Intelligent start-stop strategy of water cooling unit in air cooling main and water cooling backup cooling system

Abstract

This invention relates to an intelligent start-stop strategy for water-cooled units in a wind-cooled main-water-cooled backup cooling system, belonging to the field of power system equipment temperature control and intelligent operation and maintenance management technology. The system employs a self-growing sensing network to achieve multi-point dynamic high-density monitoring and generates a high-confidence environmental profile through proactive data integration. Utilizing a multi-stage progressive temperature control trend model combined with adaptive influencing factor adjustment, it accurately predicts temperature rise sensitive areas and overall system trends. The strategy introduces a dynamic safety margin threshold mechanism, adaptively setting alarm and redundancy intervention thresholds based on real-time temperature control prediction results to ensure system safety. A three-dimensional self-balancing decision-maker dynamically coordinates energy consumption, safety, and equipment health, optimizing the start-stop timing and combination schemes of the water-cooled units. The adaptive slow-control start-stop process effectively suppresses water hammer and icing risks through multi-stage adjustment of fluid flow and pressure; combined with fault scheduling, it achieves seamless transfer of cooling load and continuous cooling assurance.

Description

Technical Field

[0001] This invention belongs to the field of power system equipment temperature control and intelligent operation and maintenance management technology, specifically relating to an intelligent start-stop strategy for water-cooled units in an air-cooled main water-cooled backup cooling system. Background Technology

[0002] Air-cooled primary and water-cooled backup cooling systems are commonly used thermal management solutions for large-scale converter stations, data centers, and large electrical equipment. This system uses air cooling as the primary cooling method, prioritizing operation when equipment load is light or the operating environment is suitable. Its advantages include simple structure, convenient maintenance, and low energy consumption, maintaining daily thermal balance. However, when external temperatures rise, equipment load increases sharply, or the air-cooling system partially fails, it automatically or manually switches to water cooling or operates in conjunction with water cooling. The water-cooled unit, with its strong cooling capacity, ensures that the equipment does not overheat, guaranteeing safe system operation. However, in practical applications, the frequent start-stop cycles, delayed response, and manual switching of water-cooled systems still present several problems.

[0003] In traditional systems, the start-up and shutdown rules for water-cooled units are mostly based on fixed temperature thresholds or simple air-cooling failure judgments. These configurations are simplistic and lack data support, making it difficult to intelligently respond to actual equipment heat loads, environmental changes, and operational goals. Temperature control lag and frequent switching often lead to reduced system energy efficiency, increased wear and tear on water-cooled units, shortened equipment lifespan, and significantly increased energy consumption fluctuations and maintenance pressures. Furthermore, existing solutions rarely consider intelligent factors such as fault prediction, environmental adaptability, and optimal cooling allocation, failing to achieve a comprehensive balance of the system's "safety, economy, and efficiency" objectives. While some research in recent years has proposed optimizing the start-up and shutdown of cooling equipment, most focuses on single cooling units or data center scenarios. For multi-unit, highly dynamically coupled systems with air-cooled primary and water-cooled backup systems, there is a lack of practical and highly adaptive scheduling strategies.

[0004] Therefore, designing precise, intelligent, and robust start-up and shutdown strategies for water-cooled units based on the actual operating status of the equipment, environmental change patterns, and energy consumption costs is key to improving the overall efficiency and reliability of air-cooled main and water-cooled backup systems. Summary of the Invention

[0005] This invention addresses the pain points of air-cooled main and water-cooled backup systems, such as slow response, unintelligent start-up and shutdown, high energy consumption, and insufficient system reliability. It proposes a start-up and shutdown strategy for water-cooled units based on dynamic data fusion and intelligent predictive optimization. The key technical problem to be solved is how to fully utilize multi-source system operating data to achieve accurate assessment of equipment heat load, external environment, and air-cooling status.

