing power hub plus indoor unit edge computing power node". The indoor unit is no longer a pure execution terminal in traditional air conditioners that only performs temperature and humidity collection and air supply execution, but rather an edge computing power node with local computing power. The indoor unit can complete local computing tasks such as local data preprocessing, room cooling and heating load prediction, and local control and adjustment, and can also undertake the tasks to be scheduled allocated by the outdoor unit.

[0090] Obtaining the computing power resource information of the indoor unit refers to the indoor unit's real-time monitoring of its own CPU utilization, memory load, computing latency, and other indicators to obtain information on its currently available computing power resources. The indoor unit must first know "how much computing power it has left" to determine whether it can continue to handle tasks assigned by the outdoor unit. The outdoor unit also needs to use this information to determine whether the indoor unit is suitable for handling dispatched tasks. If the indoor unit has a significant amount of remaining computing power, it indicates that it has the capacity to handle external tasks; if the indoor unit's computing power is close to full capacity, it indicates that it should not be burdened with further load.

[0091] Obtaining task execution information refers to the indoor unit acquiring its currently assigned tasks and the status of its task queue. Task execution information reflects how many tasks the indoor unit has, whether its task queue is full, and whether it can accept new tasks. Computing resource information reflects the indoor unit's available computing power, while task execution information reflects how much computing power the indoor unit has already used and whether it can still use it. Only by combining these two information can the outdoor unit accurately determine whether a particular indoor unit is suitable for undertaking scheduling tasks.

[0092] Obtaining the operating parameters of the indoor unit refers to the indoor unit acquiring the temperature deviation and load change rate of the room it is located in. Temperature deviation is the difference between the actual temperature of the room and the user-set temperature, reflecting the degree of deviation from the room's current comfort level. Load change rate is the degree of change in the room's heating and cooling load over time, reflecting the stability of the room's load. These two parameters reflect the load status of the room where the indoor unit is located, allowing the outdoor unit to determine whether the indoor unit's local control tasks are heavy and whether it is suitable to undertake external tasks. If the temperature deviation and load change rate of a room where an indoor unit is located are large, it indicates that the indoor unit's local control tasks are already heavy and it is not suitable to undertake external tasks; conversely, it indicates that the indoor unit's local control tasks are light and it is suitable to undertake dispatching tasks.

[0093] The indoor and outdoor units are connected via a bidirectional computing communication bus, through which the indoor units report information. This reported information is crucial for the outdoor unit to perform computing power scheduling, determine task offloading targets, and decide whether to accept computing power support requests. Without this information from the indoor units, the outdoor unit lacks input from the indoor units and cannot determine which indoor unit is suitable to undertake the task.

[0094] After the outdoor unit performs computing power scheduling, it allocates some tasks from the pending tasks to the corresponding indoor units for execution. The indoor units receive the task allocation instructions via a bidirectional computing power communication bus. The outdoor unit performs computing power scheduling based on the operating parameters of each indoor unit, the computing power resource information of each indoor unit, and the task execution information. It determines the task allocation priority and the assigned indoor unit, and decides whether to actually allocate tasks based on cost evaluation. The pending tasks received by the indoor units are precisely the tasks that the outdoor unit has determined to allocate after the above-mentioned judgment.

[0095] The indoor unit utilizes its own computing power to complete the task and sends the results back to the outdoor unit. The tasks performed by the indoor unit may include local data preprocessing, room heating and cooling load prediction, and local control adjustment—all local computing tasks—or lightweight iterations of outdoor unit offloading and load pre-calculation. After completion, the indoor unit sends the results back to the outdoor unit. In traditional solutions, indoor units only perform simple temperature and humidity data collection and air supply, leaving their computing power idle for extended periods. Reporting indoor unit computing power and task execution information allows the outdoor unit to identify which indoor units have idle computing power and are suitable for receiving and scheduling tasks. The outdoor unit then receives and executes the tasks assigned by the outdoor unit, ensuring that idle indoor unit computing power is actually utilized and avoiding waste of computing resources.

