Water cooling control method, system, equipment and medium

Abstract

The invention is suitable for the technical field of intelligent control, and discloses a water cooling control method, system and device and a medium. Based on the preprocessed data, constructing a thermal load trend prediction model based on LSTM to obtain a cold load prediction value; based on the cold load prediction value, a PID adjusting signal is generated in combination with the temperature error value; based on the PID adjusting signal, a water pump target frequency instruction is obtained in combination with the cooling load predicted value; based on the water pump target frequency instruction, a valve target opening instruction is obtained in combination with the actually-measured water flow; based on the valve target opening instruction, the fault alarm and emergency instruction is generated, the preset emergency measures are executed, the PID control parameters are updated, the performance evaluation report and the feedback information are obtained, and the response speed, the operation stability and the autonomous decision-making capacity of the water cooling system under the dynamic load and abnormal working conditions are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and in particular to a water-cooling control method, system, device and medium. Background Technology

[0002] In recent years, with the rapid development of industrial automation and intelligent control technologies, water-cooling control systems have been widely used in various industrial cooling needs. Traditional water-cooling control systems mainly rely on PID (proportional-integral-derivative) controllers to regulate the working state of water pumps and valves to maintain system temperature stability. However, with the expansion of industrial production scale and the increase in technological complexity, the traditional PID control method has gradually shown its limitations. On the one hand, due to the lack of predictive ability for future load changes, it can often only passively adjust the current temperature deviation, resulting in slow response speed and high energy consumption. On the other hand, when faced with sudden load changes or equipment failures, traditional PID controllers are difficult to adapt quickly and make effective adjustments, thus affecting the overall operating efficiency and stability of the system.

[0003] To overcome these problems, researchers began to explore new methods that combine machine learning algorithms with traditional PID control, aiming to improve the system's adaptability and intelligence. Although existing research has improved the above problems to some extent, there are still some shortcomings. For example, although some machine learning-based prediction models can provide relatively accurate load prediction results, in practical applications, these models often require a large amount of historical data for training, and their generalization ability under different operating conditions is limited, failing to meet the requirements of real-time performance and dynamism.

[0004] Existing technologies also have certain shortcomings in handling abnormal situations and fault detection, such as failing to detect problems such as jamming of actuators or sensor failures in a timely manner. This results in the system not being able to quickly switch to a safe mode when a fault occurs, which in turn affects the safe operation of the entire production line. Therefore, how to design an efficient water-cooling control system that can accurately predict future load changes and quickly respond to abnormal situations has become an important research direction. Summary of the Invention

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides a water cooling control method, system, device and medium to solve the problems of slow response speed, high energy consumption and untimely fault detection that are common in existing technologies when dealing with complex industrial cooling needs.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a water cooling control method, comprising: Acquire and preprocess device status data; Based on the preprocessed data, an LSTM-based heat load trend prediction model is constructed to obtain the predicted cooling load value. Based on the predicted cooling load and the temperature error value, a PID control signal is generated. Based on the PID control signal and combined with the predicted cooling load, the target frequency command for the water pump is obtained. Based on the target frequency command of the water pump and the measured water flow rate, the target valve opening command is obtained; Based on the valve target opening command, fault alarms and emergency commands are generated, PID control parameters are updated, and performance evaluation reports and feedback information are obtained.

[0008] As a preferred embodiment of the water-cooling control method described in this invention, the method includes: acquiring and preprocessing equipment status data, constructing a heat load trend prediction model based on LSTM, and obtaining predicted cooling load values, including: The collected equipment status data includes: supply water temperature, return water temperature, water flow rate, and start / stop signals of terminal cooling equipment; The equipment status data is preprocessed by outlier removal, missing value imputation, and normalization to calculate the current actual cooling load. Based on the preprocessed data, historical data is used as a sliding window to train an LSTM model to learn the load change pattern and obtain the predicted cooling load value.

[0009] As a preferred embodiment of the water-cooling control method described in this invention, a PID control signal is generated based on the predicted cooling load and the temperature deviation, including: The temperature error is obtained by comparing the return water temperature with the target set temperature, and then input into the PID controller for feedback adjustment. At the same time, the predicted cooling load value is introduced as a feedforward compensation factor. Based on feedback regulation, when the predicted load change exceeds the set threshold, the PID proportional coefficient is adjusted to generate a PID regulation signal.

