CN-121993893-A - Group control method, device, equipment, storage medium and program product for machine room air conditioner

CN121993893ACN 121993893 ACN121993893 ACN 121993893ACN-121993893-A

Abstract

The application discloses a group control method, a device, equipment, a storage medium and a program product for air conditioners in a machine room, which relate to the technical field of environmental regulation and control; the method comprises the steps of determining a current local observation vector of a target air conditioner according to current operation data and current associated environment data, obtaining historical actions of the target air conditioner at the last moment, determining the current local action value of the target air conditioner according to the historical actions and the current local observation vector, determining the global joint value of a machine room according to the local action values of all the air conditioners, and determining an air conditioner group control scheme of the machine room according to the global joint value. By maximizing the global joint value, all air conditioning agents can be guided to adjust to the globally optimal direction, so that the air conditioning group control of the machine room is realized, and the problems of island operation and lack of coordination are avoided.

Inventors

YUAN XIAOYI
WANG DENUAN
ZHANG XU

Assignees

深圳力维智联技术有限公司
南京力维智联技术有限公司

Dates

Publication Date: 20260508
Application Date: 20260402

Claims (10)

1. The group control method for the air conditioners in the machine room is characterized by comprising the following steps: Acquiring current operation data of different air conditioners in a machine room and current associated environment data of associated areas corresponding to the air conditioners; For a target air conditioner, determining a current local observation vector of the target air conditioner according to the current running data and the current associated environment data, wherein the target air conditioner is any air conditioner in the machine room; Acquiring a historical action of the target air conditioner at the last moment, and determining the current local action value of the target air conditioner according to the historical action and the current local observation vector; And determining the global joint value of the machine room according to the local action value of each air conditioner, and determining the air conditioner group control scheme of the machine room according to the global joint value.
2. The group control method of air conditioners in a machine room according to claim 1, wherein the step of determining a global joint value of the machine room according to the local action value of each air conditioner comprises: constructing a joint input vector according to the local action value of each air conditioner; Correcting the combined input vector through a multi-head self-attention module to obtain a corrected output vector, wherein the corrected output vector comprises the corrected value of each air conditioner, and the corrected value is the value after correcting the local action value; And carrying out nonlinear aggregation on the correction value of each air conditioner based on the topology perception hybrid network module to obtain the global joint value of the machine room.
3. The group control method of the air conditioner in the machine room according to claim 2, wherein the multi-head self-attention module comprises three independent linear transformation units, a multi-head self-attention unit, a residual error connection unit and an output unit; the step of correcting the joint input vector through the multi-head self-attention module to obtain a corrected output vector comprises the following steps: respectively inputting the joint input vectors to three independent linear transformation units for linear projection to obtain characteristic representation of the joint input vectors; Inputting the characteristic representation to the multi-head self-attention unit to perform multi-head attention calculation to obtain attention weight; carrying out residual connection on the attention weights through the residual joint unit to obtain fusion characteristics; and carrying out layer normalization on the fusion characteristics through the output unit to obtain a corrected output vector.
4. The group control method of air conditioners in a machine room according to claim 2, wherein the topology aware hybrid network module comprises a graphic neural network state encoder, a hybrid network and a super network corresponding to each layer of the hybrid network; the step of performing nonlinear aggregation on the correction value of each air conditioner based on the topology perception hybrid network module to obtain the global joint value of the machine room comprises the following steps: Obtaining the graph structure data corresponding to the machine room, and extracting features of the graph structure data through the graph neural network state encoder to obtain a global topology embedded vector; Inputting the global topology embedded vector into the super network, and determining the super network weight and the super network bias corresponding to each layer of super network; And carrying out mixed weighting on the correction value based on the super-network weight and the super-network bias through the mixed network to obtain the global joint value of the machine room.
5. The group control method of machine room air conditioners according to claim 1, wherein before the step of obtaining the historical action of the target air conditioner at the last moment and determining the current local action value of the target air conditioner according to the historical action and the current local observation vector, the method further comprises: Acquiring an offline training data set, and performing reinforcement learning based on the offline training data set to obtain a pre-training model; performing simulation training based on the pre-training model to obtain a simulation training model; and carrying out real environment fine adjustment according to the simulation training model to obtain a real environment model, wherein the real environment model at least comprises a multi-head self-attention module and a topology perception hybrid network module.
6. The group control method of air conditioners in a machine room according to claim 5, wherein the state space of reinforcement learning is determined based on operation indexes of the machine room and graph structure data, the action space of each air conditioner is a control instruction set corresponding to the air conditioner, and the reward function is a multi-objective composite reward function.
7. The utility model provides a computer lab air conditioner group control device which characterized in that, computer lab air conditioner group control device includes: The data acquisition module is used for acquiring current operation data of different air conditioners in the machine room and current associated environment data of associated areas corresponding to the air conditioners; The vector construction module is used for determining a current local observation vector of a target air conditioner according to the current operation data and the current associated environment data for the target air conditioner, wherein the target air conditioner is any air conditioner in the machine room; the local value management module is used for acquiring the historical action of the target air conditioner at the last moment and determining the current local action value of the target air conditioner according to the historical action and the current local observation vector; And the global management and control module is used for determining the global joint value of the machine room according to the local action value of each air conditioner and determining the air conditioner group control scheme of the machine room according to the global joint value.
8. A machine room air conditioning group control device, characterized in that the device comprises a memory, a processor and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the machine room air conditioning group control method according to any of claims 1 to 6.
9. A storage medium, characterized in that the storage medium is a computer-readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements the steps of the group control method of room air conditioners as claimed in any one of claims 1 to 6.
10. A computer program product, characterized in that the computer program product comprises a computer program which, when being executed by a processor, implements the steps of the group control method of room air conditioners as claimed in any one of claims 1 to 6.

