Air conditioner, control method thereof and computer program product

Abstract

The invention discloses an air conditioner, a control method of the air conditioner and a computer program product. The method comprises the steps that the first temperature and the first humidity of all areas of a data center at the current moment and the first load power of electronic equipment are obtained; the first temperature, the first humidity and the first load power serve as input of a target evaluation network so as to output decision scores when multiple regulation and control operations are conducted on the air conditioner, the training samples are input into the evaluation network and the target network so as to obtain a first decision score and a second decision score, and the first decision score and the second decision score are obtained; when the difference between the first decision score and the second decision score reaches a preset minimum value, updating the evaluation network to obtain a target evaluation network; and regulation and control operation is determined based on the decision score, so that the air conditioner operates based on the regulation and control operation. According to the method, the load change of the electronic equipment can be quickly responded, the refrigerating capacity of the air conditioner is accurately regulated and controlled, and optimization of the overall energy consumption of the data center is achieved.

Description

Technical Field

[0001] The present application relates to the technical field of air conditioners, and in particular to an air conditioner, a control method thereof, and a computer program product. Background Art

[0002] With the continuous expansion of data centers and the rapid development of information technology, the issue of data center energy consumption has become increasingly prominent. Electronic equipment within data centers has stringent operating temperature requirements, requiring continuous cooling from air conditioning systems to ensure stable operation. This makes air conditioning systems one of the main energy consumers in data centers.

[0003] Currently, most data centers use a temperature threshold-based control strategy for air conditioning. Multiple temperature sensors are installed throughout the data center to monitor temperatures in different areas in real time. When the temperature in a particular area exceeds a preset high-temperature threshold, the air conditioning system increases cooling capacity; when the temperature falls below a preset low-temperature threshold, the system reduces or even stops cooling.

[0004] However, this approach is slow to respond. When the load on an electronic device suddenly changes, the temperature sensor detects the change and transmits it to the air conditioning system, which then makes adjustments. This process involves significant delays, potentially exposing the electronic device to high temperatures for a short period of time, impacting its performance and lifespan. Furthermore, it fails to achieve optimal global energy savings. Control based solely on local temperature fluctuations can result in overcooling in some areas and undercooling in others, resulting in energy waste. Furthermore, it lacks intelligence, and control methods based solely on temperature thresholds struggle to adapt to complex and changing operating environments. Summary of the Invention

[0005] The present application provides an air conditioner, a control method thereof, and a computer program product to at least solve the problems of slow response speed, inability to achieve global energy saving, and lack of intelligence in related technologies. It can quickly respond to changes in the load of electronic equipment, accurately regulate the cooling capacity of the air conditioner, optimize the overall energy consumption of the data center, and ensure that the electronic equipment is always within a suitable operating temperature range.

[0006] The present application provides a method for controlling an air conditioner, comprising: Obtaining a first temperature and a first humidity of each area of ​​the data center and a first load power of an electronic device at a current moment; using the first temperature, the first humidity, and the first load power as inputs of a target evaluation network to output decision scores when performing a plurality of control operations on the air conditioner, wherein training samples are respectively input into the evaluation network and the target network to obtain a first decision score and a second decision score, and when the difference between the first decision score and the second decision score reaches a preset minimum value, updating the evaluation network to obtain the target evaluation network; Determine a random number generation function based on a greedy strategy algorithm to generate a random number based on the random number generation function; In a case where the random number is less than a preset probability threshold, randomly generating a target index based on the random number generation function, and selecting a corresponding control operation from the multiple control operations as the control operation based on the target index; When the random number is greater than the preset probability threshold, the control operation corresponding to the highest decision score is used as the control operation.

[0007] The present application also provides a computer program product, including a computer program / instruction, which implements the above-mentioned air conditioner control method when executed by a processor.

[0008] The present application also provides an air conditioner, comprising a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the above-mentioned air conditioner control method is implemented.

[0009] Through this application, the first temperature and first humidity of each area of ​​the data center and the first load power of the electronic device at the current moment are obtained; the first temperature, first humidity, and first load power are used as inputs of a target evaluation network to output decision scores when performing multiple control operations on the air conditioner, wherein training samples are respectively input into the evaluation network and the target network to obtain a first decision score and a second decision score, and when the difference between the first decision score and the second decision score reaches a preset minimum value, the evaluation network is updated to obtain a target evaluation network; a random number generation function is determined based on a greedy strategy algorithm to generate a random number based on the random number generation function; when the random number is less than a preset probability threshold, a target index is randomly generated based on the random number generation function, and a corresponding control operation is selected from multiple control operations based on the target index as the control operation; when the random number is greater than the preset probability threshold, the control operation corresponding to the highest decision score is selected as the control operation. As a result, the method can quickly respond to changes in the load of the electronic device, accurately control the cooling capacity of the air conditioner, optimize the overall energy consumption of the data center, and ensure that the electronic device is always within an appropriate operating temperature range. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0011] Figure 1is a flow chart of a method for controlling an air conditioner according to an embodiment of the present application; Figure 2 is a flow chart of a method for controlling an air conditioner according to a specific example of the present application; Figure 3 Schematic diagram of a block diagram of an air conditioner according to an embodiment of the present application.

[0012] Reference numerals: 200 - air conditioner, 210 - memory, 220 - processor. DETAILED DESCRIPTION

[0013] The following describes in detail embodiments of the present application, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0014] The following describes the air conditioner control method, air conditioner and computer program product proposed in the embodiments of the present application with reference to the accompanying drawings.

[0015] Figure 1 Flowchart of a method for controlling an air conditioner according to an embodiment of the present application.

[0016] like Figure 1 As shown, the control method of the air conditioner according to the embodiment of the present application may include the following steps: S1, obtaining a first temperature and a first humidity of each area of ​​the data center and a first load power of an electronic device at a current moment; S2, using the first temperature, the first humidity, and the first load power as inputs of a target evaluation network to output decision scores when performing multiple control operations on the air conditioner, wherein the training samples are respectively input into the evaluation network and the target network to obtain a first decision score and a second decision score, and when the difference between the first decision score and the second decision score reaches a preset minimum value, updating the evaluation network to obtain a target evaluation network; S3, determining a random number generation function based on a greedy strategy algorithm to generate a random number based on the random number generation function, and proceeding to steps S4 and S5.

[0017] S4, when the random number is less than a preset probability threshold, randomly generate a target index based on a random number generation function, and select a corresponding control operation from multiple control operations as the control operation based on the target index.

[0018] S5. When the random number is greater than a preset probability threshold, the control operation corresponding to the highest decision score is used as the control operation.

[0019] Specifically, first, the first temperature and the first humidity of each area in the data center at the current time and the first load power of the electronic equipment are acquired. That is, the current operation state data of the data center is collected, which will serve as the basis for regulating the air conditioner. For example, temperature sensors and humidity sensors are installed in each key area of the data center, which are responsible for real-time monitoring and recording the environmental temperature and humidity of the area. And power monitoring devices such as electricity meters or other energy consumption monitoring tools are installed on the electronic equipment in the data center to track the energy consumption of each device and the overall load power in real time. The first load power of the electronic equipment represents the total power consumption of all servers, storage devices, network devices and other related hardware. This can be a cumulative measurement value representing the total energy consumption of all electronic equipment in the data center at a certain time. Among them, wired or wireless sensor networks can be used to transmit data to the central monitoring system, and automated data collection systems can be used to collect and store these data periodically. In this way, accurate temperature readings of each area in the data center can be obtained, including server rooms, cooling areas, storage areas, etc., which is crucial for controlling the internal environment of the data center, and real-time load power data of the electronic equipment is obtained, which reflects the current energy consumption and operating state of the equipment.

