Base station temperature control method and device, electronic equipment and medium

Abstract

The invention discloses a base station temperature control method and device, electronic equipment and a medium, and belongs to the technical field of base station air conditioners. The method comprises the following steps: acquiring sample data of a base station; inputting the sample data into a strategy network to obtain a control strategy for the working frequency of the air conditioner compressor in the base station; inputting the sample data into a value network, and estimating an expected reward value obtained by executing a control strategy; the predicted temperature of the base station after the air conditioner is controlled according to the control strategy is determined, and a real reward value is determined according to the predicted temperature and the reward function; generating a loss value according to the expected reward value and the real reward value based on the dominant function; adjusting parameters of the strategy network and the value network based on the loss value until training of the strategy network and the value network is completed; and controlling the air conditioner by using the trained strategy network so as to adjust the temperature of the base station. Therefore, the problem that the temperature control and energy saving effects of the base station are not remarkable in the prior art is solved.

Description

Technical Field

[0001] This invention belongs to the field of base station air conditioning technology, specifically relating to a base station temperature control method, device, electronic equipment, and medium. Background Technology

[0002] With the development of wireless communication technology, the number of base stations in various places has gradually increased. In order to ensure the normal operation of base station equipment, it is necessary to regulate the temperature of the base station. The air conditioner that regulates the temperature of the base station needs to run for a long time, which consumes a lot of energy.

[0003] In existing technologies, energy-saving operation of base station air conditioners is achieved by improving the structure of the air conditioner. However, changing the original air conditioner structure is too costly and the energy-saving effect is not significant enough to meet the energy-saving requirements of air conditioners in base stations. Summary of the Invention

[0004] The purpose of this invention is to provide a base station temperature control method, device, electronic device, and medium that can solve the problem that base station temperature control in the prior art relies on the modification of the air conditioning structure and the energy-saving effect is not significant.

[0005] To solve the above-mentioned technical problems, the present invention is implemented as follows: In a first aspect, embodiments of the present invention provide a base station temperature control method, the method comprising: Acquire sample data from the base station; the sample data includes the sample ambient temperature of the base station, the sample internal temperature of the base station, and the sample target temperature; The sample data is input into the policy network to obtain a control policy for the operating frequency of the air conditioning compressor in the base station; The sample data is input into the value network to estimate the expected reward value to be obtained after implementing the control strategy; After determining the control strategy for controlling the air conditioner compressor, the base station predicts the temperature and determines the actual reward value based on the predicted temperature and the reward function. The actual reward value represents the actual cooling effect after executing the control strategy. Based on the advantage function, a loss value is generated according to the expected reward value and the actual reward value; The parameters of the policy network and the value network are adjusted based on the loss value until the training of the policy network and the value network is completed; The trained policy network is used to control the air conditioner to adjust the temperature of the base station.

[0006] Optionally, determining the predicted temperature of the base station after controlling the air conditioner compressor according to the control strategy includes: Obtain the operating condition data of the air conditioning compressor in the base station when it is running according to the control strategy; The cooling capacity of the air conditioner is determined based on the operating condition data. The predicted temperature of the base station is determined based on the ambient temperature of the sample, the internal temperature of the sample at the base station, and the cooling capacity.

[0007] Optionally, determining the predicted temperature of the base station based on the sample ambient temperature, the sample internal temperature of the base station, and the cooling capacity includes: The predicted temperature of the base station is determined according to the following formula based on the ambient temperature of the sample, the internal temperature of the sample, and the cooling capacity: , in, The internal temperature of the sample. The first time period Here, B is the proportional parameter, B is the total heat of the heat sources within the base station, and C is the heat capacity. The ambient temperature is Q, and the cooling capacity of the air conditioner is Q.

[0008] Optionally, determining the true reward value based on the predicted temperature and the reward function includes: The actual reward value is determined based on the predicted temperature and the target temperature of the sample according to the following formula: , Where r is the actual reward value, For the , The target temperature of the sample.

[0009] Optionally, controlling the air conditioner using the trained policy network to adjust the temperature of the base station includes: Obtain the current ambient temperature, the current internal temperature of the base station, and the current target temperature; The current ambient temperature, the current internal temperature of the base station, and the current target temperature are input into the policy network, and the policy network outputs a target control policy for the operating frequency of the air conditioning compressor in the base station. The air conditioning compressor is controlled according to the target control strategy.

[0010] Optionally, the step of outputting a target control strategy for the operating frequency of the air conditioner compressor within the base station through the policy network includes: If the current ambient temperature is higher than a first threshold temperature, and the difference between the current internal temperature and the current target temperature is greater than a second threshold temperature, then a first control strategy is output through the strategy network; the first control strategy is to control the air conditioner to run at a first frequency for a first duration; if the temperature inside the base station is detected to be equal to the current target temperature, then the air conditioner is controlled to run at a second frequency, where the first frequency is greater than the second frequency.

[0011] Optionally, the step of outputting a target control strategy for the operating frequency of the air conditioner compressor in the base station through the policy network further includes: If the current ambient temperature is higher than the first threshold temperature, and the difference between the current internal temperature and the current target temperature is less than the second threshold but greater than the third threshold, then a second control strategy is output through the strategy network; the first control strategy is to control the air conditioner to run at a first frequency for a first duration; if the temperature inside the base station is detected to be equal to the current target temperature, then the air conditioner is controlled to run at a second frequency, where the first frequency is greater than the second frequency.

[0012] Optionally, the step of outputting a target control strategy for the operating frequency of the air conditioner compressor in the base station through the policy network further includes: If the current ambient temperature is lower than the first threshold temperature, and the difference between the current internal temperature and the current target temperature is less than the third threshold, then the third control strategy is output through the strategy network; the third control strategy is to control the air conditioner to gradually decrease to the sixth frequency at the fifth frequency and run for a fourth duration; if the temperature inside the base station is detected to be equal to the current target temperature, then the air conditioner is controlled to run at the sixth frequency.