Description

Group control method, device, equipment, storage medium and program product for machine room air conditioner Technical Field The present application relates to the field of environmental control technologies, and in particular, to a group control method, apparatus, device, storage medium, and program product for a machine room air conditioner. Background With the explosive growth of cloud computing, big data and artificial intelligence, the energy consumption problem of data centers is increasingly serious. Of the total energy consumption, heating, ventilation, and air conditioning (HeatingVentilation and Air Conditioning, HVAC) systems typically account for up to 30% -40% of the total energy consumption, being the largest energy consuming unit in addition to information equipment. How to reduce refrigeration energy consumption (reduce PUE value) through an efficient control strategy and simultaneously ensure safe operation temperature of IT equipment is a difficult problem to be solved in the current technical field. Early and most small and medium-sized machine rooms still employed Proportional-integral-derivative (pro-portonal INTEGRAL DERIVATIVE, PID) based stand-alone control or simple group control logic. Each machine room precise air conditioner (Computer Room Air Conditioning, CRAC) independently adjusts the temperature setting only according to the readings of the temperature sensor of the air return port of the machine room precise air conditioner. This approach is essentially "islanding" and lacks inter-equipment coordination. Disclosure of Invention The application mainly aims to provide a group control method, a device, equipment, a storage medium and a program product for a machine room air conditioner, which aim to solve the technical problems that the existing machine room air conditioner independently adjusts temperature setting and lacks coordination among equipment. In order to achieve the above objective, the present application provides a group control method for air conditioners in a machine room, the group control method for air conditioners in the machine room comprising: Acquiring current operation data of different air conditioners in a machine room and current associated environment data of associated areas corresponding to the air conditioners; For a target air conditioner, determining a current local observation vector of the target air conditioner according to the current running data and the current associated environment data, wherein the target air conditioner is any air conditioner in the machine room; Acquiring a historical action of the target air conditioner at the last moment, and determining the current local action value of the target air conditioner according to the historical action and the current local observation vector; And determining the global joint value of the machine room according to the local action value of each air conditioner, and determining the air conditioner group control scheme of the machine room according to the global joint value. In an embodiment, the step of determining the global joint value of the machine room according to the local action value of each air conditioner includes: constructing a joint input vector according to the local action value of each air conditioner; Correcting the combined input vector through a multi-head self-attention module to obtain a corrected output vector, wherein the corrected output vector comprises the corrected value of each air conditioner, and the corrected value is the value after correcting the local action value; And carrying out nonlinear aggregation on the correction value of each air conditioner based on the topology perception hybrid network module to obtain the global joint value of the machine room. In one embodiment, the multi-head self-attention module comprises three independent linear transformation units, a multi-head self-attention unit, a residual error connection unit and an output unit; the step of correcting the joint input vector through the multi-head self-attention module to obtain a corrected output vector comprises the following steps: respectively inputting the joint input vectors to three independent linear transformation units for linear projection to obtain characteristic representation of the joint input vectors; Inputting the characteristic representation to the multi-head self-attention unit to perform multi-head attention calculation to obtain attention weight; carrying out residual connection on the attention weights through the residual joint unit to obtain fusion characteristics; and carrying out layer normalization on the fusion characteristics through the output unit to obtain a corrected output vector. In one embodiment, the topology aware hybrid network module comprises a graph neural network state encoder, a hybrid network and a super network corresponding to each layer of the hybrid network; the step of performing nonlinear aggregation on the correction value of