[0020] In addition, the collected raw data can also be pre-processed to ensure the quality and availability of the data, providing support for subsequent analysis and decision-making. For example, removing outliers or erroneous readings, such as extreme temperature or humidity readings caused by sensor failure, converting temperature, humidity and power data to a unified measure such as Celsius, percentage and kilowatts for easy comparison and analysis, and finally integrating data from different sensors into a unified data structure such as time series data or tables.

[0021] After obtaining the first temperature, the first humidity and the first load power, the first temperature, the first humidity and the first load power can be used as input to the target evaluation network to output a decision score for multiple regulation operations of the air conditioner. That is, these real-time data are input as an input vector into the evaluation network, which converts these data into an internal representation for understanding and processing. The evaluation network can evaluate the expected effect of all possible regulation actions (such as increasing or decreasing the cooling capacity, adjusting the air speed, etc.) according to the input state data, and each action has a corresponding Q value (decision score), which represents the expected return or value of executing the action in the current state.

[0022] For example, in a data center, the following data is collected from sensors: temperature T1 = 21°C, humidity H1 = 45%, and load power P1 = 50 kW. This data is provided as input vectors to the target evaluation network. The target evaluation network is a deep neural network trained to understand the relationship between these inputs and air conditioning control actions. It consists of multiple layers, each of which learns different features of the input data. After receiving the input data, the target evaluation network processes it through its internal multi-layer neural network. The input layer receives raw data, namely the temperature, humidity, and load power values. The hidden layer consists of multiple neurons, each of which learns a specific feature of the input data. Each neuron performs a weighted summation with the input data using weights and biases, and then performs a nonlinear transformation using an activation function (such as ReLU (rectified linear unit) or sigmoid function). The activation function introduces nonlinearity, enabling the network to learn complex patterns. The hidden layer automatically extracts useful features from the input data by learning weights and biases. For example, the network might learn that the combination of temperature and humidity has a significant impact on the Q-value of a particular action. The output layer has a neuron corresponding to each possible action, outputting the predicted Q-value for that action. The activation function of the output layer might be linear, directly outputting the Q-value. Data propagates from the input layer through the hidden layers, with each layer transforming the data and extracting features. After multiple layers of processing, the final output is a Q-value for each action. These Q-values ​​represent the expected reward of performing each action under the current state. Suppose the evaluation network outputs the following Q-values ​​(decision scores) for three possible control actions: Action A1 (increase cooling capacity): Q(A1) = 0.3, Action A2 (decrease cooling capacity): Q(A2) = 0.5, and Action A3 (increase supply air temperature): Q(A3) = 0.4.

[0023] It should be noted that training samples can be input into the evaluation network and target network, respectively, to obtain a first decision score and a second decision score. When the difference between the first and second decision scores reaches a preset minimum, the evaluation network can be updated to obtain a target evaluation network. Specifically, during data center operation, real-time data such as temperature, humidity, and load power are collected and stored in an experience replay pool as training samples. Each sample can include the current state, the executed action, the reward obtained, and the new state. When the number of samples in the experience replay pool reaches a certain threshold, a batch of samples can be randomly selected for training. These samples are then input into the evaluation network, which predicts the Q-value (first decision score) for each possible action based on the current state. The same samples are then input into the target network, which predicts the Q-value (second decision score) for each possible action based on the new state. The difference between the first and second decision scores can be calculated. If the difference reaches a preset minimum (or threshold), the evaluation network's parameters θ (such as the learning rate) are updated to make its predictions closer to the output of the target network. In the Dual Deep Q-Network, the evaluation network is responsible for predicting the Q-value of each action based on the current state and selecting the action to execute. Its parameters are continuously updated during training to minimize prediction error. The target network provides a stable target Q-value (a second decision score) for training the evaluation network. Its parameters are updated less frequently, typically copied from the evaluation network at regular time steps to stabilize the training process. By separating the evaluation and target networks, the Dual Deep Q-Network mitigates the problem of Q-value overestimation and improves learning stability and efficiency. During training, the evaluation network's parameters are continuously updated to minimize the discrepancy between its predictions and those of the target network, thereby improving decision quality.

[0024] Thus, by comparing the outputs of the evaluation network and the target network, the evaluation network gradually learns how to better predict the expected reward of each action. When the evaluation network's predictions are close enough to those of the target network, the evaluation network's parameters are updated to improve its performance. This process enables the evaluation network to more accurately predict the expected rewards of performing different actions in different states, thereby making better decisions. In this way, the evaluation network can learn to select the best actions in different states to maximize the cumulative reward. The introduction of the target network helps to reduce the problem of overestimation of Q values ​​and improve the stability of learning, while the use of the experience replay pool helps to break time correlation and improve learning efficiency.

[0025] After determining multiple decision scores, a control action can be determined based on the decision scores, causing the air conditioner to operate based on the control action. For example, a random number generation function can be determined based on a greedy strategy algorithm. A random number is generated based on the random number generation function, specifically a random number between 0 and 1. This random number is used to determine whether the agent should explore or exploit. The relationship between the random number and a preset probability threshold is determined. If the random number is less than the preset probability threshold, a target index is randomly generated using the random number generation function. Based on the target index, a corresponding control action is selected from multiple control actions as the control action. Specifically, if the generated random number is less than the preset probability threshold ϵ, exploration is performed, and a target index is generated using the random number generation function. Based on the target index, an action is selected from multiple possible control actions as the control action. If the random number is greater than the preset probability threshold, the control action corresponding to the highest decision score is selected as the control action. Specifically, if the generated random number is greater than or equal to the preset probability threshold ϵ, exploitation is performed, and the action with the highest decision score is selected as the control action. In this way, the best known strategy (selecting the action with the highest Q value) can be used in most cases, while retaining a certain probability for exploration (selecting random actions) to discover potentially better strategies. This method helps the evaluation network and the target network to quickly learn and adapt when facing unknown or changing environments, while avoiding falling into local optimal solutions.

[0026] Alternatively, you can sort multiple decision scores from smallest to largest and select the control action corresponding to the decision score that falls in the median of the sorted order. It's important to note that if the environment is relatively stable and the evaluation network is already relatively accurate, you can choose the action with the highest score. This simple and direct approach selects the action with the highest expected reward, is easy to understand and implement, and helps the agent quickly learn strategies that maximize long-term cumulative rewards. If the environment is highly dynamic or the evaluation network is still training, it may be more appropriate to choose the action with the median score. Avoiding overly high or low scores can help reduce extreme or erratic behavior.

[0027] This allows the air conditioner to operate based on the control action. For example, in the above example, assume the evaluation network outputs the following Q-values ​​(decision scores) for three possible control actions: Action A1 (increase cooling capacity): Q(A1) = 0.3, Action A2 (decrease cooling capacity): Q(A2) = 0.5, and Action A3 (increase supply air temperature): Q(A3) = 0.4. The control action with the highest Q-value can be selected, meaning that increasing the supply air temperature can be selected to control the air conditioner to operate according to the action of increasing the supply air temperature. By executing the selected control action, the data center's air conditioning system will adjust its operating state accordingly, such as changing the cooling capacity, air volume, or air temperature, in order to achieve optimal temperature and humidity control while also considering energy efficiency.