[0096] In the local calculations on the indoor unit side, the indoor unit predicts the room's heating and cooling load for a preset period based on a locally deployed load prediction model. The load prediction model is a lightweight LSTM model. The input is a fixed-length time window sequence. Each time step includes 8-dimensional features: return air temperature, return air humidity, temperature setpoint, temperature deviation, human presence density, door and window opening / closing status, supply air velocity, and time code. The sampling period is 30 seconds, the historical window length is 10 time steps, the model structure is a single-layer LSTM with a fully connected output layer, the output dimension is 1, the number of parameters is less than 1000, the size is less than 10KB, and it is permanently stored in the indoor unit's Flash reserved partition. It runs as an independent task and does not perform cross-device model fragmentation. To accommodate the individual differences between rooms, static attributes such as room orientation, area, and building envelope can be encoded as low-dimensional vectors and concatenated with real-time sensor sequences before being input into an LSTM. Based on baseline data for the room, the statistical mean and variance of return air temperature and humidity are calculated and stored locally. During inference, the input features are standardized locally. For room thermal characteristic drift that occurs over long-term use, a low-rank adaptation matrix or bias increment is received from the outdoor unit, and the adaptation layer is slightly updated using local data, while the original weights are frozen. The outdoor unit collects compressed features and prediction error feedback uploaded by each indoor unit, performs global model iteration updates, generates a new model version, maintains a model version registry, and distributes the new version to each indoor unit via a bidirectional computing bus in the form of differential upgrade packages. Indoor units perform Flash partition switching during idle periods to update the model. The outdoor unit records the model version number for each indoor unit to ensure global traceability, distributing only the weight difference portion to reduce bus bandwidth usage. If the inference error of the new model exceeds a threshold, the indoor unit automatically reverts to the previous stable version and reports it.

[0097] In the local adaptive fine-tuning of the indoor unit, the indoor unit determines the local real-time load deviation based on the deviation between the indoor environmental parameters in the real-time collected multi-dimensional sensing data and the user-set values, and calculates the absolute value of the error and the error change rate. When the absolute value of the error is less than or equal to the first preset threshold, proportional-integral-derivative control is used; when the absolute value of the error is greater than the second preset threshold or the error change rate is greater than the third preset threshold, fuzzy rules are used for rapid coarse adjustment, where the second preset threshold is greater than the first preset threshold.

[0098] The fuzzy rules take temperature error and error change rate as input and control increment as output. Fuzzy subsets are divided into five levels: NB, NS, ZO, PS, and PB. Membership functions use triangular or trapezoidal functions. The rule base uses 25 IF-THEN rules (5×5). Defuzzification uses the centroid method. Taking temperature error as an example: if the actual temperature is much lower than the set temperature, the error is negative (large), corresponding to NB; if the actual temperature is slightly lower than the set temperature, the error is negative (small), corresponding to NS; if the actual temperature is equal to the set temperature, the error is zero, corresponding to ZO; if the actual temperature is slightly higher than the set temperature, the error is positive (small), corresponding to PS; if the actual temperature is much higher than the set temperature, the error is positive (large), corresponding to PB. The error change rate is similarly divided into these five levels. The membership function in fuzzy control describes the degree to which a specific value belongs to a certain fuzzy level. In traditional sets, a value either belongs to a set or does not, with only two possibilities: 0 and 1. However, in fuzzy control, a value can partially belong to a certain fuzzy level, with membership values ​​ranging from 0 to 1. In this embodiment, the membership function uses a triangular or trapezoidal function. Taking the NB level of temperature error as an example, the triangular membership function can be defined as follows: when the error is -3 degrees Celsius, the membership degree is 1, indicating that it completely belongs to NB; when the error is -1.5 degrees Celsius, the membership degree is 0, indicating that it does not belong to NB at all; when the error is between -3 and -1.5 degrees Celsius, the membership degree gradually decreases from 1 to 0, indicating that it partially belongs to NB. The rule base is a set of rules in fuzzy control that describes the correspondence between input and output. There are two variables in the input: temperature error and error change rate. Each variable has five fuzzy levels, so there are a total of 5 x 5 = 25 possible combinations, corresponding to 25 IF-THEN rules. The form of each rule is: if the temperature error is a certain level and the error change rate is a certain level, then the control increment is a certain level. For example: if the temperature error is PB and the error change rate is ZO, then the control increment is PB. This rule means that if the actual temperature is much higher than the set temperature and the temperature change trend is stable, it indicates that the temperature is too high and there is no tendency for it to decrease on its own. Therefore, a significant increase in the control input is needed to accelerate cooling. The twenty-five rules cover various combinations of temperature error and error change rate, enabling the fuzzy controller to output appropriate control input increments based on the current operating conditions.