[0010] As a preferred embodiment of the water cooling control method described in this invention, the method includes: obtaining a target frequency command for the water pump based on a PID control signal and a predicted cooling load value, comprising: Based on the PID control signal, the predicted cooling load value is compared with the load threshold, and the measured temperature is compared with the temperature threshold. At the same time, a smooth transition mechanism is introduced to generate three pump control strategies, including energy-saving mode, fast response mode, and abnormal operating condition mode, and obtain the target pump frequency command after comprehensive judgment.

[0011] As a preferred embodiment of the water-cooling control method described in this invention, the target valve opening command is obtained based on the target pump frequency command and the measured water flow rate, including: By reading the valve's current position feedback signal and combining it with the water pump's target frequency command, the valve opening is adjusted in a coordinated manner, including: simultaneously opening the valve when entering the rapid response mode, and simultaneously adjusting the valve in the energy-saving mode to stabilize the pipeline pressure difference. Meanwhile, if the actual water flow does not match the expected flow, a closed-loop compensation mechanism is activated to correct the valve opening, thereby obtaining a valve target opening command that matches the pump's operating status.

[0012] As a preferred embodiment of the water cooling control method described in this invention, the following is included: generating fault alarm and emergency commands based on the valve target opening command, comprising: Fault diagnosis is performed by comparing the target valve opening command with the actual valve opening position, combined with pump operating frequency, start / stop status, flow rate, and temperature monitoring data. If the valve does not respond, it is determined that the valve actuator is stuck. If the water pump operates at full frequency but the water flow rate is lower than the theoretical water delivery capacity, it indicates that the pipe is blocked. If the temperature continues to deviate from the set value and the command is saturated but the overall device does not respond, it is determined that the sensor is faulty. Based on the results of any fault diagnosis, an emergency command is generated and preset emergency measures are executed.

[0013] As a preferred embodiment of the water-cooling control method described in this invention, the following steps are taken: based on generating fault alarms and emergency commands, the PID control parameters are updated to obtain a performance evaluation report and feedback information, including: Based on the current fault alarm and emergency command status, the PID parameters are switched to the preset emergency parameter group, and after the operating conditions stabilize, they are restored to the normal set value through linear decay. After the operation cycle is completed, the PID parameter adjustment process, temperature control accuracy at each stage, actual operation of water pumps and valves, and data from the time of failure and handling method are summarized and recorded to build an operation performance profile. Through statistical analysis, evaluation results are obtained and optimization suggestions are generated to achieve a closed-loop iteration from fault response to strategy optimization.

[0014] In a second aspect, the present invention provides a water-cooling control system, comprising: The acquisition module is used to acquire and preprocess device status data; The load forecasting module is used to build an LSTM-based heat load trend forecasting model based on preprocessed data to obtain the cooling load forecast value. The first control module is used to generate a PID control signal based on the predicted cooling load value and the temperature error value. The second control module is used to obtain the target frequency command of the water pump based on the PID adjustment signal and the predicted value of the cooling load. The first execution module is used to obtain the target valve opening command based on the target frequency command of the water pump and the measured water flow rate. The second execution module is used to generate fault alarms and emergency commands based on the valve target opening command, update PID control parameters, and obtain performance evaluation reports and feedback information.

[0015] Thirdly, the present invention provides an electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the water-cooling control method.

[0016] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the water-cooling control method.

[0017] Compared with existing technologies, the beneficial effects of this invention are as follows: By constructing a heat load trend prediction model based on an LSTM neural network, this invention effectively overcomes the shortcomings of traditional water-cooling control systems that rely solely on real-time temperature deviations for passive adjustment, resulting in delayed response and high energy consumption, thus achieving forward-looking prediction of cooling load changes. Combined with the prediction results, the PID controller undergoes feedforward compensation and parameter adaptive adjustment, significantly improving the dynamic response speed and control accuracy under sudden load changes. Furthermore, the invention introduces a pump-valve collaborative control strategy and a real-time fault diagnosis mechanism based on command-feedback deviation, solving the problems of difficulty in timely identification of actuator anomalies and untimely emergency response in existing technologies, ensuring the safety and continuity of system operation. Through closed-loop performance evaluation and feedback optimization mechanisms, the prediction model and control strategy are continuously improved, overcoming the limitations of traditional methods in adapting to long-term operating condition changes, and achieving a synergistic improvement in control accuracy, energy efficiency, and reliability. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the overall process of a water cooling control method according to an embodiment of the present invention.