[0028] Furthermore, to maintain a stable data center environment and protect electronic equipment from temperature fluctuations, cooling capacity adjustment steps can be set at 1%-5% of the system's rated cooling capacity, depending on the data center's size and needs. For example, for a system with a rated cooling capacity of 1000kW, the adjustment step can be set from 10kW to 50kW. When cooling capacity adjustment is required, the air conditioning control system will increase or decrease cooling output in steps of 10kW to 50kW, depending on the current load demand and ambient conditions, until the required cooling capacity is achieved. The air conditioning control system also monitors the supply air temperature and makes fine adjustments as needed, ensuring that temperature fluctuations do not exceed an accuracy of ±0.5°C. This helps maintain a stable data center environment and prevents temperature fluctuations from adversely affecting electronic equipment. By optimizing cooling capacity and supply air temperature control, the air conditioning system's energy efficiency can be improved, energy consumption can be reduced, and the stability of the data center environment can be maintained, protecting electronic equipment from temperature fluctuations.

[0029] Thus, using a dual-deep Q-network algorithm, an adaptive air conditioner control method is implemented. It can quickly make decisions based on real-time data and respond to environmental changes. Through continuous learning and optimization, the evaluation network gradually improves its prediction accuracy, making data center air conditioning control more efficient and energy-efficient. Furthermore, by periodically updating the evaluation network's parameters to the target network, learning stability is maintained. This approach reduces the need for manual intervention and improves the system's automation and intelligence. It can quickly respond to changes in electronic equipment load, precisely control air conditioner cooling capacity, optimize overall data center energy consumption, and ensure that electronic equipment always maintains an appropriate operating temperature range.

[0030] According to one embodiment of the present application, after the air conditioner is operated based on the regulation operation, the control method of the air conditioner further includes: obtaining a reward value corresponding to the operation of the air conditioner based on the regulation operation, and obtaining the second temperature and second humidity of each area of ​​the data center and the second load power of the electronic device after the air conditioner is operated based on the regulation operation; storing the reward value, the second temperature, the second humidity, and the second load power in a training sample, wherein the training sample is used to train the evaluation network.

[0031] Specifically, based on the action selected by the evaluation network, the air conditioner performs the corresponding control operation, such as adjusting the cooling capacity, air flow rate, or air temperature. After executing the control operation, the data center's environmental conditions change, requiring remeasurement of the temperature, humidity, and electronic equipment load power in each area of ​​the data center. This data reflects the effectiveness of the control operation. A reward value is calculated based on the new conditions and a preset goal (such as maintaining a specific temperature range or minimizing energy consumption). The reward value reflects the effectiveness of the action, with a positive reward indicating that the action contributes to achieving the goal, while a negative reward indicates that the action negatively impacts the goal. The reward value, along with the second temperature, second humidity, and second load power, can then be stored as training samples. The reward value represents a quantitative indicator of the action's effectiveness, the second temperature represents the new temperature in each area of ​​the data center after the control operation, the second humidity represents the new humidity in each area of ​​the data center after the control operation, and the second load power represents the load power of the electronic equipment after the control operation. These training samples are then used to train the evaluation network to optimize its decision-making strategy. The trained evaluation network then serves as the target evaluation network to determine the optimal control operation for the air conditioner.

[0032] By collecting feedback after executing actions, the evaluation network can continuously learn and adjust its strategies, thereby achieving intelligent control of the data center air conditioning system, improving energy efficiency and reliability. This approach enables the agent to make decisions based on real-time data and continuously optimize its behavior as the environment changes.

[0033] According to one embodiment of the present application, the training sample also includes a first temperature, a first humidity, a first load power and the control operation at the current moment, and the training sample is input into the evaluation network and the target network respectively to obtain a first decision score and a second decision score, including: taking the first temperature, the first humidity, the first load power and the control operation at the current moment as inputs to the evaluation network to obtain the first decision score; taking the second temperature, the second humidity, and the second load power as inputs to the evaluation network to output the decision score when multiple control operations are performed on the air conditioner; determining the updated control operation corresponding to the highest decision score; taking the second temperature, the second humidity, the second load power and the updated control operation as inputs to the target network to obtain the second decision score.

[0034] Specifically, the training samples include not only the new state (second temperature, second humidity, second load power) and reward value after the action is executed, but also the state before the action is executed (first temperature, first humidity, first load power) and the current control operation. This information is used to train the evaluation network and target network to optimize the decision-making process. The state before the control operation (first temperature, first humidity, first load power) and the action performed are collected. After the action is executed, the new state (second temperature, second humidity, second load power) and the corresponding reward value are collected. This data is stored as training samples in the experience replay pool.

[0035] The evaluation network takes the first temperature, first humidity, first load power, and the current control operation as input to generate a first decision score (first Q-value). For example, this data is provided as an input vector to the evaluation network. This data reflects the current operating state of the data center and the action the agent plans to perform. The input data enters the evaluation network through its input layer, beginning the forward propagation process. The data then passes through multiple hidden layers, each containing multiple neurons. These neurons perform a weighted summation with the input data using weights and biases, and then undergo a nonlinear transformation using an activation function. The hidden layers learn features of the input data that are crucial for predicting the Q-value. For example, the network may learn that the combination of temperature and humidity has a significant impact on the Q-value of a particular action. After multiple layers of processing, the data reaches the output layer, which typically has a neuron corresponding to each possible action. The output layer generates a Q-value for each action, representing the expected reward of performing each action under the current state. Based on this, an action is selected from the Q-values ​​output by the evaluation network. This action can be randomly selected or the one with the highest Q-value.

[0036] Similarly, the second temperature, second humidity, and second load power are used as inputs to the evaluation network to output decision scores for multiple control operations performed on the air conditioner. The updated control operation with the highest decision score is then determined. These real-time collected parameters serve as inputs to the evaluation network. The evaluation network evaluates the Q value (i.e., decision score) of each possible control operation (e.g., increasing cooling capacity, decreasing air flow), based on the current state. A higher Q value indicates a higher likelihood of achieving a positive effect (e.g., reducing energy consumption, maintaining stable temperature), given the current state. The control operation with the highest Q value is selected from the Q values ​​output by the evaluation network. This operation is considered the most likely to achieve the best effect under the current state and is called the updated control operation. The second temperature, second humidity, second load power, and updated control operation are then used as inputs to the target network to obtain a second decision score. The target network provides a stable benchmark for calculating the target Q value. Based on the input state and the updated control operation, the target network outputs a target Q value (i.e., the second decision score). This target Q value is used to evaluate the actual effectiveness of the updated control operation and serves as a reference for training the evaluation network.

[0037] Thus, using real-time collected data as input, the network output decision scores are evaluated, the optimal control operation is selected, and verification and optimization are performed using the target network. This approach not only enables rapid response to changes in electronic device loads, but also optimizes global energy consumption, improves control accuracy and reliability, and achieves significant energy savings and operational stability.