[0013] Secondly, embodiments of the present invention provide a base station temperature control device, comprising: a sample acquisition module, used to acquire sample data of the base station; the sample data includes the sample ambient temperature of the base station, the sample internal temperature of the base station, and the sample target temperature; The strategy network module is used to input the sample data into the strategy network to obtain a control strategy for the operating frequency of the air conditioning compressor in the base station. The value network module is used to input the sample data into the value network to estimate the expected reward value to be obtained after executing the control strategy; The reward function module is used to determine the predicted temperature of the base station after controlling the air conditioner compressor according to the control strategy, and to determine the actual reward value based on the predicted temperature and the reward function. The actual reward value represents the actual cooling effect after executing the control strategy. The advantage function module is used to generate a loss value based on the expected reward value and the actual reward value according to the advantage function; The debugging module is used to adjust the parameters of the policy network and the value network based on the loss value until the training of the policy network and the value network is completed; An execution module is used to control the air conditioner using the trained policy network to adjust the temperature of the base station.

[0014] Optionally, the reward function module includes: The data acquisition submodule is used to acquire the operating condition data of the air conditioning compressor in the base station when it is running according to the control strategy; The cooling capacity calculation submodule is used to determine the cooling capacity of the air conditioner based on the operating condition data. The predicted temperature calculation submodule is used to determine the predicted temperature of the base station based on the sample ambient temperature, the sample internal temperature of the base station, and the cooling capacity.

[0015] Optionally, the predicted temperature calculation submodule includes: The predicted temperature calculation unit is used to determine the predicted temperature of the base station according to the sample ambient temperature, the sample internal temperature, and the cooling capacity using the following formula: , in, The internal temperature of the sample. The first time period Here, B is the proportional parameter, B is the total heat of the heat sources within the base station, and C is the heat capacity. The ambient temperature is Q, and the cooling capacity of the air conditioner is Q.

[0016] Optionally, the reward function module includes: The reward value calculation unit is used to determine the actual reward value based on the predicted temperature and the target temperature of the sample according to the following formula: , Where r is the actual reward value, For the , The target temperature of the sample.

[0017] Optionally, the execution module includes: The current data collection submodule is used to acquire the current ambient temperature, the current internal temperature of the base station, and the current target temperature; The control strategy determination submodule is used to input the current ambient temperature, the current internal temperature of the base station and the current target temperature into the strategy network, and output the target control strategy for the operating frequency of the air conditioning compressor in the base station through the strategy network. An execution submodule is used to control the air conditioning compressor according to the target control strategy.

[0018] Optionally, the control strategy determination submodule includes: The first control strategy unit is configured to output a first control strategy through the strategy network if the current ambient temperature is higher than a first threshold temperature and the difference between the current internal temperature and the current target temperature is greater than a second threshold temperature; the first control strategy is to control the air conditioner to run at a first frequency for a first duration; if the temperature inside the base station is detected to be equal to the current target temperature, the air conditioner is controlled to run at a second frequency, wherein the first frequency is greater than the second frequency.

[0019] Optionally, the control strategy determination submodule further includes: The second control strategy unit is configured to output a second control strategy through the strategy network if the current ambient temperature is higher than a first threshold temperature, and the difference between the current internal temperature and the current target temperature is less than a second threshold but greater than a third threshold; the first control strategy is to control the air conditioner to run at a first frequency for a first duration; if the temperature inside the base station is detected to be equal to the current target temperature, the air conditioner is controlled to run at a second frequency, wherein the first frequency is greater than the second frequency.

[0020] Optionally, the control strategy determination submodule further includes: The third control strategy unit outputs a third control strategy through the strategy network if the current ambient temperature is lower than the first threshold temperature and the difference between the current internal temperature and the current target temperature is less than the third threshold temperature. The third control strategy is to control the air conditioner to gradually decrease from the fifth frequency to the sixth frequency and run for a fourth duration. If the temperature inside the base station is detected to be equal to the current target temperature, the air conditioner is controlled to run at the sixth frequency.

[0021] Thirdly, embodiments of the present invention provide an electronic device including a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the method described in the first aspect.

[0022] Fourthly, embodiments of the present invention provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0023] In this embodiment of the invention, a base station temperature control method includes acquiring sample data of the base station, wherein the sample data includes the sample ambient temperature, the sample internal temperature, and the sample target temperature of the base station; inputting the sample data into a policy network and a value network respectively to obtain a control policy for the operating frequency of the air conditioning compressor in the base station and to estimate the expected reward value to be obtained after executing the control policy; determining the predicted temperature of the base station after controlling the air conditioning compressor according to the control policy, and determining the actual reward value based on the predicted temperature and the reward function, wherein the actual reward value represents the actual cooling effect after executing the control policy; generating a loss value based on the expected reward value and the actual reward value based on the dominance function; adjusting the parameters of the policy network and the value network based on the loss value until the policy network and the value network are trained; and using the trained policy network to control the air conditioning to adjust the temperature of the base station. This invention trains a policy network by combining sample data, a value network, a reward function, and a dominance function. The output policy network can adapt to different operating conditions and control the operation of the air conditioning compressor, reducing frequent start-stop cycles or overcooling. Furthermore, this method does not rely on modifications to the air conditioning hardware or alter the existing air conditioning structure, resulting in low cost and easy deployment. Optimization of the software algorithm achieves energy-saving effects for base station air conditioning. Moreover, the collaborative training of the policy network and the value network ensures the long-term efficiency of the control strategy. The use of the reward function and the dominance function improves the accuracy and speed of the training process. This allows the invention to maintain stable base station temperature while improving energy efficiency and adapting to diverse base station environments, providing a low-cost, high-efficiency solution for temperature control within base stations.