[0099] In some embodiments, the method further includes the following steps: Step 21: Before executing the task to be scheduled, determine whether the local computing power of the indoor unit is sufficient to meet the computational requirements of the task to be executed by the indoor unit; Step 22: If the local computing power of the indoor unit is insufficient to meet the computational requirements of the task to be executed by the indoor unit, a computing power support request is sent to the outdoor unit. The computing power support request carries the task deadline and the temperature deviation of the room where the indoor unit is located. Step 23: If a support confirmation message is received from the outdoor unit, the task to be executed is handed over to the outdoor unit for execution. Step 24: If a rejection message is received from the outdoor unit, the task to be performed will be downgraded.

[0100] In this embodiment of the invention, although the indoor unit is an edge computing node capable of performing tasks such as local data preprocessing, room heating and cooling load prediction, and local control and adjustment, its computing power is limited. When the outdoor unit assigns a task to be scheduled to the indoor unit, the indoor unit needs to determine whether it can handle the computation before execution. If its local computing power is insufficient to complete the task, the indoor unit should not attempt to do so but should proactively request computing power support from the outdoor unit. Upon receiving the request, the outdoor unit determines whether to accept the task and how much it can handle; the indoor unit then decides whether to transfer the task or degrade it based on the outdoor unit's response.

[0101] Before executing scheduled tasks, it is determined whether the indoor unit's local computing power is sufficient to handle the computational load of the tasks. The indoor unit's tasks consist of two parts: first, local computational tasks determined by the indoor unit based on its own computing resources and task attributes, such as local data preprocessing, room heating / cooling load prediction, and local control adjustments; second, tasks assigned to the outdoor unit. Before executing these tasks, the indoor unit first assesses its local CPU utilization, memory load, computational latency, and other computing resources to determine if it is sufficient to complete the tasks before the deadline.

[0102] If the indoor unit's local computing power is insufficient to handle the computational load of its pending tasks, it sends a computing power support request to the outdoor unit. This request includes the task deadline and the temperature deviation of the room where the indoor unit is located. The deadline is used by the outdoor unit to determine if the task can be completed in time; the temperature deviation is used by the outdoor unit to determine which room is more urgent during multi-request arbitration. If the indoor unit receives a support confirmation message from the outdoor unit, it transfers the pending tasks to the outdoor unit for execution. The support confirmation message is sent by the outdoor unit when it meets the computing power resource allocation conditions and has sufficient idle computing power, indicating that it is willing and able to help. Upon receiving this message, the indoor unit transfers the pending tasks that it cannot compute locally to the outdoor unit, avoiding redundant computation and consolidation of computing power. If the indoor unit receives a rejection message from the outdoor unit, it downgrades the pending tasks. The rejection message is sent by the outdoor unit when it does not meet the computing power resource allocation conditions or has insufficient idle computing power. Upon receiving this message, the indoor unit cannot transfer the tasks but must downgrade them itself, for example, by switching the load forecasting model to a linear regression approximation model or extending the local computation cycle.

[0103] Before executing a task, the system determines whether local computing power is sufficient. If it is insufficient, it proactively requests support to prevent indoor units from overloading complex tasks, which could lead to excessive local load prediction delays and slow control responses, thus ensuring effective room temperature control. In traditional solutions, only the outdoor unit assigns tasks to the indoor unit. However, the indoor unit can also proactively request assistance from the outdoor unit, allowing both idle computing power in the indoor and outdoor units to be utilized, avoiding waste of computing resources.