[0038] According to one embodiment of the present application, the control method of the air conditioner also includes: determining a target loss function based on the first decision score and the second decision score; minimizing the target loss function based on a gradient descent algorithm so that the difference between the first decision score and the second decision score reaches a preset minimum value.

[0039] Specifically, determining and minimizing the target loss function is a key step in the training process. The target loss function can be determined based on the first and second decision scores. Specifically, the first decision score is the Q-value output by the evaluation network based on the current state, representing the evaluation network's expected utility for each possible control operation. The second decision score is the target Q-value output by the target network based on the current state and the updated control operation, representing the target network's estimate of the actual utility of the updated control operation. The target loss function can be determined based on the difference between the Q-value output by the evaluation network and the target Q-value output by the target network.

[0040] After determining the target loss function, the target loss function can be minimized based on the gradient descent algorithm so that the difference between the first decision score and the second decision score reaches a preset minimum value. The gradient descent algorithm is an optimization algorithm used to minimize the loss function L(θ). The basic idea is to gradually reduce the value of the loss function by calculating the gradient of the loss function with respect to the parameter θ and updating the parameter in the opposite direction of the gradient. In each iteration, a batch of samples are randomly drawn from the experience replay pool, and the loss function L(θ) of these samples is calculated. The gradient of the loss function with respect to the parameter θ is calculated, and the parameter θ of the evaluation network is updated according to the gradient descent update rule. During the training process, it is hoped that the difference between the first decision score and the second decision score is as small as possible, that is, the value of the loss function L(θ) is as small as possible. A preset minimum value is usually set. When the value of the loss function is below this threshold, it can be considered that the training has converged and the algorithm has learned a relatively accurate control strategy.

[0041] By minimizing the objective loss function, the evaluation network can learn more accurate Q values, enabling it to make better control decisions in actual operation. This helps improve the energy efficiency and operational stability of data center air conditioning systems, enabling rapid response to changes in electronic equipment loads, and precisely controlling the cooling and air flow rates of the air conditioning system, saving energy and extending equipment life.

[0042] According to one embodiment of the present application, determining a target loss function based on a first decision score and a second decision score includes: taking the product of a discount factor and the second decision score and the sum of the reward values ​​as the target decision score; and determining the target loss function based on the difference between the first decision score and the target decision score.

[0043] Specifically, when determining the target loss function based on the first decision score and the second decision score, the product of the discount factor and the second decision score and the sum of the reward value can be used as the target decision score, and the target loss function can be determined based on the difference between the first decision score and the target decision score. For example, the formula Determine the target loss function. is the target loss function, Is to evaluate the network in state Next action The output of is the first decision score. is the reward value, is the discount factor (a value between 0 and 1 that balances the importance of immediate rewards and future rewards), is the maximum Q value of all possible actions a' of the target network in the next state s'. It should be noted that the state s in the above formula is the first temperature and first humidity of each area in the current data center and the first load power of the electronic equipment. Executing action a is to evaluate the action selected by the network in the current state s, such as increasing the cooling capacity, decreasing the cooling capacity, increasing the air supply volume, decreasing the air supply volume, adjusting the air supply temperature, etc. The state s' represents the adjusted temperature, humidity, and load power of the air conditioner. a' is the action corresponding to the maximum Q value among all possible actions considered by the target network in the new state s', such as increasing the cooling capacity, decreasing the cooling capacity, increasing the air supply volume, decreasing the air supply volume, adjusting the air supply temperature, etc.

[0044] Therefore, by introducing the objective loss function, the algorithm's stability and efficiency are improved, enabling it to better handle high-dimensional state and action spaces, as well as complex environmental dynamics. This can lead to better control strategies and improved performance in practical applications, such as data center air conditioning control.

[0045] According to one embodiment of the present application, obtaining a reward value corresponding to the operation of the air conditioner based on the control operation includes: obtaining a temperature balance index of the data center, a temperature stability index of the electronic equipment, and an energy efficiency index of the air conditioner; and determining the reward value based on the sum of the product of the first weight coefficient and the temperature balance index, the product of the second weight coefficient and the temperature stability index, and the product of the third weight coefficient and the energy efficiency index.

[0046] Specifically, rewards are feedback received by the agent (in this case, the air conditioning control system) from the environment, used to evaluate the effectiveness of its actions. The design of reward values ​​is crucial to the training process and the performance of the resulting policy. When obtaining the reward corresponding to the air conditioner's control actions, the data center's temperature uniformity, the electronic equipment's temperature stability, and the air conditioner's energy efficiency can be obtained. The temperature uniformity index measures the consistency of temperatures across different areas within the data center. High temperature uniformity means minimal temperature variation across the data center, helping to protect electronic equipment from localized overheating. The electronic equipment's temperature stability index assesses the temperature fluctuations in the electronic equipment's operating environment. High temperature stability means minimal temperature variation around the electronic equipment, which helps improve its operating efficiency and lifespan. The air conditioner's energy efficiency index measures the energy efficiency of the air conditioner while providing cooling. High energy efficiency means less energy is consumed to achieve the same cooling effect.

[0047] Because different performance indicators may have varying degrees of impact on the operational efficiency and costs of a data center, each indicator needs to be assigned a weighting factor. Specifically, a first weighting factor is assigned to the temperature balance indicator, reflecting its importance within the overall performance; a second weighting factor is assigned to the temperature stability indicator, reflecting its importance within the overall performance; and a third weighting factor is assigned to the energy efficiency indicator, reflecting its importance within the overall performance. Thus, the reward value can be determined based on the sum of the product of the first weighting factor and the temperature balance indicator, the product of the second weighting factor and the temperature stability indicator, and the product of the third weighting factor and the energy efficiency indicator.

[0048] Therefore, by comprehensively considering temperature balance, temperature stability, and energy efficiency, the reward value can more comprehensively reflect the effectiveness of air conditioning control operations. A well-designed reward value can guide the air conditioning control system to make better decisions, thereby improving the overall performance of the data center. Furthermore, the weight coefficient can be adjusted based on actual conditions to accommodate different operational requirements and optimization goals. In this way, the reward value not only reflects the direct effect of air conditioning control operations, but also considers their impact on the long-term operational efficiency and costs of the data center, thereby helping to train more efficient and reliable air conditioning control strategies.

[0049] According to one embodiment of the present application, obtaining the temperature balance index of the data center includes: calculating the temperature average value based on the actual temperature of each area of ​​the data center; and determining the temperature balance index based on the actual temperature of each area of ​​the data center and its corresponding weight coefficient and temperature average value.

[0050] Specifically, when obtaining the temperature balance index of a data center, the temperature average value can be calculated based on the actual temperature of each area of ​​the data center. That is, first, it is necessary to collect the actual temperature data of each area of ​​the data center, which can be obtained through temperature sensors deployed in different locations of the data center. Then, the average value of these temperature data is calculated. That is, the average temperature is equal to the sum of the actual temperatures of each area divided by the total number of data area centers. The temperature balance index is usually calculated based on the deviation between the actual temperature of each area and the average temperature. In order to more accurately reflect the importance of different areas, a weight coefficient can be introduced. The weight coefficient can be determined based on factors such as the importance of the area, equipment density or heat load. For example, the temperature balance index is calculated by the formula OK. Among them, is the weight coefficient of the ith region, is the actual temperature of the ith region, is the average temperature of all areas in the data center.