[0006] To achieve the above objectives, the present invention employs the following technical solution: the strategy includes:

[0007] A self-growing sensing network and data integration are constructed, and a self-growing sensing network activation algorithm is adopted to dynamically deploy and activate monitoring nodes according to changes in equipment operating heat load, so as to achieve multi-point high-density sensing.

[0008] Multi-stage progressive temperature control trend prediction: Using a multi-stage progressive temperature control trend extrapolation model, first, through short-term local fluctuation analysis, potential temperature rise sensitive areas are captured, and then gradually extended to long-term prediction of the entire system.

[0009] Dynamic safety margin adaptive judgment is based on progressive temperature control trend results. A dynamic safety margin threshold judgment mechanism is introduced to automatically adjust the single / multi-level alarm and redundant intervention thresholds in real time according to near-future trends, rather than fixed thresholds.

[0010] Multi-objective self-balancing start-up and shutdown decision-making: Construct a three-dimensional trade-off self-balancing decision-maker, combine the safety margin range, equipment health status and energy consumption assessment results output in step three, realize automatic transfer of multi-objective weights—dynamic inter-adjustment of energy consumption, safety and equipment health, and calculate the optimal start-up and shutdown timing and the required combination of water-cooling units;

[0011] Adaptive start-stop and fault scheduling: Based on intelligent decision-making instructions, the adaptive start-stop program is activated, and the automatic fluid flow and pressure are adjusted in multiple steps to prevent water hammer and icing risks during start-stop.

[0012] In one scheme, the self-growing sensing network and data integration include: dynamically deploying and activating monitoring nodes according to changes in equipment operating heat load to achieve multi-point high-density sensing, and ensuring the flexibility and integrity of the monitoring network through density adaptive activation rules; data integration standardizes, filters and corrects real-time heterogeneous data of temperature, humidity and wind speed collected by each monitoring node, adopts an active data integration method based on weighted fuzzy consistency criteria to eliminate outliers and anomalies, and forms a high-confidence environmental profile matrix through spatiotemporal interpolation.

[0013] In one scheme, the multi-stage progressive temperature control trend prediction includes: a multi-stage progressive temperature control trend deduction model, which takes a high-confidence environmental profile matrix as input, firstly uses a progressive interval perturbation detection method to perform short-term series fluctuation analysis on local areas or key equipment nodes, locates temperature rise sensitive areas through a sensitivity weighting function, automatically expands the influence domain when the sensitivity continuously exceeds the adaptive threshold, and enters the medium- and long-term trend stage, and adopts a time-progressive convolution-regression joint modeling strategy to perform multi-scale modeling and weighted aggregation of the temperature control state.

[0014] In one scheme, the aforementioned dynamic safety margin adaptive judgment includes: a dynamic safety margin thresholding mechanism, which is based on the multi-stage temperature control trend prediction results and the prediction values ​​of each local node, analyzes the temperature rise critical rate, temperature prediction extreme value, and real-time performance parameters of the air-cooled unit in real time, and comprehensively forms a dynamic margin envelope interval to replace the traditional fixed temperature threshold.

[0015] By normalizing and aggregating key indicators, alarm and redundancy intervention thresholds are dynamically set. When the comprehensive threshold meets the water cooling intervention conditions, the water cooling unit preparation or startup process is automatically initiated, realizing real-time adaptive safety margin management and alarm hierarchical redundancy decision-making.

[0016] In one embodiment, the multi-objective self-balancing start-up and shutdown decision-making includes a three-dimensional trade-off self-balancing decision-maker. This decision-maker integrates the safety margin range, equipment health status, and energy consumption assessment results as decision inputs. Through a multi-objective weight self-transfer mechanism, it achieves dynamic weight inter-adjustment of energy consumption, safety, and equipment health. It comprehensively evaluates the risk, lifespan, and energy efficiency performance of each water-cooling unit combination scheme in future start-up and shutdown cycles, adaptively outputs the optimal start-up and shutdown timing and water-cooling unit configuration, and continuously optimizes the weights and decision-making strategies based on the cumulative losses and maintenance expenditures throughout the entire cycle, thereby achieving full lifecycle self-evolutionary management of start-up and shutdown behavior.