[0104] Reference Figure 3 The diagram shows a flowchart of a computing power collaborative control method for a multi-split air conditioning control system provided by an embodiment of the present invention. Figure 3 This illustrates the complete process of an embodiment of the present invention, such as... Figure 3 As shown, the left side represents the outdoor unit. The process proceeds sequentially: Based on its own computing resources, it identifies tasks to be scheduled from the pending tasks; it receives computing resources, task execution information, and operating parameters reported by each indoor unit; it acquires its own operating parameters (outdoor ambient temperature, compressor frequency, superheat, condensing / evaporating pressure); and then performs computing power scheduling, allocating tasks to the corresponding indoor units. Computing power scheduling internally includes four sub-steps: determining the load fluctuation level of the room corresponding to each indoor unit; determining the computational load level of the tasks to be scheduled; determining the task allocation priority and assigning indoor units; and allocating tasks to the corresponding indoor units according to priority. This is followed by computing power support request processing, i.e., determining whether the allocation conditions are met; if so, accepting the task or accepting it according to arbitration sorting; otherwise, rejecting it. Finally, the scheduling strategy and model weights are updated based on task execution results, real-time energy consumption, room temperature fluctuations, and scheduling logs.

[0105] On the right side is the indoor unit, which executes the following steps in sequence: It acquires its own computing power resource information, task execution information, and operating parameters and reports them to the outdoor unit; it receives the tasks to be scheduled from the outdoor unit and executes them; before execution, it determines whether its local computing power is sufficient; if not, it requests computing power support from the outdoor unit; if support is confirmed, it transfers the task; if rejected, it downgrades the process; finally, it sends the task execution result back to the outdoor unit. The cross-lane arrows on the left and right sides of the diagram represent the information interaction between the indoor and outdoor units: the indoor unit reports information, the outdoor unit assigns tasks, the indoor unit initiates a computing power support request, and the indoor unit sends back the task execution result, forming a complete process.

[0106] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0107] Reference Figure 4 The diagram shows a structural block diagram of a multi-split air conditioning control system provided by an embodiment of the present invention. The system includes an outdoor unit 301 and multiple indoor units 302, wherein the indoor units 302 are connected to the outdoor unit 301. The outdoor unit 301 is used to obtain the computing power resource information and tasks to be executed of the outdoor unit, and to determine the tasks to be scheduled from the tasks to be executed based on the computing power resource information of the outdoor unit. The indoor unit 302 is used to acquire its own computing power resource information and task execution information, as well as the operating parameters of the indoor unit, and to report the computing power resource information, the task execution information, and the operating parameters of each indoor unit to the outdoor unit; the task execution information includes the tasks currently undertaken by the indoor unit and the task queue status; the operating parameters include the temperature deviation and load change rate of the room where the indoor unit is located; The outdoor unit 301 is used to receive computing resource information and task execution information of each indoor unit, as well as the operating parameters of each indoor unit; to obtain the operating parameters of the outdoor unit; the operating parameters of the outdoor unit include at least one of outdoor ambient temperature, compressor operating frequency, superheat and condensing / evaporating pressure; and to perform computing power scheduling according to the operating parameters of each indoor unit, the operating parameters of the outdoor unit, the computing resource information and task execution information of each indoor unit, and to allocate the task to be scheduled to the corresponding indoor unit for execution. The indoor unit 302 is used to receive the scheduled tasks assigned by the outdoor unit and execute the scheduled tasks.

[0108] In some embodiments, the outdoor unit 301 is used to obtain the computational load, deadline, and data dependency of each task to be executed; based on the computational load, deadline, data dependency, and computing power resource information of each task to be executed, tasks whose computational load exceeds the remaining computing power of the outdoor unit, whose deadline is earlier than the estimated completion time of the outdoor unit, and whose data dependency cannot be satisfied by the outdoor unit are identified as the tasks to be scheduled.