[0051] Therefore, by ensuring uniform temperature distribution within the data center, unnecessary energy waste can be avoided. For example, there is no need for overcooling in certain areas. Temperature balance helps prevent damage to electronic equipment due to local overheating, thereby improving the reliability and life of the equipment. By adjusting the weight coefficient, it can flexibly adapt to the specific needs of different areas within the data center.

[0052] According to one embodiment of the present application, obtaining a temperature stability index of an electronic device includes: calculating the maximum and minimum temperature values ​​of the electronic device within a preset time period; determining a temperature fluctuation difference based on the maximum and minimum values; and determining a temperature stability index based on the temperature fluctuation difference and a preset mapping relationship, wherein the preset mapping relationship is used to indicate the relationship between the temperature fluctuation difference and the temperature stability index.

[0053] Specifically, the temperature stability of electronic equipment is a key factor in ensuring its normal operation and extending its service life. Temperature stability indicators can help assess temperature fluctuations in the equipment's operating environment, thus providing a basis for air conditioning control strategies. To obtain the temperature stability indicator for electronic equipment, temperature data from the electronic equipment must first be collected over a preset time period. This data can be obtained using built-in temperature sensors or externally deployed temperature monitoring equipment. The maximum and minimum temperature values ​​within that time period are then calculated. Based on these values, the temperature fluctuation difference is determined. The temperature fluctuation difference is the difference between the maximum and minimum temperature values, reflecting the magnitude of the temperature fluctuation over the preset time period.

[0054] After determining the temperature fluctuation difference, the temperature stability index can be determined based on the temperature fluctuation difference and the preset mapping relationship. That is, the temperature stability index is determined based on the temperature fluctuation difference and the preset mapping relationship. The preset mapping relationship is a function or rule used to convert the temperature fluctuation difference into a temperature stability index. For example, the relationship between the temperature fluctuation difference and the temperature stability index is determined in advance. After the temperature fluctuation difference is determined, the temperature stability index can be obtained by directly calling the preset mapping relationship. For example, this mapping relationship can be designed according to actual needs and equipment characteristics, and the mapping relationship may include linear mapping, nonlinear mapping, etc. For example, a linear mapping relationship can be temperature stability index = 1-α*ΔT / preset threshold. Among them, α is a coefficient used to adjust the sensitivity of the mapping, ΔT is the temperature fluctuation difference, and the preset threshold is an upper limit of temperature fluctuation set according to equipment characteristics and operating requirements.

[0055] Therefore, by monitoring temperature stability, temperature fluctuations that may lead to equipment failure can be discovered and resolved in a timely manner. The temperature stability index can serve as an important tool for evaluating and optimizing data center environmental control, helping to ensure that electronic equipment operates stably under appropriate temperature conditions.

[0056] According to one embodiment of the present application, obtaining an energy efficiency index of an air conditioner includes: determining an energy efficiency ratio based on a ratio of the cooling capacity of the air conditioner to the power consumption of the air conditioner; when the energy efficiency ratio is greater than a preset energy efficiency ratio threshold, calculating a first energy efficiency ratio difference between the energy efficiency ratio and the preset energy efficiency ratio threshold, and determining the energy efficiency index based on the product of the first energy efficiency ratio difference and a positive reward value; when the energy efficiency ratio is less than the preset energy efficiency ratio threshold, calculating a second energy efficiency ratio difference between the energy efficiency ratio and the preset energy efficiency ratio threshold, and determining the energy efficiency index based on the product of the second energy efficiency ratio difference and a negative reward value. The preset energy efficiency ratio threshold can be determined based on actual conditions.

[0057] Specifically, when obtaining an air conditioner's energy efficiency index, the energy efficiency ratio (EER) can be determined based on the ratio of the air conditioner's cooling capacity to its power consumption. Cooling capacity refers to the cooling energy provided by the air conditioner per unit time, and power consumption refers to the electrical energy consumed while providing that cooling capacity. A preset EER threshold is a benchmark used to assess whether the air conditioner's energy efficiency meets the expected standard. The calculation method for the energy efficiency index can be determined based on the comparison of the actual EER with the preset EER threshold.

[0058] When the EER is greater than the preset EER threshold, the difference between the actual EER and the preset EER threshold is calculated (i.e., the first EER difference = EER - the preset EER threshold). The energy efficiency index is determined based on the product of the first EER difference and the positive reward value (i.e., the energy efficiency index = the first EER difference × the positive reward value). In this case, the energy efficiency index is positive, indicating that the air conditioner's energy efficiency exceeds the preset standard and a positive reward should be given. When the EER is less than the preset EER threshold, the difference between the preset EER threshold and the actual EER is calculated (i.e., the second EER difference = the preset EER threshold - EER). The energy efficiency index is determined based on the product of the second EER difference and the negative reward value (i.e., the energy efficiency index = the second EER difference × the negative reward value). In this case, the energy efficiency index is negative, indicating that the air conditioner's energy efficiency is lower than the preset standard and a negative penalty should be given. The preset EER threshold can be 3, the first EER difference is 10, and the second EER difference is -5.

[0059] Therefore, by rewarding high-efficiency operation and penalizing low-efficiency operation, it is possible to incentivize the optimization of air conditioner operation strategies, thereby improving overall energy efficiency. In this way, energy efficiency indicators can serve as an important tool for evaluating and optimizing air conditioner operation efficiency, helping to achieve energy savings and improve operational quality.

[0060] According to one embodiment of the present application, the air conditioner control method further includes: monitoring the status of the air conditioner; and adjusting the training parameters of the evaluation network when an abnormality occurs in the air conditioner, wherein the abnormality of the air conditioner includes at least one of the following: the temperature of at least one area in the data center is greater than a first preset temperature threshold and lasts for a first preset time period, or the temperature of at least one area is less than a second preset temperature threshold and lasts for a second preset time period, the suction pressure and exhaust pressure of the air conditioner compressor are not within a preset pressure range, and the flow rate of chilled water and cooling water are not within a preset flow range; adjusting the training parameters of the evaluation network includes: reducing at least one of the experience replay pool capacity, batch training size, learning rate, and target network update frequency. The first preset time period, the second preset time period, the first preset temperature threshold, the second preset temperature threshold, the preset pressure range, and the preset flow range can be determined according to actual conditions.

[0061] Specifically, while controlling the air conditioner, the air conditioner's status can also be monitored. Sensors and monitoring systems collect real-time operating data from the air conditioner, including but not limited to temperature, humidity, pressure, current, and voltage. This data is analyzed to assess the air conditioner's health and performance. When an air conditioner anomaly is detected, measures are taken to adjust the training parameters of the evaluation network to adapt to the current operating conditions and prevent further deterioration of system performance. For example, if the temperature of at least one area in the data center exceeds a first preset temperature threshold and this condition persists for a first preset duration (e.g., 30 minutes), a high temperature anomaly is considered to have occurred. If the temperature of at least one area falls below a second preset temperature threshold and this condition persists for a second preset duration (e.g., 20 minutes), a low temperature anomaly is considered to have occurred. The air conditioner compressor's suction and discharge pressures are monitored. If these pressures are outside the preset pressure range, a pressure anomaly is considered to have occurred. The chilled water and cooling water flow rates are monitored. If the flow rates are outside the preset flow range, a flow anomaly is considered to have occurred. Therefore, when an air conditioner anomaly is detected, the training parameters of the evaluation network need to be adjusted to adapt to the new operating conditions and optimize the control strategy.