[0024] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0025] Figure 1 This is a flowchart of a base station temperature control method according to an embodiment of the present invention; Figure 2 This is a flowchart of a base station temperature control method according to another embodiment of the present invention; Figure 3 This is a flowchart of a base station temperature control method according to an embodiment of the present invention; Figure 4 This is a structural block diagram of a base station temperature control device according to an embodiment of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] The terms "first," "second," etc., used in the specification and claims of this invention are used to distinguish similar objects and are not used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0028] With the development of wireless communication technology, the number of base stations in various places has gradually increased. In order to ensure the normal operation of base station equipment, it is necessary to regulate the temperature of the base station. The air conditioner that regulates the temperature of the base station needs to run for a long time, which consumes a lot of energy.

[0029] In existing technologies, energy-saving operation of base station air conditioners is achieved by improving the structure of the air conditioner. However, changing the original air conditioner structure is too costly and the energy-saving effect is not significant enough to meet the energy-saving requirements of air conditioners in base stations.

[0030] Therefore, one of the core concepts of this invention is that by combining sample data, value network, reward function and advantage function to train the policy network, the output policy network can guide the operation of the air conditioner without changing the existing air conditioner structure, and achieve energy-saving effect through software algorithm optimization.

[0031] This invention provides a base station temperature control method, device, electronic equipment, and medium, which can solve the problem that the existing technology of base station temperature control relies on the modification of the air conditioning structure and the energy-saving effect is not significant.

[0032] The base station temperature control method provided by the present invention will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0033] Figure 1 This is a flowchart of a base station temperature control method according to an embodiment of the present invention, as follows: Figure 1 As shown, this base station temperature control method is applicable to electronic devices. Figure 1 The illustrated embodiment includes at least the following steps 101-107.

[0034] In step 101, sample data of the base station is acquired; the sample data includes the sample ambient temperature of the base station, the sample internal temperature of the base station, and the sample target temperature.

[0035] First, sample data of the base station is acquired, specifically including ambient temperature, internal temperature, and target temperature. This sample data provides the necessary data for training, ensuring that the control strategy of the subsequently trained policy network can capture the thermal dynamic characteristics of the base station. Furthermore, by covering multi-dimensional data, such as ambient temperature reflecting the external temperature of the base station, internal temperature reflecting the internal temperature during operation, and target temperature serving as a standard for the internal temperature to determine whether it exceeds a reasonable range, this data better reflects the actual operating conditions of the base station and its air conditioning system, laying a data foundation for generating a reasonable control strategy.

[0036] In step 102, the sample data is input into the policy network to obtain a control policy for the operating frequency of the air conditioning compressor in the base station.

[0037] In step 102, sample data is input into the strategy network to obtain a control strategy for the operating frequency of the air conditioner compressor. The strategy network learns and outputs frequency adjustment control strategies (such as frequency increase, frequency decrease, or frequency hold), which can be used to respond to different operating conditions in real time, avoiding the traditional operation of frequently starting and stopping the air conditioner. In this embodiment of the invention, the strategy network can adaptively adjust the output control strategy according to the changes in the input sample ambient temperature, sample internal temperature, and sample target temperature, so as to achieve precise conditions for the air conditioner compressor, thereby reducing the unnecessary high-frequency operation time of the air conditioner compressor, thereby reducing energy consumption and reducing the wear and tear of the air conditioner compressor equipment.

[0038] In step 103, the sample data is input into the value network to estimate the expected reward value to be obtained after executing the control strategy.

[0039] In the above steps, sample data is input into the value network to estimate the expected reward value after implementing the control policy. It can be understood that the value network provides a long-term reference for evaluating the control policy by predicting the long-term cumulative reward, which is used to balance immediate and long-term effects during training. Step 103 introduces a value estimation mechanism to ensure that the control policy not only focuses on short-term cooling within the base station but also predicts and considers overall energy efficiency optimization during long-term operation, thereby improving the stability and long-term sustainability of the control policy generated by the policy network.

[0040] In step 104, the predicted temperature of the base station is determined after the air conditioner compressor is controlled according to the control strategy, and the actual reward value is determined based on the predicted temperature and the reward function. The actual reward value represents the actual cooling effect after the control strategy is executed.

[0041] The predicted temperature of the base station is determined after the air conditioning compressor is controlled according to the control strategy. The actual reward value is determined based on the predicted temperature and the reward function. After the control strategy controls the air conditioning compressor, the temperature inside the base station will gradually change from the original sample internal temperature to the predicted temperature. The reward function processes the predicted temperature to obtain the actual reward value to quantify the actual cooling effect. It can be understood that the reward function can provide immediate feedback on the effect of strategy execution, so that the training process has a clear optimization goal.

[0042] In step 105, a loss value is generated based on the expected reward value and the actual reward value using the advantage function.

[0043] In step 105 above, the advantage function calculates the difference between the expected reward value and the actual reward value, which is used to quantify the direction of improvement of the control strategy. The advantage function generates a loss value, which can be used to update the value network and the policy network, thereby ensuring the collaborative optimization of the policy network and the value network. Furthermore, by reducing estimation errors, the accuracy and stability of the control strategy are improved while increasing the training efficiency of the policy network.

[0044] In step 106, the parameters of the policy network and the value network are adjusted based on the loss value until the training of the policy network and the value network is completed.

[0045] By iteratively updating the parameters of the policy network and the value network, the policy network can gradually converge to an efficient control policy, and the value network can more accurately predict the expected reward value. This improves the adaptability of the fir tree method and ensures that the policy network can be continuously updated in the simulation environment to improve the corresponding control policy.

[0046] In step 107, the trained policy network is used to control the air conditioner to adjust the temperature of the base station.