[0109] In some embodiments, the outdoor unit 301 is configured to: determine the load fluctuation level of the room corresponding to each indoor unit based on the temperature deviation of the room where each indoor unit is located and the load change rate; determine the computational workload level of the task to be scheduled based on the compressor operating frequency, the condensing / evaporating pressure, the superheat, and the outdoor ambient temperature; determine the task allocation priority of the task to be scheduled and the indoor unit to receive the task based on the load fluctuation level and the computational workload level of the task to be scheduled; and allocate the task to be scheduled to the corresponding indoor unit for execution based on the task allocation priority and the indoor unit to receive the task.

[0110] In some embodiments, the outdoor unit 301 is used to determine whether the task to be scheduled is related to the room where the indoor unit is located or related to the global optimization of the outdoor unit; for the task to be scheduled related to the room where the indoor unit is located, the task allocation priority of the task to be scheduled is determined according to the load fluctuation level of the room where the indoor unit is located; the higher the load fluctuation level, the higher the task allocation priority; for the task to be scheduled related to the global optimization of the outdoor unit, the task allocation priority of the task to be scheduled is determined according to the computational load level of the task to be scheduled; the higher the computational load level, the higher the task allocation priority.

[0111] In some embodiments, the outdoor unit 301 is used to determine the indoor unit screening level based on the computational load level of the task to be scheduled; the higher the computational load level of the task to be scheduled, the higher the indoor unit screening level; based on the load fluctuation level of the room where each indoor unit is located, indoor units whose load fluctuation level of the room is lower than the indoor unit screening level are selected as candidate indoor units; the hardware capability tags and current remaining computing resources of each candidate indoor unit are obtained; the computational load, deadline, and data dependency of the task to be scheduled are matched with the hardware capability tags of each candidate indoor unit to select indoor units with matching hardware capabilities; among the indoor units with matching hardware capabilities, the indoor unit with the largest remaining computing resources and the lowest load fluctuation level of the room where it is located is selected as the indoor unit to undertake the task to be scheduled.

[0112] In some embodiments, the outdoor unit 301 is configured to: sort the tasks to be scheduled according to their task allocation priority; estimate the execution time of the tasks to be scheduled locally on the outdoor unit; estimate the execution time of the tasks to be scheduled on the indoor unit receiving the tasks based on the hardware capability tags and the current remaining computing resources of the indoor unit receiving the tasks; obtain the bus communication delay between the outdoor unit and the indoor unit receiving the tasks; determine the total execution time based on the execution time and the bus communication delay; and when the total execution time is less than the execution time of the outdoor unit, sequentially allocate the tasks to be scheduled in the sorting result to the corresponding indoor units for execution.

[0113] In some embodiments, the indoor unit 302 is used to determine whether the local computing power of the indoor unit is sufficient for the computational load of the task to be executed before executing the task to be scheduled; if the local computing power of the indoor unit is insufficient for the computational load of the task to be executed, a computing power support request is sent to the outdoor unit, and the computing power support request carries the task deadline and the temperature deviation of the room where the indoor unit is located. The outdoor unit 301 is used to receive computing power support requests sent by the indoor unit; determine whether the outdoor unit meets the preset computing power resource allocation conditions; if the outdoor unit does not meet the computing power resource allocation conditions, it replies with a rejection message to the indoor unit; if the outdoor unit meets the computing power resource allocation conditions, it determines whether the outdoor unit's idle computing power can handle the pending tasks of all indoor units corresponding to the computing power support requests; if the outdoor unit's idle computing power can handle the pending tasks of all indoor units corresponding to the computing power support requests, it handles all the pending tasks of the indoor units, adds the pending tasks to the task queue of the outdoor unit, and sends a support confirmation message to the indoor unit. The indoor unit 302 is configured to, if it receives a support confirmation message from the outdoor unit, transfer the task to be executed to the outdoor unit for execution; if it receives a rejection message from the outdoor unit, it will downgrade the task to be executed.