[0062] Training parameters may include reducing or adjusting at least one of the experience replay pool capacity, batch training size, learning rate, and target network update frequency. Adjusting the experience replay pool capacity, i.e., reducing the capacity of the experience replay pool, can reduce the number of samples during training, which helps the network adapt to new operating conditions more quickly. Reducing the batch size, i.e., reducing the size of the batch training, can increase the update frequency during training, helping the network adapt to changes more quickly. Reducing the learning rate, i.e., reducing the learning rate, can make network parameter updates smoother and reduce instability caused by excessive parameter updates. Reducing the update frequency, i.e., reducing the update frequency of the target network, can reduce fluctuations during training and improve learning stability. Furthermore, the extent of the adjustment can also be determined based on the severity of the air conditioner anomaly. For example, a more significant adjustment of the training parameters may be required for severe temperature anomalies.

[0063] Therefore, by adjusting the evaluation network's training parameters when an air conditioner experiences an anomaly, the system's robustness can be improved, enabling it to better cope with operational uncertainties and changes. Timely adjustments can prevent the air conditioner from continuing to operate under abnormal conditions, thereby protecting the equipment from damage. By adjusting the training parameters, the air conditioner's performance can be optimized under different operating conditions, improving energy efficiency and reliability.

[0064] According to one embodiment of the present application, the air conditioner control method further includes: obtaining a difference between a control parameter determined by the evaluation network during a control operation and a control parameter used by the feedback controller to control the air conditioner; and, if the difference exceeds a preset difference threshold and persists for a preset time, controlling the air conditioner to operate based on the control parameter used by the feedback controller to control the air conditioner. The preset difference threshold and the preset time can be determined based on actual conditions.

[0065] Specifically, the difference between the control parameters determined by the evaluation network during the control operation and the control parameters used by the feedback controller to control the air conditioner is obtained. The evaluation network determines a set of control parameters based on the current environmental conditions (such as temperature and humidity) and historical data. These parameters may include cooling capacity, air volume, and air temperature. A feedback controller (such as a PID (Proportional-Integral-Derivative) controller) calculates another set of control parameters based on real-time sensor data and preset control logic. The difference between the two sets of control parameters is calculated. This difference can be an absolute difference or some form of norm (such as the Euclidean distance).

[0066] A difference threshold is set to determine whether the difference between the two sets of control parameters is large enough to require action, and a time threshold is set to determine whether the difference exceeds the threshold for a sufficient period of time. If the difference exceeds the preset difference threshold and this state persists for a preset time, the output of the evaluation network is considered to be unstable or unreliable. If the judgment conditions are met, the control system switches to operating the air conditioner based on the control parameters of the feedback controller. This is a safety mechanism used to prevent potential errors in the evaluation network from adversely affecting the operation of the air conditioner. In addition, during the switching process, a smooth transition is required to avoid impacting the operation of the air conditioner. This can be achieved by gradually adjusting the control parameters.

[0067] Thus, by switching to the feedback controller when the evaluation network's output is unreliable, the system's robustness can be improved. When the evaluation network's output could potentially damage the air conditioner, switching to the feedback controller can enhance system safety. When the evaluation network's output is stable and reliable, it can be used for more refined regulation to optimize performance. In this way, the air conditioner's control method can be made more flexible and reliable, enabling intelligent switching between the evaluation network and the feedback controller to adapt to varying operating conditions, improving overall operational efficiency and reliability.

[0068] According to one embodiment of the present application, the control method of the air conditioner also includes: obtaining the suction pressure and exhaust pressure of the compressor of the air conditioner, and obtaining the temperature difference between the cooling water and the chilled water in the cooling circuit of the air conditioner; using the suction pressure, exhaust pressure, temperature difference, first temperature, first humidity and first load power as inputs of the evaluation network to output decision scores when performing multiple control operations on the air conditioner.

[0069] Specifically, to enable the evaluation network to more accurately predict the effects of different control operations, the air conditioner's compressor suction and discharge pressures, as well as the temperature difference between the cooling water and chilled water in the air conditioner's cooling circuit, are also obtained. The compressor suction and discharge pressures reflect the operating status of the air conditioner's compressor. Low suction pressure or high discharge pressure may indicate a compressor problem or the need to adjust operating parameters to improve efficiency. The temperature difference between the cooling water and chilled water reflects the heat exchange efficiency of the cooling circuit. Ideally, the temperature difference between the cooling water and chilled water should be large enough to ensure effective heat exchange.

[0070] After obtaining the air conditioner's compressor suction and discharge pressures and the temperature difference between the cooling water and chilled water in the air conditioner's cooling circuit, the suction pressure, discharge pressure, temperature difference, first temperature, first humidity, and first load power are used as inputs to an evaluation network to output decision scores for performing multiple control operations on the air conditioner. Specifically, the evaluation network outputs decision scores for performing multiple control operations on the air conditioner based on the input parameters. These scores reflect the expected effect of each possible control operation (such as adjusting the compressor operating frequency, changing the air supply volume, or adjusting the cooling water flow rate) under the current conditions. A higher score indicates a higher likelihood that the control operation will produce a positive effect, such as improving energy efficiency or maintaining temperature stability, under the current conditions. The specific method for determining the decision score is similar to the method described above for outputting decision scores for performing multiple control operations on the air conditioner using the first temperature, first humidity, and first load power as inputs to the target evaluation network, and will not be further elaborated here.

[0071] Therefore, by considering the working status of the compressor and the heat exchange efficiency of the cooling circuit, the air conditioner can be regulated more accurately to adapt to the current operating conditions. Accurate regulation can improve the energy efficiency of the air conditioner and reduce energy waste, thereby being able to automatically adjust the regulation strategy according to real-time operating parameters and environmental conditions to achieve the best operating effect.

[0072] According to one embodiment of the present application, the ratio of the power difference of the first load power at adjacent time points to the time interval between adjacent time points is obtained as the thermal load change rate; when the thermal load change rate is greater than zero, the air conditioner is pre-controlled based on the first strategy, wherein the first strategy includes at least one of increasing the cooling capacity, increasing the air supply volume, and decreasing the air supply temperature; when the thermal load change rate is less than zero, the air conditioner is pre-controlled based on the second strategy, wherein controlling the air conditioner includes at least one of reducing the cooling capacity, reducing the air supply volume, and increasing the air supply temperature.

[0073] Specifically, the electronic equipment heat load change rate parameter is a key predictive indicator. It helps us understand the heat load trend of electronic equipment within a data center. This allows us to predict heat load trends based on the heat load change rate parameter and adjust the air conditioner's cooling strategy in advance. Specifically, the heat load change rate is calculated by calculating the difference in load power between two adjacent time points, t and t-1, and dividing it by the time interval Δt. After determining the heat load change rate, it can be assessed. If the heat load change rate is greater than zero, it indicates a predicted heat load increase. Cooling capacity can be increased in advance to address the upcoming heat load peak. The supply air volume and temperature can also be adjusted. Increasing the supply air volume and decreasing the supply air temperature can be adjusted to more effectively distribute the cold air and improve cooling efficiency.