[0020] Figure 2This is a flowchart illustrating the fault diagnosis and parameter optimization process of the water cooling control method according to an embodiment of the present invention. Detailed Implementation

[0021] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0022] Example 1, referring to Figure 1 As an embodiment of the present invention, a water cooling control method is provided, comprising: S100: Acquire and preprocess device status data; S200: Based on the preprocessed data, construct an LSTM-based heat load trend prediction model to obtain the predicted cooling load value; S300: Generates a PID control signal based on the predicted cooling load and the temperature error value; S400: Based on the PID control signal and combined with the predicted cooling load, the target frequency command for the water pump is obtained; S500: Based on the target frequency command of the water pump and combined with the measured water flow rate, the target valve opening command is obtained; S600: Based on the valve target opening command, generate fault alarms and emergency commands, update PID control parameters, and obtain performance evaluation reports and feedback information.

[0023] It should be noted that in water-cooled control systems, cooling pipes are core components. During operation, parameters such as temperature, pressure, and medium flow rate fluctuate significantly inside and outside the system, leading to substantial changes in equipment temperature distribution and stress gradient. Factors such as cooling water corrosion, scaling, micro-leakage, and water quality fluctuations can cause non-uniform changes in equipment wall thickness or structure. Furthermore, operating cooling systems often operate in corrosive water environments, under temperature gradient fields, and with fluctuating pressure, making real-time and accurate measurement of the temperature and stress fields difficult, and safety warning responses often exhibit delays. Simultaneously, the continuous scouring of cooling water and the synergistic effect of corrosive media can easily cause localized thinning of equipment walls, surface damage, or the propagation of internal microcracks, thereby jeopardizing system health and leading to equipment performance degradation. Therefore, real-time monitoring of the health status of water-cooled control systems is particularly important.

[0024] Therefore, to address the aforementioned operational monitoring and health prediction issues, the following steps (S100-S600) are used to construct an LSTM-based heat load trend prediction model, obtain the cooling load prediction value, calculate the temperature deviation and combine it with the cooling load prediction value to generate a PID control signal, obtain the water pump target frequency command, generate the valve target opening command, generate fault alarm information and emergency control commands, update the PID control parameters, and use the updated PID control parameters to generate a performance evaluation report and feedback information.

[0025] Example 2, refer to Figure 1 As an embodiment of the present invention, a water cooling control method is provided based on the above embodiment.

[0026] In this embodiment of the application, steps S100-S200 involve acquiring and preprocessing equipment status data, constructing a heat load trend prediction model based on LSTM, and obtaining predicted cooling load values, including the following steps A1-A3: A1: Collect equipment status data and preprocess it to calculate the current actual cooling load; Specifically, temperature sensors WT1 and WT2 installed on the water supply and return pipes are used to continuously acquire the current water temperature value. The real-time water flow rate measured by the flow sensors installed on the main water supply and return pipes of the water cooling system is read. Combined with the start and stop operation signals of various end-user cooling equipment, such as air conditioning units and process cooling units, as the basic information for determining the source of the current cooling load.

[0027] The collected signals of supply water temperature, return water temperature, water flow rate, and start / stop status of terminal cooling equipment are preprocessed, including outlier removal, missing value imputation, and normalization. A time series indexed by time is generated and stored. Specifically, data from temperature sensors, flow sensors, and terminal equipment status signals are synchronously collected at a unified timestamp (e.g., once per second). Each collection of four data points is treated as a single data point and stored chronologically in the database, forming a time-indexed sequence structure, such as timestamp → [T supply water, T return water, flow rate, equipment 1 status, equipment 2 status, ...]. This sequence is used to calculate the actual heat exchange at each moment, providing an accurate data foundation for load analysis. By multiplying the water flow rate by the specific heat capacity of water, and then multiplying by the difference between the return water temperature and the supply water temperature, the instantaneous heat exchange capacity of the current system, i.e., the current actual cooling load, is obtained.

[0028] In an optional implementation, the data preprocessing in step S100 can also be performed using advanced anomaly detection, which can not only eliminate jumps, but also introduce the intelligent anomaly detection algorithm of isolated forest to dynamically identify and handle complex situations such as sensor drift and periodic anomalies.

[0029] In another alternative implementation, the data preprocessing in step S100 can also be carried out through dynamic feature engineering. In addition to the original data, more predictive derived features can be constructed, such as quantifying the start-stop status of the terminal device into instantaneous total power or equivalent load factor as a strong feature output of the model.