[0114] In some embodiments, the outdoor unit 301 is configured to: if the outdoor unit's idle computing power is insufficient to handle all pending tasks corresponding to computing power support requests, prioritize handling pending tasks of the indoor unit corresponding to the computing power support request with the earliest task deadline based on the task deadline order of each indoor unit's computing power support requests; if multiple indoor units have the same task deadline for their computing power support requests, compare the absolute value of the temperature deviation of the room where each indoor unit is located, and prioritize handling pending tasks of the indoor unit corresponding to the computing power support request with the larger absolute value of the temperature deviation; if the absolute values ​​of the temperature deviation are still the same, obtain the computing power mutual assistance contribution of each indoor unit corresponding to the computing power support request within a preset time period, and prioritize handling pending tasks of the indoor unit corresponding to the computing power support request with the higher computing power mutual assistance contribution; the computing power mutual assistance contribution is the ratio of the number of times the indoor unit has accepted and successfully completed tasks assigned by the outdoor unit within a preset time period to the total number of times it has accepted tasks assigned by the outdoor unit; for computing power support requests that are not handled, return a computing power insufficiency prompt to the corresponding indoor unit.

[0115] As the system implementation is basically similar to the method implementation, it is described in a relatively simple way. For relevant details, please refer to the description of the method implementation.

[0116] This invention also provides a multi-split air conditioner, including: a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, it implements the various processes of the computing power collaborative control method embodiment of the multi-split air conditioner control system as described in any of the above embodiments, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0117] This invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the computing power collaborative control method embodiment of the multi-split air conditioning control system as described above, and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0118] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0119] Furthermore, it should be noted that the scope of the methods and apparatus in the embodiments of the present invention is not limited to performing functions in the order shown or discussed. It may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0120] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0121] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other modifications under the guidance of the present invention without departing from the spirit and scope of the claims, and all of these modifications are within the protection scope of the present invention.

Claims

1. A computational collaborative control method for a multi-split air conditioning control system, characterized in that, The method is applied to an outdoor unit, which is connected to multiple indoor units; the method includes: Obtain the computing power resource information and tasks to be executed of the outdoor unit, and determine the tasks to be scheduled from the tasks to be executed based on the computing power resource information of the outdoor unit; Receive computing resource information and task execution information of each indoor unit, as well as the operating parameters of each indoor unit; the operating parameters of each indoor unit include the temperature deviation and load change rate of the room where each indoor unit is located; Obtain the operating parameters of the outdoor unit; the operating parameters of the outdoor unit include at least one of the following: outdoor ambient temperature, compressor operating frequency, superheat, and condensing / evaporating pressure; Based on the operating parameters of each indoor unit, the operating parameters of the outdoor unit, the computing power resource information and task execution information of each indoor unit, computing power scheduling is performed, and the tasks to be scheduled are assigned to the corresponding indoor units for execution.

2. The computational power collaborative control method for a multi-split air conditioning control system according to claim 1, characterized in that, The step of determining the tasks to be scheduled from the tasks to be executed based on the computing power resource information of the outdoor unit includes: Obtain the computational cost, deadline, and data dependencies for each task to be executed; Based on the computational load, deadline, data dependencies, and computing power resource information of each task to be executed, tasks whose computational load exceeds the remaining computing power of the outdoor unit, whose deadline is earlier than the estimated completion time of the outdoor unit, and whose data dependencies cannot be satisfied by the outdoor unit are identified as the tasks to be scheduled.

3. The computational collaborative control method for a multi-split air conditioning control system according to claim 1, characterized in that, The step of performing computing power scheduling based on the operating parameters of each indoor unit, the operating parameters of the outdoor unit, the computing power resource information and task execution information of each indoor unit includes: Based on the temperature deviation of the room where each indoor unit is located and the load change rate, determine the load fluctuation level of the room corresponding to each indoor unit; The computational load level of the task to be scheduled is determined based on the compressor operating frequency, the condensing / evaporating pressure, the superheat, and the outdoor ambient temperature. Based on the load fluctuation level and the computational load level of the task to be scheduled, determine the task allocation priority of the task to be scheduled and the indoor unit to undertake the task to be scheduled. Based on the task allocation priority and the indoor unit that will receive the task to be scheduled, the task to be scheduled is assigned to the corresponding indoor unit for execution.