[0074] When the heat load change rate is less than zero, indicating a predicted decrease in heat load, appropriate control and adjustments are necessary to avoid energy waste or other potential problems caused by overcooling. For example, the cooling capacity of the air conditioning system can be reduced to match the current heat load demand based on the degree of heat load reduction. The supply air volume and temperature can also be adjusted, such as decreasing the supply air volume and increasing the supply air temperature, to more effectively utilize cold air, improve cooling efficiency, and avoid overcooling.

[0075] As a result, appropriate control and adjustments can be made according to the increase or decrease in heat load to ensure stable operation and efficient energy saving of the data center.

[0076] In addition, one embodiment of the application can monitor the rate of change of the heat load of electronic devices to more accurately predict future heat load trends and proactively adjust the air conditioner's cooling strategy. For example, real-time operating data from electronic devices, including CPU (Central Processing Unit) usage, memory usage, and network traffic, is collected and analyzed to calculate the rate of change of the heat load. A combined LSTM (Long Short-Term Memory) and Kalman filter model is employed. This model combines the advantages of LSTM and Kalman filters to improve the accuracy and real-time performance of heat load prediction. The LSTM network structure consists of three hidden layers, each with 128 neurons. This structure helps capture long-term dependencies in time series data. The model inputs are load data from the past 24 hours and current traffic volume. These data provide sufficient historical information and current status for the LSTM network to learn patterns in heat load changes. The process noise covariance is set to 0.01 to represent the uncertainty of the system model, and the measurement noise covariance is set to 0.05 to represent the uncertainty of the measurement data. The Kalman filter uses these parameters to correct the prediction error of the LSTM network in real time, thereby improving the accuracy of the prediction.

[0077] This collects real-time operating data from electronic devices, including heat load, CPU usage, memory usage, and network traffic. The rate of change of the heat load is calculated and used as a feature. This data, along with the load data from the past 24 hours and current service traffic, is fed into the LSTM network. The LSTM network then predicts future heat loads based on this input data. The LSTM predictions are then fed into a Kalman filter, which uses the process noise covariance and measurement noise covariance to correct prediction errors in real time, resulting in more accurate heat load forecasts. Based on these corrected heat load forecasts, the cooling strategy can be adjusted in advance, such as adjusting the compressor operating frequency or changing the air supply volume, to ensure that the data center temperature remains within an appropriate range.

[0078] Therefore, combining LSTM and Kalman filters can improve the accuracy of heat load prediction, thereby more effectively managing the temperature of the data center. By predicting the heat load in advance and adjusting the cooling strategy, energy efficiency can be improved and energy consumption can be reduced. In this way, the air conditioning system of the data center can operate more intelligently and efficiently, providing a stable and reliable operating environment for electronic equipment.

[0079] The following combination Figure 2 To describe the method of this application.

[0080] As a specific example, the control method of the air conditioner of the present application may include the following steps: S101 , obtaining a first temperature and a first humidity of each area of ​​a data center and a first load power of an electronic device at a current moment.

[0081] S102: Taking the first temperature, the first humidity, and the first load power as inputs of a target evaluation network to output decision scores when performing multiple control operations on the air conditioner.

[0082] S103: Determine a random number generation function based on a greedy strategy algorithm to generate a random number based on the random number generation function.

[0083] S104: Determine whether the random number is less than a preset probability threshold. If yes, go to step S105; if not, go to step S113.

[0084] S105 , randomly generating a target index based on a random number generation function, and selecting a corresponding control operation from a plurality of control operations as the control operation based on the target index, so that the air conditioner operates based on the control operation.

[0085] S106, obtaining the reward value corresponding to the operation of the air conditioner based on the control operation, and obtaining the second temperature and second humidity of each area of ​​the data center and the second load power of each electronic device after the air conditioner is operated based on the control operation, and storing the reward value, second temperature, second humidity, and second load power in the training sample.

[0086] S107: Determine whether the number of training samples meets a preset threshold. If yes, execute step S108; if not, execute step S101.

[0087] S108 , inputting the training samples into the evaluation network and the target network respectively to obtain a first decision score and a second decision score.

[0088] S109 , taking the sum of the product of the discount factor and the second decision score and the reward value as the target decision score, and determining the target loss function based on the difference between the first decision score and the target decision score.

[0089] S110 , taking the sum of the product of the discount factor and the second decision score and the reward value as the target decision score, and determining the target loss function based on the difference between the first decision score and the target decision score.

[0090] S111 , minimizing the target loss function based on a gradient descent algorithm so that the difference between the first decision score and the second decision score reaches a preset minimum value.

[0091] S112, updating the evaluation network to obtain a target evaluation network, so that the next time the air conditioner is regulated, a regulation operation is determined based on the target evaluation network, and the parameters of the updated evaluation network are copied to the target network.

[0092] S113: The control operation corresponding to the highest decision score is used as the control operation, so that the air conditioner operates based on the control operation, and then enters step S106.

[0093] In summary, according to the air conditioner control method of the embodiment of the present application, the first temperature and first humidity of each area of ​​the data center and the first load power of the electronic device at the current moment are obtained; the first temperature, first humidity and first load power are used as inputs of the target evaluation network to output decision scores when performing multiple control operations on the air conditioner, wherein the training samples are respectively input into the evaluation network and the target network to obtain a first decision score and a second decision score, and when the difference between the first decision score and the second decision score reaches a preset minimum value, the evaluation network is updated to obtain a target evaluation network; a random number generation function is determined based on a greedy strategy algorithm to generate a random number based on the random number generation function; when the random number is less than a preset probability threshold, a target index is randomly generated based on the random number generation function, and a corresponding control operation is selected from multiple control operations based on the target index as the control operation; when the random number is greater than the preset probability threshold, the control operation corresponding to the highest decision score is selected as the control operation. As a result, the method can quickly respond to changes in the load of the electronic equipment, accurately control the cooling capacity of the air conditioner, optimize the overall energy consumption of the data center, and ensure that the electronic equipment is always within a suitable operating temperature range.

[0094] Corresponding to the above embodiments, the present application also proposes a computer program product.

[0095] The computer program product of the embodiment of the present application includes a computer program / instruction, and when the computer program / instruction is executed by a processor, the above-mentioned air conditioner control method is implemented.

[0096] According to the computer program product of the embodiment of the present application, by executing the above-mentioned air conditioner control method, it is possible to quickly respond to changes in the load of electronic equipment, accurately adjust the cooling capacity of the air conditioner, optimize the overall energy consumption of the data center, and ensure that the electronic equipment is always in an appropriate operating temperature range.

[0097] Corresponding to the above embodiment, the present application also proposes an air conditioner.

[0098] like Figure 3 As shown, the air conditioner 200 of the embodiment of the present application may include: a memory 210, a processor 220 and a program stored in the memory 210 and executable on the processor 220. When the processor 220 executes the program, the above-mentioned air conditioner control method is implemented.

[0099] According to the air conditioner of the embodiment of the present application, by executing the above-mentioned air conditioner control method, it is possible to quickly respond to changes in the load of electronic equipment, accurately adjust the cooling capacity of the air conditioner, optimize the overall energy consumption of the data center, and ensure that the electronic equipment is always in an appropriate operating temperature range.

[0100] It should be noted that the logic and / or steps represented in flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic device), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner if necessary, and then storing it in a computer memory.