[0047] Then, the trained policy network is used to regulate the air conditioner, specifically the air conditioner compressor, thereby further adjusting the internal temperature of the base station. The trained policy network is deployed to achieve intelligent real-time control, automatically adjusting the compressor frequency according to the current operating conditions to achieve energy-saving control, improve system reliability and speed up response, and transform the training results into practical applications, effectively reducing the energy consumption of the air conditioner compressor during operation.

[0048] This invention trains a policy network by combining sample data, a value network, a reward function, and a dominance function. The output policy network can adapt to different operating conditions and control the operation of the air conditioning compressor, reducing frequent start-stop cycles or overcooling. Furthermore, this method does not rely on modifications to the air conditioning hardware or alter the existing air conditioning structure, resulting in low cost and easy deployment. Optimization of the software algorithm achieves energy-saving effects for base station air conditioning. Moreover, the collaborative training of the policy network and the value network ensures the long-term efficiency of the control strategy. The use of the reward function and the dominance function improves the accuracy and speed of the training process. This allows the invention to maintain stable base station temperature while improving energy efficiency and adapting to diverse base station environments, providing a low-cost, high-efficiency solution for temperature control within base stations.

[0049] In some embodiments, both the policy network and the value network are constructed using a fully connected network structure.

[0050] Secondly, Figure 2 This is a flowchart of a base station temperature control device according to another embodiment of the present invention. Figure 3 This is a flowchart of a base station temperature control method according to an embodiment of the present invention; please refer to... Figure 2 and Figure 3 This invention provides a base station temperature control method, comprising at least the following steps 201-209: In step 201, sample data of the base station is obtained; the sample data includes the sample ambient temperature of the base station, the sample internal temperature of the base station, and the sample target temperature.

[0051] First, sample data of the base station is acquired, specifically including ambient temperature, internal temperature, and target temperature. This sample data provides the necessary data for training, ensuring that the control strategy of the subsequently trained policy network can capture the thermal dynamic characteristics of the base station. Furthermore, by covering multi-dimensional data, such as ambient temperature reflecting the external temperature of the base station, internal temperature reflecting the internal temperature during operation, and target temperature serving as a standard for the internal temperature to determine whether it exceeds a reasonable range, this data better reflects the actual operating conditions of the base station and its air conditioning system, laying a data foundation for generating a reasonable control strategy.

[0052] In step 202, the sample data is input into the policy network to obtain a control policy for the operating frequency of the air conditioning compressor in the base station.

[0053] In step 202, sample data is input into the strategy network to obtain a control strategy for the operating frequency of the air conditioner compressor. The strategy network learns and outputs frequency adjustment control strategies (such as frequency increase, frequency decrease, or frequency hold), which can be used to respond to different operating conditions in real time, avoiding the traditional operation of frequent start-stop control of the air conditioner. In this embodiment of the invention, the strategy network can adaptively adjust the output control strategy according to the changes in the input sample ambient temperature, sample internal temperature, and sample target temperature, so as to achieve precise conditions for the air conditioner compressor, thereby reducing the unnecessary high-frequency operation time of the air conditioner compressor, thereby reducing energy consumption and reducing the wear and tear of the air conditioner compressor equipment.

[0054] In step 203, the sample data is input into the value network to estimate the expected reward value to be obtained after executing the control strategy.

[0055] In the above steps, sample data is input into the value network to estimate the expected reward value after implementing the control policy. It can be understood that the value network provides a long-term reference for evaluating the control policy by predicting the long-term cumulative reward, which is used to balance immediate and long-term effects during training. Step 203 introduces a value estimation mechanism to ensure that the control policy not only focuses on short-term cooling within the base station but also predicts and considers overall energy efficiency optimization during long-term operation, thereby improving the stability and long-term sustainability of the control policy generated by the policy network.

[0056] In step 204, the predicted temperature of the base station is determined after the air conditioner compressor is controlled according to the control strategy, and the actual reward value is determined based on the predicted temperature and the reward function. The actual reward value represents the actual cooling effect after the control strategy is executed.

[0057] The predicted temperature of the base station is determined after the air conditioning compressor is controlled according to the control strategy. The actual reward value is determined based on the predicted temperature and the reward function. After the control strategy controls the air conditioning compressor, the temperature inside the base station will gradually change from the original sample internal temperature to the predicted temperature. The reward function processes the predicted temperature to obtain the actual reward value to quantify the actual cooling effect. It can be understood that the reward function can provide immediate feedback on the effect of strategy execution, so that the training process has a clear optimization goal.

[0058] Step 204 also includes the following steps: Sub-step S11: Obtain the operating condition data of the air conditioning compressor in the base station running according to the control strategy.

[0059] Specifically, the system acquires operating condition data of the air conditioning compressor within the base station as it operates according to the control strategy, thereby transforming the control commands (such as frequency changes) output by the strategy network into specific and quantifiable operating condition data during the air conditioning operation. In some embodiments, the operating condition data includes relevant data such as the frequency and speed of the air conditioning compressor, which can accurately reflect the actual working state of the compressor under different control strategies, providing accurate input for subsequent calculations and ensuring the effectiveness of the training process.

[0060] Sub-step S12: Determine the cooling capacity of the air conditioner based on the operating condition data.

[0061] In some embodiments, a cooling capacity model of the base station is constructed. By inputting corresponding operating condition data, such as the frequency and speed of the air conditioner compressor, into the cooling capacity model, the output is the cooling capacity (i.e., cooling output) of the air conditioner. This simulates the operation of the air conditioner in the actual physical environment, thereby establishing a quantitative relationship between control actions and cooling effect. Furthermore, the cooling capacity model maps the operating condition data such as compressor frequency to specific cooling output, thereby effectively quantifying the energy consumption and cooling capacity of the control strategy. This allows the reward function to be evaluated based on the actual energy conversion efficiency, thereby guiding the policy network to learn a control strategy that cools down quickly and consumes less energy.