[0030] A2: Based on the preprocessed data, historical data is used as a sliding window to train the LSTM model to learn the load change pattern and obtain the predicted value of cooling load. Specifically, a Long Short-Term Memory (LSTM) neural network model is trained using historical data, such as high-frequency data collected over 30 consecutive days, covering different operating conditions, including day and night, high and low loads, and start-up and shutdown processes, to ensure the model has good generalization ability. During training, a supervised learning approach is adopted, dividing the historical data into training and validation sets. The mean squared error (MSE) is used as the loss function, and the error between the predicted and true values ​​is minimized through the backpropagation algorithm and the Adam optimizer. The LSTM network weights are iteratively updated until the model's performance converges on the validation set, resulting in a cold load prediction model that can be deployed online. The model outputs a predicted value for the cold load change trend in the next 10 to 60 seconds, providing a forward-looking basis for subsequent control actions.

[0031] It should be noted that by using LSTM to establish a heat load trend prediction model, after the model is trained, it can output the predicted value of future cooling load change trend in real time, realizing the transformation from "passive response" to "active prediction". This effectively overcomes the lag caused by traditional PID control relying solely on the current error for adjustment, significantly improves the system's adaptability to dynamic load changes, and provides a reliable forward-looking decision-making basis for subsequent precise and energy-saving pump and valve coordinated control.

[0032] In an optional implementation, the LSTM model prediction in step S200 can also be a CNN-LSTM model, that is, a one-dimensional convolutional layer is added before LSTM. First, CNN is used to extract local short-term features of the data within the sliding window, and then LSTM is used to capture long-term dependencies, thereby improving the ability to capture mutation loads.

[0033] In another alternative implementation, the LSTM model prediction in step S200 can also introduce an attention mechanism after the LSTM layer, allowing the model to automatically focus on the most critical moments in the historical sequence for predicting future loads, such as the moment when a high-power terminal device starts up, thereby improving the interpretability and prediction accuracy of the model.

[0034] In this embodiment of the application, step S300 generates a PID control signal based on the predicted cooling load value and the temperature deviation, including the following steps B1-B2: B1: The temperature error is obtained by comparing the return water temperature with the target set temperature, and then input into the PID controller for feedback adjustment. At the same time, the predicted cooling load value is introduced as a feedforward compensation factor. Specifically, the target set temperature issued by the central control module DB27 in the distributed control system (DCS) according to the process requirements standard is used as a benchmark. It is compared with the actual water temperature fed back by the return water temperature sensor WT2 to calculate the current temperature error value. The error value is sent as the core input to the PID control logic. The proportional part responds to the current deviation, the integral part eliminates long-term residual error, and the derivative part suppresses overshoot in the control process to ensure effective suppression of the existing deviation. Meanwhile, the cold load forecast value is introduced as a feedforward compensation factor to realize a composite control mechanism that combines feedback and prediction: the difference between the future cold load forecast value output by the LSTM model and the current actual cold load is multiplied by the pre-calibrated feedforward gain coefficient Kf to obtain the feedforward compensation factor, where Kf is determined by regression of historical data.

[0035] B2: Based on feedback regulation, when the predicted load change exceeds the set threshold, the PID proportional coefficient is adjusted to generate a PID regulation signal.

[0036] Specifically, a threshold A is set by combining system identification with historical data statistical analysis. When the predicted increase in future load value compared to the current load value exceeds the preset threshold A, the proportional coefficient Kp in the PID controller is automatically adjusted to make the adjustment action more rapid. If the predicted load value is lower than the preset threshold A, it is determined that the load tends to stabilize or decrease, and the normal adjustment intensity is maintained to avoid overreaction. Finally, a PID pump speed adjustment signal that integrates feedback control and trend prediction is generated as the basis for the next stage of decision-making.

[0037] It should be noted that by dynamically adjusting the proportional gain of the PID controller based on the load forecast results, this strategy effectively overcomes the inherent defects of traditional PID control, which relies solely on the current error and lacks foresight. It significantly improves the control accuracy and stability of the system under varying operating conditions and achieves synergistic optimization of rapid response and smooth adjustment.