4. The computational collaborative control method for a multi-split air conditioning control system according to claim 3, characterized in that, The step of determining the task allocation priority of the task to be scheduled based on the load fluctuation level and the computational workload level of the task to be scheduled includes: Determine whether the task to be scheduled is related to the room where the indoor unit is located, or related to the global optimization of the outdoor unit; For tasks to be scheduled related to the room where the indoor unit is located, the task allocation priority of the task to be scheduled is determined according to the load fluctuation level of the room where the indoor unit is located; the higher the load fluctuation level, the higher the task allocation priority. For tasks to be scheduled that are related to the global optimization of the outdoor unit, the task allocation priority of the task to be scheduled is determined according to the computational load level of the task to be scheduled; the higher the computational load level, the higher the task allocation priority.

5. The computational collaborative control method for a multi-split air conditioning control system according to claim 3, characterized in that, The step of determining the indoor unit to undertake the task to be scheduled based on the load fluctuation level and the computational workload level of the task to be scheduled includes: The indoor unit screening level is determined based on the computational load level of the task to be scheduled; the higher the computational load level of the task to be scheduled, the higher the indoor unit screening level. Based on the load fluctuation level of the room where each indoor unit is located, indoor units whose load fluctuation level is lower than the indoor unit screening level are selected as candidate indoor units. Obtain the hardware capability tags and current remaining computing resources for each candidate indoor unit; The computational load, deadline, and data dependencies of the tasks to be scheduled are matched with the hardware capability tags of each candidate indoor unit to select indoor units with matching hardware capabilities. Among indoor units with matching hardware capabilities, the indoor unit with the largest remaining computing power resources and the lowest load fluctuation level in the room is selected as the indoor unit to undertake the task to be scheduled.

6. The computational power collaborative control method for a multi-split air conditioning control system according to claim 5, characterized in that, The step of assigning the task to be scheduled to the corresponding indoor unit for execution includes: The tasks to be scheduled are sorted according to their task allocation priority. Estimate the execution time of the task to be scheduled on the outdoor unit; Based on the hardware capability tags of the indoor unit that will undertake the task to be scheduled and the current remaining computing resources, the execution time of the task on the indoor unit that will undertake the task to be scheduled is estimated. Obtain the bus communication delay between the outdoor unit and the indoor unit that undertakes the task to be scheduled; The total execution time is determined based on the execution time and the bus communication delay; When the total execution time is less than the local execution time of the outdoor unit, the tasks to be scheduled in the sorting results are sequentially assigned to the corresponding indoor units for execution.

7. The computational collaborative control method for a multi-split air conditioning control system according to claim 1, characterized in that, The method further includes: Receive computing power support request sent by indoor unit, the computing power support request carrying the task deadline and the temperature deviation of the room where the indoor unit is located; Determine whether the outdoor unit meets the preset computing resource allocation conditions; If the outdoor unit does not meet the computing resource allocation conditions, it replies with a rejection message to the indoor unit; If the outdoor unit meets the computing power resource allocation conditions, then determine whether the idle computing power of the outdoor unit can handle the pending tasks of the indoor units corresponding to all computing power support requests. If the outdoor unit's idle computing power is sufficient to handle all pending tasks of the indoor units corresponding to the computing power support requests, then it will handle all pending tasks of the indoor units, add the pending tasks to the task queue of the outdoor unit, and send support confirmation information to the indoor units.