[0062] Sub-step S13: Determine the predicted temperature of the base station based on the sample ambient temperature, the sample internal temperature of the base station, and the cooling capacity.

[0063] In some embodiments, a room load model is constructed to simulate the dynamic thermodynamic process of cooling capacity acting on the interior space of the base station. This enables accurate and repeatable prediction of the consequences of control policy execution without the need for costly trial and error in a real base station environment. The room load model couples air conditioning cooling, heat exchange between the external and internal environments of the base station, and indoor temperature changes to form a closed-loop simulation system. This allows reinforcement learning algorithms to perform efficient iterative learning in a near-realistic environment, enabling the trained policy network to effectively cope with various complex and dynamically changing operating conditions.

[0064] In steps S11 to S13, the above steps together constitute a complete simulation environment. The cooling capacity model is used to simulate the operation of the air conditioner, and the room load model is used to simulate the internal state of the base station. By constructing a simulation environment that closely resembles reality, the reliability, security, and efficiency of reinforcement learning training of the policy network are significantly improved. By optimizing the policy network under the constraints defined by the cooling capacity model and the room load model, it is ensured that the instructions of the control policy generated by the finally trained policy network not only meet the energy-saving target but also satisfy the laws of thermodynamics. This avoids the risk of invalid oscillations or equipment damage that may occur when the control policy is executed in a real system, providing a solid foundation for generating an energy-saving and robust temperature control policy.

[0065] Furthermore, sub-step S13 also includes: determining the predicted temperature of the base station according to the sample ambient temperature, the sample internal temperature, and the cooling capacity using the following formula: , in, The internal temperature of the sample. The first time period Here, B is the proportional parameter, B is the total heat of the heat sources within the base station, and C is the heat capacity. The ambient temperature is Q, and the cooling capacity of the air conditioner is Q.

[0066] The first time period refers to a period during which the control strategy is implemented, and during the first time period, the temperature inside the base station changes from the internal temperature of the sample to the predicted temperature.

[0067] The above formula comprehensively considers environmental heat exchange, internal heat sources of the base station, and the cooling effect of the air conditioner, thus forming a complete heat balance equation. This ensures that the calculated predicted temperature reflects the actual thermodynamic behavior inside and outside the base station. By optimizing the policy network under the constraints defined by the physical model, it ensures that the control strategy generated by the policy network not only meets energy-saving requirements but also follows thermodynamic laws. This avoids the risks of temperature overshoot, fluctuations, or equipment damage that may occur in the real base station environment. Furthermore, the training process can be iterated safely and efficiently in a virtual environment without trial and error in the real base station environment, reducing training costs and risks. At the same time, it ensures that the generated policy network can adapt to various complex operating conditions, such as different ambient temperatures and changes in internal heat load.

[0068] In some embodiments, the proportional parameters, total heat, and heat capacity need to be updated in real time. Specifically, each update process updates only one parameter, while the other two are treated as known quantities. These are then substituted with the predicted temperature value, the internal temperature of the sample, the first time period, the cooling capacity, and the other two known parameter values ​​to update the relevant parameters. It is understandable that the environment inside a base station is not static. For example, the total heat inside the base station may increase during peak electricity consumption periods. Therefore, adjustments to the relevant parameters are necessary to ensure that the room load model reflects a more realistic base station situation, thereby guaranteeing that the control strategy output by the trained policy network can better cope with different display conditions.

[0069] Step 204 above also includes the following step: determining the true reward value based on the predicted temperature and the target temperature of the sample according to the following formula: , Where r is the actual reward value, For the , The target temperature of the sample.

[0070] The above reward function formula has the characteristics of being continuous, smooth, and unimodal. Under the action of the above reward function formula, control strategies with different temperature rates will be trained to achieve different cooling effects.

[0071] The design of this reward function enables the policy network to learn two key behaviors simultaneously. Specifically, on the one hand, the exponential term... Large temperature deviations can cause a sharp drop in the actual reward value. Therefore, the incentive strategy network takes effective measures to quickly reduce the temperature difference when the initial temperature deviation is large, demonstrating optimization of the "cooling speed." On the other hand, when the predicted temperature is close to the target temperature of the sample, the reward function reaches an extreme point (…). The temperature changes are relatively flat in the vicinity, which reduces the penalty for small temperature fluctuations. This encourages the policy network to adopt gentler and smoother frequency adjustments to maintain temperature stability, rather than frequently and drastically adjusting the air conditioner compressor frequency. This effectively suppresses temperature overshoot and fluctuations, reduces the impact on the air conditioner compressor, and lowers the energy consumption of the air conditioner compressor.

[0072] Correspondingly, the constants 5 and 0.1 in the formula play a role in adjusting the reward scale and sensitivity in the reward function. The coefficient 5 keeps the true reward value at a suitable level, facilitating stable gradient calculation and parameter updates for the policy network, while the parameter 0.1 controls the tolerance to temperature deviations. The setting of these constants allows the reward function to provide significant reward changes within a reasonable temperature deviation range, reducing the problem of insufficient effective guidance in the early stages of training due to an overly flat reward function, and also reducing the limitation on the exploration space of the policy network due to an overly steep reward function.

[0073] The reward function provides an evaluation criterion for training the policy network that combines clear objectives with flexible constraints. By driving the reinforcement learning algorithm through the reward function, the system can automatically discover and optimize control strategies that can achieve both rapid cooling and precise temperature stabilization under different operating conditions. This reduces the high energy consumption and equipment wear inherent in existing start-stop control methods, and enables efficient, stable and energy-saving operation of base station air conditioners.

[0074] In step 205, a loss value is generated based on the expected reward value and the actual reward value according to the advantage function.

[0075] In step 205 above, the advantage function calculates the difference between the expected reward value and the actual reward value, which is used to quantify the direction of improvement of the control strategy. The advantage function generates a loss value, which can be used to update the value network and the policy network, thereby ensuring the collaborative optimization of the policy network and the value network. Furthermore, by reducing estimation errors, the accuracy and stability of the control strategy are improved while increasing the training efficiency of the policy network.