[0017] In one scheme, the adaptive slow start-stop and fault scheduling includes: after the start-stop command of the water cooling unit is issued, a multi-step progressive control method of liquid flow and pipeline pressure is adopted. The flow rate and pressure are slowly adjusted in stages through the intelligent valve control system to suppress water hammer and freezing risks. Combined with real-time adjustment of refrigerant flow rate and pipeline temperature distribution, the temperature drop rate is kept stable.

[0018] When an abnormality is detected in the air-cooled unit, the system automatically identifies the nature and scope of the abnormality, activates the redundant water-cooled unit activation plan, completes the cold load migration based on the load ratio algorithm, and allocates the target load and schedules water-cooled backup resources in real time. The entire process automatically archives data and switching trajectories to achieve seamless connection and continuous cooling guarantee.

[0019] In one scheme, the multi-stage progressive temperature control trend prediction model also includes an adaptive influence factor adjustment unit, which uses a feedback correction mechanism to update the weights of each node and model parameters in real time after each stage of calculation, so as to adapt to the thermal inertia of the equipment and changes in the external environment, and achieve sensitive and robust temperature control trend prediction.

[0020] Beneficial effects of this invention:

[0021] This invention, through a combination of data-driven and intelligent prediction methods, significantly improves the scientific rigor and foresight of the start-up and shutdown of water-cooled units in air-cooled main and water-cooled backup cooling systems. It eliminates the energy waste caused by traditional, simplistic rules, delayed responses, and unnecessary frequent start-ups and shutdowns of water cooling systems. The intelligent control algorithm ensures that backup water-cooled units switch smoothly only during necessary periods, greatly reducing overall system energy consumption and unnecessary wear and tear on equipment, and effectively extending the lifespan of water-cooled equipment.

[0022] Multi-objective optimization and health status rotation mechanisms enhance the overall fault tolerance of the cooling system and improve the safety assurance level of critical equipment. Under various adverse conditions such as complementary hot and cold environments, sudden load changes, weakened air cooling, and extreme climates, the system can maintain a highly adaptive operating strategy, significantly reducing the risk of thermal runaway. Meanwhile, the centralized data platform and automated operation and maintenance logs not only facilitate the full lifecycle management of the equipment but also provide a solid data foundation for subsequent intelligent optimization and maintenance. Attached Figure Description

[0023] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0024] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Typical embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0025] Unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. To facilitate understanding, the invention will now be described more fully with reference to the accompanying drawings. Typical embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to make the disclosure of the invention more thorough and complete.

[0026] like Figure 1 As shown, an intelligent start-stop strategy for the water-cooling unit in an air-cooled main cooling system with water-cooled backup cooling is described. It includes:

[0027] Step 1: Constructing a self-growing sensing network and integrating data

[0028] First, the system adopts a self-growing sensing network activation algorithm to dynamically deploy and activate monitoring nodes according to changes in the thermal load of the equipment, thereby achieving high-density sensing at multiple points and ensuring that the monitoring network remains flexible and complete under any condition.

[0029] Heterogeneous data (temperature, humidity, wind speed, etc.) collected by nodes are standardized, filtered, and corrected in real time through innovative proactive data integration algorithms, forming a high-confidence original environmental profile and providing a solid data foundation for trend modeling.

[0030] First, the system deploys an innovative self-growing sensing network activation algorithm. Its core idea is to adaptively and dynamically optimize the spatial distribution of monitoring nodes based on the real-time heat load distribution of the monitored equipment. Let the entire monitoring area be Ω, with N sensing nodes distributed therein, and the location of each node be Ω. The node's activation state can be indicated by a function. This indicates that 1 represents active and 0 represents dormant. Assume the heat load distribution at any point on the device is as follows: The system analyzes in real time gradient field A density-adaptive activation rule is adopted: if a certain position heat load gradient Exceeding the threshold If the node is activated, a nearby dormant node will be activated; otherwise, the node will be put into dormancy. This allows nodes to dynamically "grow" and shrink according to the heat load, ensuring the flexibility and high density of monitoring coverage.