[0038] In this embodiment of the application, step S400, based on the PID control signal and combined with the predicted cooling load value, obtains the target frequency command for the water pump, including: Based on the PID control signal, the predicted cooling load value is compared with the load threshold, and the measured temperature is compared with the temperature threshold. At the same time, a smooth transition mechanism is introduced to generate three pump control strategies, including energy-saving mode, fast response mode and abnormal working condition mode, and the target frequency command of the water pump after comprehensive judgment is obtained. Specifically, based on the PID control signal, the threshold K is set using statistical analysis of historical operating data, and the threshold D is set using an adaptive threshold algorithm. If the predicted cooling load is lower than the preset threshold D and the temperature is lower than the preset threshold K, it indicates that the current cooling demand is stable. In this case, the energy-saving mode is activated, and the water pump operating frequency is actively reduced to a safe level that maintains the minimum circulation flow, thereby reducing power consumption. If the predicted cooling load is greater than the preset threshold D and the temperature is greater than the preset threshold K, it is determined to be a sudden high load scenario. It immediately switches to the fast response mode, skips the gradual adjustment process, and directly outputs higher frequency commands to make the water pump quickly increase its output to meet the upcoming cooling demand. If the predicted cooling load is lower than the preset load threshold D and the temperature is higher than the preset threshold K, or if the predicted cooling load is higher than the preset threshold D and the temperature is lower than the preset threshold K, it is determined to be an abnormal operating condition, triggering the fault diagnosis mechanism, pausing the energy-saving mode, maintaining the current water pump frequency and starting deviation verification, avoiding misjudging as a stable state and entering inefficient regulation, relying on PID output to continuously adjust the pump speed to achieve stable temperature control.

[0039] Furthermore, a smooth transition mechanism is set between the three modes. The current operating mode is determined by the "if-else" condition, which significantly reduces ineffective energy consumption and overcomes the energy waste caused by the continuous high-frequency operation of the water pump in traditional control. At the same time, corresponding control commands are output to prevent hydraulic shock or temperature fluctuation caused by sudden changes in control commands. Finally, the target operating frequency command of the water pump after comprehensive judgment is output.

[0040] It should be noted that when a sudden increase in load and rapid expansion of error are predicted, the system immediately switches to a fast response mode, skipping the conventional gradual adjustment process and directly outputting high-frequency commands to rapidly increase the pump's output, effectively avoiding the risk of temperature exceeding limits due to response lag. These three strategies overcome the limitations of the traditional single PID control mode in terms of poor adaptability, achieving a balance between energy saving, response speed, and operational safety, and significantly improving the intelligence level and practical application efficiency of the pump speed control.

[0041] In an alternative implementation, the thresholds K and D set in step S400 can also be context-aware dynamic thresholds, allowing the thresholds to be automatically adjusted according to the macroscopic state of the system. For example, the thresholds D and K are no longer fixed values, but functions or lookup tables related to factors such as ambient temperature, seasonal patterns, and weekdays / holidays.

[0042] In another optional implementation, the smooth transition mechanism in step S400 can also be a trend-based predictive switching. For example, even if the current value does not exceed the threshold, if the LSTM predicts that the load or temperature will change at a very high rate within the next 30 seconds and will inevitably exceed the threshold, the system can enter a "rapid response" or "abnormal alert" state in advance. This extends proactive control from load forecasting to the mode switching itself.

[0043] In this embodiment of the application, step S500, based on the target frequency command of the water pump and combined with the measured water flow rate, obtains the target valve opening command, including: Specifically, the system reads the feedback signal of the valve's current position. When the water pump speeds up to cope with high load, the regulating valve is opened simultaneously to reduce water flow resistance, improve water delivery efficiency, and reduce the burden on the water pump. Under low-load energy-saving operation, the valve is adjusted to maintain stable pipeline pressure difference and prevent uneven water flow distribution, while ensuring the minimum circulation flow.

[0044] Simultaneously, flow sensor data is used for verification. If the actual flow rate does not match the expected flow rate (where the expected flow rate refers to the theoretical water delivery calculated based on the current target operating frequency of the pump and the initial valve opening, combined with the pump performance curve and pipeline resistance characteristic curve), a closed-loop compensation mechanism is activated. The deviation is used as input, and a fine-tuning command is generated through a low-gain PID or proportional control algorithm to dynamically correct the valve opening. To prevent over-adjustment and ensure precise and controllable hydraulic balance, the fine-tuning range is limited to ±10% of the current valve opening, and the single adjustment amplitude does not exceed 5%. At the same time, a flow deviation dead zone (such as ±3%) is set, and adjustment is only triggered when the deviation exceeds the dead zone to ensure hydraulic balance. Finally, a valve target opening command matching the pump operating status is output to optimize the overall operating efficiency.

[0045] It should be noted that this pump-valve coordination strategy overcomes the blind spots and energy waste caused by adjusting only a single device such as a pump or valve in traditional control, and significantly improves the overall operating efficiency, response accuracy and reliability of the system.