[0101] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0102] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present application. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0103] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of such features. Throughout the description of this application, "plurality" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0104] In this application, unless otherwise specified or limited, the terms "installed," "connected," "connect," "fixed," etc. should be understood in a broad sense. For example, they can refer to fixed connection, detachable connection, or integration; mechanical connection or electrical connection; direct connection or indirect connection through an intermediate medium; internal communication between two elements or interaction between two elements, unless otherwise specified. Those skilled in the art will understand the specific meanings of the above terms in this application based on specific circumstances.

[0105] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for controlling an air conditioner, characterized in that: The method comprises: Obtaining a first temperature and a first humidity of each area of ​​the data center and a first load power of the electronic equipment at the current moment; using the first temperature, the first humidity, and the first load power as inputs of a target evaluation network to output decision scores when performing a plurality of control operations on the air conditioner, wherein training samples are respectively input into the evaluation network and the target network to obtain a first decision score and a second decision score, and when the difference between the first decision score and the second decision score reaches a preset minimum value, updating the evaluation network to obtain the target evaluation network; Determine a random number generation function based on a greedy strategy algorithm to generate a random number based on the random number generation function; In a case where the random number is less than a preset probability threshold, randomly generating a target index based on the random number generation function, and selecting a corresponding control operation from the multiple control operations as the control operation based on the target index; When the random number is greater than the preset probability threshold, the control operation corresponding to the highest decision score is used as the control operation.

2. The air conditioner control method according to claim 1, characterized in that: After the air conditioner is operated based on the control operation, the method further includes: Obtaining a reward value corresponding to the operation of the air conditioner based on the control operation, and obtaining a second temperature and a second humidity of each area of ​​the data center and a second load power of each electronic device after the air conditioner is operated based on the control operation; The reward value, the second temperature, the second humidity, and the second load power are stored in the training samples, wherein the training samples are used to train the evaluation network.

3. The air conditioner control method according to claim 2, characterized in that: The training sample further includes the first temperature, the first humidity, the first load power, and the control operation at the current moment. Inputting the training sample into the evaluation network and the target network respectively to obtain a first decision score and a second decision score includes: using the first temperature, the first humidity, the first load power, and the control operation at the previous moment as inputs to the evaluation network to obtain a first decision score; using the second temperature, the second humidity, and the second load power as inputs of the evaluation network to output decision scores when performing multiple control operations on the air conditioner; Determine the update control operation corresponding to the highest decision score; The second temperature, the second humidity, the second load power, and the updated control operation are used as inputs of the target network to obtain a second decision score.

4. The air conditioner control method according to claim 3, characterized in that: The method further comprises: determining a target loss function based on the first decision score and the second decision score; The objective loss function is minimized based on a gradient descent algorithm so that the difference between the first decision score and the second decision score reaches a preset minimum value.

5. The air conditioner control method according to claim 4, characterized in that: The determining of a target loss function based on the first decision score and the second decision score includes: The product of the discount factor and the second decision score and the sum of the reward value are used as the target decision score; The target loss function is determined based on a difference between the first decision score and the target decision score.

6. The air conditioner control method according to claim 2, characterized in that: The obtaining of a reward value corresponding to the operation of the air conditioner based on the control operation includes: Obtaining a temperature balance index of the data center, a temperature stability index of the electronic equipment, and an energy efficiency index of the air conditioner; The reward value is determined based on the sum of a product of a first weight coefficient and the temperature balance index, a product of a second weight coefficient and the temperature stability index, and a product of a third weight coefficient and the energy efficiency index.

7. The air conditioner control method according to claim 6, characterized in that: Obtaining a temperature balance index of the data center includes: Calculate the average temperature based on the actual temperature of each area in the data center; The temperature balance index is determined based on the actual temperature of each area of ​​the data center, the corresponding weight coefficient, and the temperature average value.

8. The air conditioner control method according to claim 6, characterized in that: Obtaining a temperature stability index of the electronic device includes: Calculating the maximum and minimum temperature values ​​of the electronic device within a preset time period; determining a temperature fluctuation difference based on the maximum value and the minimum value; The temperature stability index is determined based on the temperature fluctuation difference and a preset mapping relationship, wherein the preset mapping relationship is used to indicate the relationship between the temperature fluctuation difference and the temperature stability index.

9. The air conditioner control method according to claim 6, characterized in that: Obtaining the energy efficiency index of the air conditioner, including: determining an energy efficiency ratio based on a ratio of a cooling capacity of the air conditioner to a power consumption of the air conditioner; When the energy efficiency ratio is greater than a preset energy efficiency ratio threshold, calculating a first energy efficiency ratio difference between the energy efficiency ratio and the preset energy efficiency ratio threshold, and determining the energy efficiency index based on a product of the first energy efficiency ratio difference and a positive reward value; When the energy efficiency ratio is less than the preset energy efficiency ratio threshold, a second energy efficiency ratio difference between the energy efficiency ratio and the preset energy efficiency ratio threshold is calculated, and the energy efficiency index is determined based on the product of the second energy efficiency ratio difference and the negative reward value.

10. The air conditioner control method according to claim 1, wherein: The method further comprises: monitoring the status of the air conditioner; In the event that an abnormality occurs in the air conditioner, the training parameters of the evaluation network are adjusted, wherein the abnormality in the air conditioner includes at least one of the following: a temperature in at least one area of ​​the data center is greater than a first preset temperature threshold and lasts for a first preset time period, or a temperature in at least one area is less than a second preset temperature threshold and lasts for a second preset time period, a suction pressure and a discharge pressure of the compressor of the air conditioner are not within a preset pressure range, and a flow rate of chilled water and cooling water is not within a preset flow range; Adjusting the training parameters of the evaluation network includes reducing at least one of an experience replay pool capacity, a batch training size, a learning rate, and a target network update frequency.

11. The air conditioner control method according to claim 1, wherein: The method further comprises: Obtaining a difference between a control parameter determined by the evaluation network when the control operation is performed and a control parameter determined by the feedback controller when the air conditioner is controlled; When the difference exceeds a preset difference threshold and lasts for a preset time, the air conditioner is controlled to operate based on the control parameters when the feedback controller controls the air conditioner.

12. The air conditioner control method according to claim 1, wherein: The method further comprises: Obtaining the suction pressure and exhaust pressure of the compressor of the air conditioner, and obtaining the temperature difference between the cooling water and the chilled water in the cooling circuit of the air conditioner; The suction pressure, the exhaust pressure, the temperature difference, the first temperature, the first humidity and the first load power are used as inputs of the evaluation network to output decision scores when performing multiple control operations on the air conditioner.

13. The air conditioner control method according to claim 1, wherein: The method further comprises: Obtaining a ratio of a power difference between the first load powers at adjacent time points to a time interval between adjacent time points as a thermal load change rate; When the heat load change rate is greater than zero, the air conditioner is controlled in advance based on a first strategy, wherein the first strategy includes at least one of increasing cooling capacity, increasing air supply volume, and decreasing air supply temperature; When the heat load change rate is less than zero, the air conditioner is controlled in advance based on a second strategy, wherein controlling the air conditioner includes at least one of reducing cooling capacity, reducing air supply volume, and increasing air supply temperature.

14. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the control method for the air conditioner according to any one of claims 1 to 13 is implemented.

15. An air conditioner, characterized in that: include: A memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for controlling the air conditioner according to any one of claims 1 to 13 is implemented.

Images