[0076] In some embodiments, the advantage function outputs accurate estimates of the expected reward value and the actual reward value, which are used to measure the value state after performing an action in the current state, and to evaluate the merits of the current strategy.

[0077] In step 206, the parameters of the policy network and the value network are adjusted based on the loss value until the training of the policy network and the value network is completed.

[0078] By iteratively updating the parameters of the policy network and the value network, the policy network can gradually converge to an efficient control policy, and the value network can more accurately predict the expected reward value. This improves the adaptability of the fir tree method and ensures that the policy network can be continuously updated in the simulation environment to improve the corresponding control policy.

[0079] In step 207, the current ambient temperature, the current internal temperature of the base station, and the current target temperature are obtained.

[0080] In the above steps, the current ambient temperature, the current internal temperature of the base station, and the current target temperature are obtained. By obtaining these temperature values, the control strategy is formulated based on the real-time status parameters of the system, rather than historical or fixed data, thus providing dynamic input conditions. This allows for timely and complete capture of changes in the external environment and fluctuations in the heat load inside the base station, laying a data foundation for the strategy network to generate accurate control strategies and avoiding control deviations or energy efficiency losses caused by information lag.

[0081] In step 208, the current ambient temperature, the current internal temperature of the base station, and the current target temperature are input into the policy network, and the policy network outputs a target control policy for the operating frequency of the air conditioning compressor in the base station. The current ambient temperature, the current internal temperature of the base station, and the current target temperature are input into the policy network. The policy network then outputs a target control policy for the operating frequency of the air conditioning compressor within the base station. After training, the policy network exhibits practicality and fast inference capabilities, encapsulating complex control decision-making processes within a neural network. Without online optimization or manual intervention, it can output control policies (such as frequency increase, frequency decrease, or maintaining the current frequency) based on the current ambient temperature, the current internal temperature of the base station, and the current target temperature. This not only improves control response speed but also ensures the matching of the control policy with the current operating conditions. It can adaptively apply energy-saving modes learned during training (such as rapid cooling and stable maintenance) to directly optimize the compressor's operating efficiency.

[0082] Step 208 includes the following sub-steps: In step S21, the step of outputting a target control strategy for the operating frequency of the air conditioner compressor in the base station through the strategy network includes: if the current ambient temperature is higher than a first threshold temperature and the difference between the current internal temperature and the current target temperature is greater than a second threshold temperature, then a first control strategy is output through the strategy network; the first control strategy is to control the air conditioner to run at a first frequency for a first duration; if the temperature in the base station is detected to be equal to the current target temperature, then the air conditioner is controlled to run at a second frequency, wherein the first frequency is greater than the second frequency.

[0083] In step S21, the determination condition is that the current ambient temperature is high and the difference between the current internal temperature and the target temperature is large. This operating condition usually corresponds to an emergency situation where the base station equipment faces the risk of overheating. Under this condition, the policy network outputs the first control policy. Specifically, firstly, the air conditioner is controlled to run at a higher first frequency for a first duration, which can quickly provide a large amount of cooling and forcibly and rapidly reduce the internal temperature of the base station, thus prioritizing the safe operation of the base station equipment under high temperature and harsh conditions. Subsequently, after detecting that the temperature has been reached, the frequency is immediately switched to a lower second frequency. This avoids the energy waste caused by the air conditioner compressor continuing to run at high power after the target temperature has been reached, and achieves a smooth transition from "full protection" to "precise maintenance", while effectively preventing temperature overshoot caused by inertia.

[0084] In step S22, the step of outputting a target control strategy for the operating frequency of the air conditioner compressor in the base station through the strategy network further includes: if the current ambient temperature is higher than a first threshold temperature, and the difference between the current internal temperature and the current target temperature is less than a second threshold and greater than a third threshold, then a second control strategy is output through the strategy network; the first control strategy is to control the air conditioner to run at a first frequency for a first duration; if the temperature in the base station is detected to be equal to the current target temperature, then the air conditioner is controlled to run at a second frequency, wherein the first frequency is greater than the second frequency.

[0085] In step S22, the determination condition is that the ambient temperature is high but the internal temperature difference is in a medium range. At this time, the second control strategy is output. Specifically, the second control strategy instructs the air conditioner compressor to run at the third frequency for a short period of time. This stage aims to quickly accumulate enough cooling capacity to suppress the temperature rise caused by the high external temperature. Then, the instruction frequency is gradually reduced to the fourth frequency. This is a smooth transition process, which ensures that cooling capacity is accumulated first and then gradually converged. This avoids the energy loss caused by blindly using the highest frequency under non-extreme operating conditions. It also prevents drastic temperature fluctuations and frequent start-stop of the compressor through the gradual adjustment of the frequency, achieving a better balance between safety and energy saving.

[0086] In step S23, the step of outputting a target control strategy for the operating frequency of the air conditioner compressor in the base station through the strategy network further includes: if the current ambient temperature is lower than a first threshold temperature and the difference between the current internal temperature and the current target temperature is less than a third threshold, then a third control strategy is output through the strategy network; the third control strategy is to control the air conditioner to gradually decrease from a fifth frequency to a sixth frequency and run for a fourth duration; if the temperature in the base station is detected to be equal to the current target temperature, then the air conditioner is controlled to run at the sixth frequency.

[0087] In step S23, the determination condition is that the ambient temperature is mild and the internal temperature difference is small. At this time, the base station heat load is the lowest. The strategy network outputs the third control strategy, which controls the compressor to start from a relatively high fifth frequency and gradually reduce it to the sixth frequency. The mild start and continuous slow drop method makes the cooling process almost match the natural heating process of the base station. This avoids the high energy consumption behavior of the compressor repeatedly starting to compensate for the small temperature rise in traditional start-stop control. It allows the compressor to always operate stably in a lower frequency range, thereby minimizing energy consumption. At the same time, it achieves extremely stable temperature control, which is suitable for base stations with small heat loads to maintain long-term stable operation.