[0046] In this embodiment of the application, step S600 generates a fault alarm and emergency command based on the valve target opening command, updates the PID control parameters, and obtains a performance evaluation report and feedback information, including the following steps C1-C4: C1: By comparing the target valve opening command with the actual valve opening position and combining the collected data, fault diagnosis is performed. Specifically, upon receiving the valve target opening command, it compares it in real time with the actual opening position reported by the valve. Combined with the actual operating frequency, start / stop status, and flow and temperature monitoring data of the water pump, a comprehensive judgment is made as to whether the actuator is responding normally to the control command. If the valve is found to be unable to adjust to the position as instructed, such as when the instruction has been issued but the position has not changed, it is determined that the valve actuator is stuck or lacks power.

[0047] Furthermore, the current pump frequency is compared with the target operating frequency command for the pump: If the water pump is running at full load but the flow rate is still far below the theoretical value expected based on the pump frequency and valve opening, where the theoretical value of the valve opening is determined by the water pump performance curve, and the valve opening required to meet the flow rate is calculated using the pressure difference-flow-opening relationship model, valve Cv value and fluid mechanics formula as the expected value, then it indicates that the pipeline is blocked or the pump is faulty. If the temperature continues to deviate from the set value and control commands are continuously output, but the cooling water circulation control device, which consists of a water pump, valves, sensors, and controller, does not respond, it indicates that the sensor or control link has failed. C2: Based on the result of any fault diagnosis, generate emergency instructions and execute preset emergency measures.

[0048] Specifically, if any abnormal situation occurs, the alarm mechanism will be triggered immediately, sending the fault type and location information to the monitoring personnel, and automatically executing the preset emergency measures. For example, if it is determined that the valve actuator is stuck, the backup pump will be started automatically and the key regulating valve of the branch where the faulty valve is located will be fully opened, switching to bypass operation mode. If it is detected that the water pump is running at full load but the flow rate is far below the theoretical value (indicating pipe blockage or pump failure), the current water pump will be stopped immediately, the backup pump will be started, and the main pipeline regulating valve will be fully opened to maximize the flow capacity. If the temperature continues to deviate from the set value and the control command has been saturated but the system does not respond, it is determined that the sensor or control link has failed. Then, it will automatically switch to a fixed high-frequency operation mode based on timing logic, such as the water pump running at 75% frequency, while all key valves will be adjusted to the fully open state, switching to safe constant flow operation condition to ensure basic cooling capacity.

[0049] It should be noted that upon confirmation of a fault, an automatic emergency response is immediately triggered. Safety strategies such as starting the backup pump, switching to a fixed high-frequency operating mode, and fully opening key valves can be implemented without manual intervention, ensuring uninterrupted cooling water circulation and effectively preventing equipment damage due to overheating.

[0050] C3: Based on the current fault alarm and emergency command status, the PID parameters are switched to the preset emergency parameter group, and after the operating conditions stabilize, they are restored to the normal set value through linear decay. Specifically, when entering emergency mode, the PID parameters immediately switch to the pre-calibrated "emergency parameter group" to adapt to high flow and low resistance conditions.

[0051] After the system stabilizes, the emergency parameters are gradually restored from their emergency values ​​to the standard values ​​under normal operating conditions using a linear or exponential decay method, achieving a smooth transition. During the initial equipment switchover, the proportional gain is proactively reduced to suppress shocks and overshoot caused by sudden changes in control inputs, preventing secondary disturbances to the system. After the emergency mode is activated, the stability of temperature control is continuously monitored. If the deviation change rate is less than ±0.2℃ / min for 60 seconds, confirming system stability, the PID parameters are linearly restored from the emergency values ​​to the normal setpoints by increasing the proportional gain Kp by 5% every 30 seconds. The entire restoration process lasts at least 5 minutes to prevent secondary disturbances caused by sudden parameter changes. Simultaneously, the control effect is continuously evaluated by combining real-time temperature deviation, pump frequency response characteristics, and cooling load prediction trends. If oscillations or slow responses are detected during the adjustment process, the parameter combination is dynamically corrected. This adaptive adjustment mechanism ensures that the controller maintains good adjustment performance under both normal and fault recovery conditions, providing stable support for restoring normal operation.

[0052] C4: After completing the operation cycle, summarize the records and build an operation performance profile. Statistical analysis is used to obtain evaluation results and generate optimization suggestions, realizing a closed-loop iteration from fault response to strategy optimization.