8. The computational power collaborative control method for a multi-split air conditioning control system according to claim 7, characterized in that, The method further includes: If the idle computing power of the outdoor unit is insufficient to handle all pending tasks corresponding to computing power support requests, then based on the task deadlines of the computing power support requests of each indoor unit, the pending tasks of the indoor unit corresponding to the computing power support request with the closest task deadline will be handled first. If multiple indoor units have the same deadline for their computing power support requests, the absolute value of the temperature deviation of the room where the corresponding indoor unit is located is compared, and the task to be executed for the indoor unit corresponding to the computing power support request with the larger absolute value of the temperature deviation is given priority. If the absolute value of the temperature deviation is still the same, the computing power mutual assistance contribution of the indoor unit corresponding to each computing power support request in the past preset time period is obtained, and the pending tasks of the indoor unit corresponding to the computing power support request with the higher computing power mutual assistance contribution are given priority. The computing power mutual assistance contribution is the ratio of the number of times the indoor unit has accepted and successfully completed the outdoor unit's assigned tasks in the past preset time period to the total number of times the indoor unit has accepted the outdoor unit's assigned tasks. For computing power support requests that are not accepted, a message indicating insufficient computing power is returned to the corresponding indoor unit.

9. A computational collaborative control method for a multi-split air conditioning control system, characterized in that, Applied to an indoor unit, wherein the indoor unit is connected to an outdoor unit, the method includes: The system acquires the computing power resource information and task execution information of the indoor unit, as well as the operating parameters of the indoor unit, and reports the computing power resource information, the task execution information, and the operating parameters of each indoor unit to the outdoor unit. The task execution information includes the tasks currently undertaken by the indoor unit and the task queue status. The operating parameters include the temperature deviation and load change rate of the room where the indoor unit is located. Receive the scheduled tasks assigned by the outdoor unit and execute the scheduled tasks.

10. The computational power collaborative control method for a multi-split air conditioning control system according to claim 9, characterized in that, The method further includes: Before executing the task to be scheduled, determine whether the local computing power of the indoor unit is sufficient to meet the computational requirements of the task to be executed by the indoor unit; If the local computing power of the indoor unit is insufficient to meet the computational requirements of the task to be executed by the indoor unit, a computing power support request is sent to the outdoor unit. The computing power support request carries the task deadline and the temperature deviation of the room where the indoor unit is located. If a support confirmation message is received from the outdoor unit, the task to be performed will be handed over to the outdoor unit for execution. If a rejection message is received from the outdoor unit, the task to be performed will be downgraded.

11. A multi-split air conditioning control system, characterized in that, The system includes an outdoor unit and multiple indoor units, with the outdoor unit connected to the multiple indoor units; The outdoor unit is used to obtain the computing power resource information and tasks to be executed from the outdoor unit, and to determine the tasks to be scheduled from the tasks to be executed based on the computing power resource information of the outdoor unit. The indoor unit is used to acquire its own computing power resource information and task execution information, as well as the operating parameters of the indoor unit, and to report the computing power resource information, the task execution information, and the operating parameters of each indoor unit to the outdoor unit; the task execution information includes the tasks currently undertaken by the indoor unit and the task queue status; the operating parameters include the temperature deviation and load change rate of the room where the indoor unit is located; The outdoor unit is used to receive computing resource information and task execution information from each indoor unit, as well as the operating parameters of each indoor unit; to acquire the operating parameters of the outdoor unit; the operating parameters of the outdoor unit include at least one of outdoor ambient temperature, compressor operating frequency, superheat, and condensing / evaporating pressure; and to perform computing power scheduling based on the operating parameters of each indoor unit, the operating parameters of the outdoor unit, the computing resource information and task execution information of each indoor unit, and to allocate the tasks to be scheduled to the corresponding indoor units for execution. The indoor unit is used to receive the scheduled tasks assigned by the outdoor unit and execute the scheduled tasks.

12. A multi-split air conditioner, characterized in that, include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, it implements the steps of the computational power collaborative control method of the multi-split air conditioning control system as described in any one of claims 1-8, or implements the steps of the computational power collaborative control method of the multi-split air conditioning control system as described in any one of claims 9-10.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the computing power collaborative control method of the multi-split air conditioning control system as described in any one of claims 1-8, or implements the steps of the computing power collaborative control method of the multi-split air conditioning control system as described in any one of claims 9-10.