[0088] In step 209, the air conditioning compressor is controlled according to the target control strategy.

[0089] In some embodiments, the air conditioning compressor is controlled according to the target control strategy. The above steps translate the control strategy into actual execution actions, directly adjusting the operating frequency of the air conditioning compressor. This achieves continuous and precise frequency control, replacing the binary switch mode in traditional start-stop control. This significantly reduces the number of frequent start-stop cycles and the full-speed operation time of the air conditioning compressor. It not only reduces equipment wear of the compressor caused by start-stop shocks, but also further compresses energy consumption space by avoiding excessive cooling and temperature fluctuations.

[0090] Steps 207 to 209 above complete the closed loop of the policy network from simulation training to practical application, enabling the system to achieve efficient and adaptive temperature control of the base station in a real environment. Through real-time data-driven and rapid deployment of the policy network, the operating strategy of the air conditioning compressor is dynamically optimized while ensuring the safe operation of the base station, effectively suppressing energy consumption peaks and equipment losses. At the same time, no hardware modification is required, reducing implementation costs, thereby solving the core problems of high cost of structural improvement and insufficient energy-saving effect in the background technology.

[0091] Secondly, please refer to Figure 4 This invention provides a base station temperature control device, comprising: The sample acquisition module 301 is used to acquire sample data from the base station; the sample data includes the sample ambient temperature of the base station, the sample internal temperature of the base station, and the sample target temperature. The strategy network module 302 is used to input the sample data into the strategy network to obtain a control strategy for the operating frequency of the air conditioning compressor in the base station. Value network module 303 is used to input the sample data into the value network to estimate the expected reward value to be obtained after executing the control strategy; The reward function module 304 is used to determine the predicted temperature of the base station after controlling the air conditioner compressor according to the control strategy, and to determine the actual reward value based on the predicted temperature and the reward function. The actual reward value represents the actual cooling effect after executing the control strategy. Advantage function module 305 is used to generate a loss value based on the expected reward value and the actual reward value according to the advantage function; The debugging module 306 is used to adjust the parameters of the policy network and the value network based on the loss value until the training of the policy network and the value network is completed. The execution module 307 is used to control the air conditioner using the trained policy network to adjust the temperature of the base station.

[0092] This invention trains a policy network by combining sample data, a value network, a reward function, and a dominance function. The output policy network can adapt to different operating conditions and control the operation of the air conditioning compressor, reducing frequent start-stop cycles or overcooling. Furthermore, this method does not rely on modifications to the air conditioning hardware or alter the existing air conditioning structure, resulting in low cost and easy deployment. Optimization of the software algorithm achieves energy-saving effects for base station air conditioning. Moreover, the collaborative training of the policy network and the value network ensures the long-term efficiency of the control strategy. The use of the reward function and the dominance function improves the accuracy and speed of the training process. This allows the invention to maintain stable base station temperature while improving energy efficiency and adapting to diverse base station environments, providing a low-cost, high-efficiency solution for temperature control within base stations.

[0093] Optionally, the reward function module 304 includes: The data acquisition submodule is used to acquire the operating condition data of the air conditioning compressor in the base station when it is running according to the control strategy; The cooling capacity calculation submodule is used to determine the cooling capacity of the air conditioner based on the operating condition data. The predicted temperature calculation submodule is used to determine the predicted temperature of the base station based on the sample ambient temperature, the sample internal temperature of the base station, and the cooling capacity.

[0094] Optionally, the predicted temperature calculation submodule includes: The predicted temperature calculation unit is used to determine the predicted temperature of the base station according to the sample ambient temperature, the sample internal temperature, and the cooling capacity using the following formula: , in, The internal temperature of the sample. The first time period Here, B is the proportional parameter, B is the total heat of the heat sources within the base station, and C is the heat capacity. The ambient temperature is Q, and the cooling capacity of the air conditioner is Q.

[0095] Optionally, the reward function module 304 includes: The reward value calculation unit is used to determine the actual reward value based on the predicted temperature and the target temperature of the sample according to the following formula: , Where r is the actual reward value, For the , The target temperature of the sample.

[0096] Optionally, the execution module 307 includes: The current data collection submodule is used to acquire the current ambient temperature, the current internal temperature of the base station, and the current target temperature; The control strategy determination submodule is used to input the current ambient temperature, the current internal temperature of the base station and the current target temperature into the strategy network, and output the target control strategy for the operating frequency of the air conditioning compressor in the base station through the strategy network. An execution submodule is used to control the air conditioning compressor according to the target control strategy.

[0097] Optionally, the control strategy determination submodule includes: The first control strategy unit is configured to output a first control strategy through the strategy network if the current ambient temperature is higher than a first threshold temperature and the difference between the current internal temperature and the current target temperature is greater than a second threshold temperature; the first control strategy is to control the air conditioner to run at a first frequency for a first duration; if the temperature inside the base station is detected to be equal to the current target temperature, the air conditioner is controlled to run at a second frequency, wherein the first frequency is greater than the second frequency.

[0098] Optionally, the control strategy determination submodule further includes: The second control strategy unit is configured to output a second control strategy through the strategy network if the current ambient temperature is higher than a first threshold temperature, and the difference between the current internal temperature and the current target temperature is less than a second threshold but greater than a third threshold; the first control strategy is to control the air conditioner to run at a first frequency for a first duration; if the temperature inside the base station is detected to be equal to the current target temperature, the air conditioner is controlled to run at a second frequency, wherein the first frequency is greater than the second frequency.