[0053] Specifically, for load change conditions, the mean square error (MSE) between the predicted value and the actual load and the response delay time are calculated to assess prediction timeliness. For fault switching processes, the overshoot, settling time and number of oscillations after PID parameter switching are obtained to determine adaptive stability. For energy-saving modes, the standard deviation of temperature fluctuations and energy consumption reduction during operation are statistically analyzed to assess stability and energy-saving effects. A weighted performance score is generated by combining various indicators to compare the effectiveness of different strategies horizontally.

[0054] Based on the evaluation results, optimization suggestions are generated, such as adjusting the training frequency of the LSTM model, correcting the switching threshold of the multimodal control mode, and optimizing the initial setpoints of PID parameters under emergency conditions. The optimization suggestions are then fed back to the long short-term memory neural network prediction model and control logic. Specifically, after manual confirmation or automatic approval, the optimization suggestions are converted into executable instructions. For example, the adjusted LSTM model is retrained and written into the training scheduling module, the corrected mode switching thresholds such as the load change rate threshold and temperature error threshold are written into the control logic configuration table, and the optimized emergency PID initial parameters are written into the parameter library. The updated parameters are automatically loaded in the next cycle of operation, realizing a closed-loop iteration from fault response to long-term strategy optimization, and continuously improving the overall stability and energy efficiency of operation.

[0055] For example, such as Figure 2The diagram shows the fault diagnosis and parameter optimization process. The valve target opening command is compared with the actual valve opening position. If the command and feedback are inconsistent, a fault alarm is triggered. If they are consistent, normal operation continues. Finally, the operating data is recorded, a performance report is generated, and feedback is provided to the optimization model and control logic.

[0056] In summary, this invention effectively overcomes the shortcomings of traditional water-cooling control systems that rely solely on real-time temperature deviations for passive adjustment, resulting in delayed response and high energy consumption. It achieves forward-looking prediction of cooling load changes by constructing a heat load trend prediction model based on an LSTM neural network. Combined with the prediction results, feedforward compensation and adaptive parameter adjustment are applied to the PID controller, significantly improving dynamic response speed and control accuracy under sudden load changes. Furthermore, the introduction of a pump-valve collaborative control strategy and a real-time fault diagnosis mechanism based on command-feedback deviation solves the problems of difficulty in timely identification of actuator anomalies and delayed emergency response in existing technologies, ensuring the safety and continuity of system operation. Finally, through closed-loop performance evaluation and feedback optimization mechanisms, the prediction model and control strategy are continuously improved, overcoming the limitations of traditional methods in adapting to long-term operating condition changes, and achieving a synergistic improvement in control accuracy, energy efficiency, and reliability.

[0057] Example 3 illustrates a water-cooling control method. It should be noted that the technical solution of this water-cooling control system belongs to the same concept as the water-cooling control method described above. Details not described in detail in this example can be found in the description of the water-cooling control method described above.

[0058] This embodiment also provides a water-cooling control system, including: The acquisition module is used to acquire and preprocess device status data; The load forecasting module is used to build an LSTM-based heat load trend forecasting model based on preprocessed data to obtain the cooling load forecast value. The first control module is used to generate a PID control signal based on the predicted cooling load value and the temperature error value. The second control module is used to obtain the target frequency command of the water pump based on the PID adjustment signal and the predicted value of the cooling load. The first execution module is used to obtain the target valve opening command based on the target frequency command of the water pump and the measured water flow rate. The second execution module is used to generate fault alarms and emergency commands based on the valve target opening command, update PID control parameters, and obtain performance evaluation reports and feedback information.

[0059] This embodiment also provides an electronic device suitable for water-cooling control, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the water-cooling control method proposed in the above embodiment.

[0060] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the water cooling control method proposed in the above embodiments.

[0061] The storage medium proposed in this embodiment and the water-cooling control method proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0062] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. 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 can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0063] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A water-cooling control method, characterized in that, include: Acquire and preprocess device status data; Based on the preprocessed data, an LSTM-based heat load trend prediction model is constructed to obtain the predicted cooling load value. Based on the predicted cooling load and the temperature error value, a PID control signal is generated. Based on the PID control signal and combined with the predicted cooling load, the target frequency command for the water pump is obtained. Based on the target frequency command of the water pump and the measured water flow rate, the target valve opening command is obtained; Based on the valve target opening command, fault alarms and emergency commands are generated, PID control parameters are updated, and performance evaluation reports and feedback information are obtained.