[0031] Subsequently, each node collected data including temperature. ,humidity Wind speed The system uses heterogeneous raw data from multiple sources. To improve data reliability, an active data integration algorithm based on a weighted fuzzy consistency criterion is introduced. Specifically, the original data matrix is ​​first processed... Perform normalization processing, and then define data consistency metrics. This represents the fuzzy similarity between node i and node j on the same metric. For each dimension, a Gaussian similarity function is used:

[0032]

[0033] in, Let be the normalized value of the i-th node under the k-th sensing quantity (temperature, humidity, etc.). The standard deviation is the empirical value. The system aggregates the similarity of all node pairs and calculates the globally consistent weighted mean. If a node deviates from the global mean by more than a threshold, active correction or removal is triggered, thereby eliminating outliers and anomalies. Finally, all standardized and filtered data are used for spatiotemporal interpolation (Kriging interpolation) to form a high-confidence environmental profile matrix. This provides a solid data foundation for subsequent trend modeling and intelligent decision-making.

[0034] Step 2: Multi-stage progressive temperature control trend prediction

[0035] Based on reliable data streams, a multi-stage progressive temperature control trend inference model is used. First, through short-term local fluctuation analysis, potential temperature rise sensitive areas are captured, and then gradually extended to long-term prediction of the entire system, realizing the progressive reasoning of temperature control trends from "point to surface".

[0036] The model incorporates an adaptive influence factor adjustment unit, which adjusts the influence weights in real time based on the phased calculation results, making the predictions more consistent with the actual thermal inertia of the equipment and external environmental disturbances.

[0037] The high-confidence environmental profile matrix output in step 1 As input, short-time fluctuation analysis is first performed for each local area or key equipment node. Specifically, for the temperature time series of each monitoring point... An innovative progressive interval perturbation detection method is employed. This method uses a sliding window... Calculate local increment Furthermore, using a sensitivity weighting function (in For node weights, (Using a nonlinear amplification factor) to assess the temperature rise sensitivity at the current moment. Compare the sensitivity of all nodes. It can quickly locate potential temperature-sensitive areas.

[0038] When the temperature rise sensitivity of one or more regions continuously exceeds the adaptive threshold At this point, the system automatically expands to adjacent areas, constructing an influence domain that progresses from "point" to "line" to "surface." Subsequently, in the medium- to long-term trend phase, the system performs multi-scale modeling of the temperature control state. Let the global temperature control state vector be denoted as... The trend prediction employs a self-developed time-progressive convolution-regression joint modeling strategy. First, multi-scale convolution is used to extract local trend features from the sequence of each point. ,in Indicates the convolution stride. This is the convolution parameter set. Subsequently, the local trend features from various regions are aggregated using weighted methods to form a system-level temperature control trend prediction. ,in Determined by the system's adaptive influence factor. To predict the step size, This is the bias constant.

[0039] In the adaptive influence factor adjustment unit of the model, the system evaluates the prediction error in real time after each stage of calculation is completed. Then, the feedback correction mechanism is used to update the weights of each node. And model parameters. The following adaptive recursion is used:

[0040]

[0041] in This is the learning rate coefficient. This is the prediction error for the current node. This mechanism can adjust the prediction weights in real time when there is high thermal inertia, severe external disturbances, or the emergence of new anomalies, making the prediction more consistent with the actual thermal field evolution, taking into account both sensitivity and robustness, and accurately supporting subsequent safety threshold judgments and start-up / shutdown decisions.