[0099] Optionally, the control strategy determination submodule further includes: The third control strategy unit outputs a third control strategy through the strategy network if the current ambient temperature is lower than the first threshold temperature and the difference between the current internal temperature and the current target temperature is less than the third threshold temperature. The third control strategy is to control the air conditioner to gradually decrease from the fifth frequency to the sixth frequency and run for a fourth duration. If the temperature inside the base station is detected to be equal to the current target temperature, the air conditioner is controlled to run at the sixth frequency.

[0100] As the apparatus embodiment is basically similar to the method embodiment, it is described in a relatively simple manner. For relevant details, please refer to the description of the method embodiment.

[0101] This invention also provides an electronic device, including: a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, it implements the various processes of the above-described base station temperature control method embodiments and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0102] This invention also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described base station temperature control method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0103] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0104] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of the present invention is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0105] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, electronic device, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0106] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of the present invention.

Claims

1. A base station temperature control method, characterized by, The method comprises: obtaining sample data of a base station; the sample data comprises a sample ambient temperature of the base station, a sample internal temperature of the base station and a sample target temperature; inputting the sample data into a policy network to obtain a control strategy for a working frequency of an air conditioner compressor in the base station; inputting the sample data into a value network to estimate an expected reward value expected to be obtained after the control strategy is executed; determining a predicted temperature of the base station after the air conditioner compressor is controlled according to the control strategy, and determining a real reward value according to the predicted temperature and a reward function, the real reward value representing an actual cooling effect after the control strategy is executed; generating a loss value based on the expected reward value and the real reward value according to an advantage function; adjusting parameters of the policy network and the value network based on the loss value until the policy network and the value network are trained; controlling the air conditioner using the trained policy network to adjust the temperature of the base station.

2. The base station temperature control method of claim 1, wherein The method further comprises: obtaining working condition data of the air conditioner compressor when the air conditioner compressor operates according to the control strategy; determining a refrigeration capacity of the air conditioner according to the working condition data; determining the predicted temperature of the base station according to the sample ambient temperature, the sample internal temperature of the base station and the refrigeration capacity.

3. The base station temperature control method of claim 2, wherein The method further comprises: determining the predicted temperature of the base station according to the sample ambient temperature, the sample internal temperature and the refrigeration capacity according to the following formula: ， wherein, is the internal temperature of the sample, is the first time period, is a proportional parameter, B is the total heat of the heat source within the base station, and C is the heat capacity, is the sample ambient temperature, and Q is the refrigerating capacity of the air conditioner.

4. The base station temperature control method of claim 1, wherein The method further comprises: determining the real reward value according to the predicted temperature and the sample target temperature according to the following formula: ， wherein r is a true reward value, is the , is the sample target temperature.

5. The base station temperature control method of claim 1, wherein The method further comprises: obtaining a current ambient temperature, a current internal temperature of the base station and a current target temperature; inputting the current ambient temperature, the current internal temperature of the base station and the current target temperature into the policy network to output a target control strategy for the working frequency of the air conditioner compressor in the base station through the policy network; controlling the air conditioner compressor according to the target control strategy.

6. The base station temperature control method of claim 5, wherein, The method further comprises: if the current ambient temperature is higher than a first threshold temperature and a difference between the current internal temperature and the current target temperature is greater than a second threshold temperature, outputting a first control strategy through the policy network; the first control strategy is to control the air conditioner to operate at a first frequency for a first time length; if it is detected that the temperature in the base station is equal to the current target temperature, controlling the air conditioner to operate at a second frequency, the first frequency being greater than the second frequency.

7. The base station temperature control method of claim 6, wherein The method further comprises: if the current environment temperature is higher than a first threshold temperature, a difference between the current internal temperature and the current target temperature is less than a second threshold and greater than a third threshold, outputting a second control strategy through the policy network; the first control strategy is to control the air conditioner to run at a first frequency for a first time length; if it is detected that the temperature in the base station is equal to the current target temperature, controlling the air conditioner to run at a second frequency, the first frequency being greater than the second frequency.

8. The base station temperature control method of claim 7, wherein, The outputting, by the policy network, of the target control strategy for the working frequency of the air conditioner compressor in the base station further includes: if the current environment temperature is lower than the first threshold temperature, the difference between the current internal temperature and the current target temperature is less than the third threshold, outputting a third control strategy through the policy network; the third control strategy is to control the air conditioner to gradually reduce from a fifth frequency to a sixth frequency and run for a fourth time length; if it is detected that the temperature in the base station is equal to the current target temperature, controlling the air conditioner to run at the sixth frequency.

9. A device for temperature control of a base station, c h a r a c t e r i s e d i n that The method further includes: a sample acquisition module configured to acquire sample data of a base station; the sample data including a sample environment temperature of the base station, a sample internal temperature of the base station, and a sample target temperature; a policy network module configured to input the sample data into a policy network to obtain a control strategy for a working frequency of an air conditioner compressor in the base station; a value network module configured to input the sample data into a value network to estimate an expected reward value expected to be obtained after the control strategy is executed; a reward function module configured to determine a predicted temperature of the base station after the air conditioner compressor is controlled according to the control strategy, and determine a real reward value according to the predicted temperature and a reward function, the real reward value representing an actual cooling effect after the control strategy is executed; an advantage function module configured to generate a loss value based on an advantage function according to the expected reward value and the real reward value; a debugging module configured to adjust parameters of the policy network and the value network based on the loss value until the policy network and the value network are trained completely; an execution module configured to control the air conditioner using the trained policy network to adjust the temperature of the base station.

10. An electronic device, comprising: A processor, a memory, and a program or instructions stored on the memory and executable on the processor, the program or instructions being executed by the processor to implement the steps of the base station temperature control method of claims 1-8.

11. A readable storage medium, characterized by, A readable storage medium storing a program or instructions, the program or instructions being executed by a processor to implement the steps of the base station temperature control method of claims 1-8.

Images