2. The water-cooling control method as described in claim 1, characterized in that, Acquire and preprocess equipment status data, construct an LSTM-based heat load trend prediction model, and obtain the predicted cooling load values, including: The collected equipment status data includes: supply water temperature, return water temperature, water flow rate, and start / stop signals of terminal cooling equipment; The equipment status data is preprocessed by outlier removal, missing value imputation, and normalization to calculate the current actual cooling load. Based on the preprocessed data, historical data is used as a sliding window to train an LSTM model to learn the load change pattern and obtain the predicted cooling load value.

3. The water-cooling control method as described in claim 2, characterized in that, Based on the predicted cooling load and temperature deviation, a PID control signal is generated, including: The temperature error is obtained by comparing the return water temperature with the target set temperature, and then input into the PID controller for feedback adjustment. At the same time, the predicted cooling load value is introduced as a feedforward compensation factor. Based on feedback regulation, when the predicted load change exceeds the set threshold, the PID proportional coefficient is adjusted to generate a PID regulation signal.

4. The water-cooling control method as described in claim 3, characterized in that, Based on the PID control signal and combined with the predicted cooling load, the target frequency command for the water pump is obtained, including: Based on the PID control signal, the predicted cooling load value is compared with the load threshold, and the measured temperature is compared with the temperature threshold. At the same time, a smooth transition mechanism is introduced to generate three pump control strategies, including energy-saving mode, fast response mode, and abnormal operating condition mode, and obtain the target pump frequency command after comprehensive judgment.

5. The water-cooling control method as described in claim 4, characterized in that, Based on the target pump frequency command and the measured water flow rate, the target valve opening command is obtained, including: By reading the valve's current position feedback signal and combining it with the water pump's target frequency command, the valve opening is adjusted in a coordinated manner, including: simultaneously opening the valve when entering the rapid response mode, and simultaneously adjusting the valve in the energy-saving mode to stabilize the pipeline pressure difference. Meanwhile, if the actual water flow does not match the expected flow, a closed-loop compensation mechanism is activated to correct the valve opening, thereby obtaining a valve target opening command that matches the pump's operating status.

6. The water-cooling control method as described in claim 5, characterized in that, Based on the valve target opening command, generate fault alarms and emergency commands, including: Fault diagnosis is performed by comparing the target valve opening command with the actual valve opening position, combined with pump operating frequency, start / stop status, flow rate, and temperature monitoring data. If the valve does not respond, it is determined that the valve actuator is stuck. If the water pump operates at full frequency but the water flow rate is lower than the theoretical water delivery capacity, it indicates that the pipe is blocked. If the temperature continues to deviate from the set value and the command is saturated but the overall device does not respond, it is determined that the sensor is faulty. Based on the results of any fault diagnosis, an emergency command is generated and preset emergency measures are executed.

7. The water-cooling control method as described in claim 6, characterized in that, Based on the generated fault alarms and emergency commands, the PID control parameters are updated, and a performance evaluation report and feedback information are obtained, including: Based on the current fault alarm and emergency command status, the PID parameters are switched to the preset emergency parameter group, and after the operating conditions stabilize, they are restored to the normal set value through linear decay. After the operation cycle is completed, the PID parameter adjustment process, temperature control accuracy at each stage, actual operation of water pumps and valves, and data from the time of failure and handling method are summarized and recorded to build an operation performance profile. Through statistical analysis, evaluation results are obtained and optimization suggestions are generated to achieve a closed-loop iteration from fault response to strategy optimization.

8. A water-cooling control system, employing the method as described in any one of claims 1-7, characterized in that, include: The acquisition module is used to acquire and preprocess device status data; The load forecasting module is used to build an LSTM-based heat load trend forecasting model based on preprocessed data to obtain the cooling load forecast value. The first control module is used to generate a PID control signal based on the predicted cooling load value and the temperature error value. The second control module is used to obtain the target frequency command of the water pump based on the PID adjustment signal and the predicted value of the cooling load. The first execution module is used to obtain the target valve opening command based on the target frequency command of the water pump and the measured water flow rate. The second execution module is used to generate fault alarms and emergency commands based on the valve target opening command, update PID control parameters, and obtain performance evaluation reports and feedback information.

9. An electronic device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the water-cooling control method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It includes storing computer-executable instructions that, when executed by a processor, implement the steps of the water-cooling control method according to any one of claims 1 to 7.

Images