[0042] Step 3: Adaptive Judgment of Dynamic Safety Margin

[0043] Based on the progressive temperature control trend results, an innovative dynamic safety margin judgment mechanism is introduced, which automatically adjusts the single / multi-level alarm and redundant intervention thresholds in real time according to the near future trend, rather than a fixed threshold.

[0044] This mechanism analyzes the critical rate of temperature rise, predicts extreme values, and the real-time performance of the air-cooled unit, and comprehensively forms a margin envelope to provide a scientific decision-making window for whether to activate water cooling.

[0045] The multi-stage temperature control trend prediction results obtained in step 2 and the predicted values ​​of its local nodes Based on this, the safety margin of the cooling system is calculated in real time. The system first designs a dynamic margin envelope interval to replace the traditional fixed temperature threshold. Specifically, the upper limit of the system's safe allowable temperature is set as follows: The predicted extreme values ​​for temperature within the current and future window period (h minutes) are Safety margin is defined as

[0046]

[0047] This margin This reflects the closest the system temperature will be to the safe upper limit in the near future.

[0048] Furthermore, to enhance the dynamism and foresight of the threshold-judging mechanism, the system introduces a critical temperature rise rate. As an auxiliary criterion, that is

[0049]

[0050] And compare it with historical fluctuation thresholds in real time. To make comparisons and prevent the risk of rapid, short-term temperature increases. In addition, the system evaluates the performance parameters of the air-cooled unit in real time. (such as fan operating efficiency, heat exchanger cleanliness), and compared with the design optimal value. Compare and calculate the current air-cooling redundancy factor.

[0051]

[0052] The system normalizes and aggregates the above multiple indicators to dynamically determine alarm and water cooling intervention strategies. Threshold for judgment. Consider the following parameters:

[0053]

[0054] in This refers to a function defined through weighting or fuzzy logic. For example, a weighted sum can be defined:

[0055]

[0056] in As empirical weights. The system uses... Based on this, dynamic hierarchical activation of single / multi-level alarms is performed, and redundant intervention is triggered for judgment: if (Safety threshold) The system will then enter the water cooling unit preparation or start-up process.

[0057] Step 4: Multi-objective self-balancing start-up and shutdown decision

[0058] Entering the decision-making stage, a three-dimensional self-balancing decision-maker is constructed. Combining the safety margin range, equipment health status (including historical fatigue and current health factors), and energy consumption assessment results output in step three, an innovative automatic transfer of multi-objective weights is achieved—dynamic inter-adjustment of energy consumption, safety, and equipment health, to calculate the optimal start-up and shutdown timing and the required combination of water-cooling units.

[0059] The decision-maker considers not only individual start-up and shutdown, but also the lifecycle utilization efficiency and maintenance minimization throughout the entire process.

[0060] A three-dimensional self-balancing decision-maker is introduced to organically integrate safety margin, equipment health status, and energy efficiency assessment, achieving multi-objective dynamic optimization decision-making. First, the safety margin range output in step 3 is used... Equipment health status indicators (such as historical fatigue damage) With current health factors , Larger values ​​indicate better health), and energy consumption assessments (such as real-time power consumption). Weighted periodic energy consumption As input for decision-making. For each decision cycle, a three-objective loss function is defined:

[0061]

[0062] in:

[0063] (The smaller the margin, the greater the loss.) (as a scale factor)

[0064] (The worse the health or the higher the fatigue, the greater the loss);

[0065] Or the normalized value of periodic energy consumption (the higher the energy consumption, the greater the loss).

[0066] The self-transfer of target weights intelligently adjusts the weights of the three targets based on feedback. Maintain weight normalization ( If the system approaches a safety threshold or a significant temperature rise is predicted, it will automatically upgrade. (like , Indicates the main part. (As a regulating factor), if the equipment health is close to the fatigue limit, then increase it. During normal, stable periods, energy efficiency is prioritized.

[0067] In actual start-up and shutdown decisions, let the possible water-cooling combination schemes be: (e.g., the number and configuration of different water-cooling units in use), for each alternative solution Forecast the three target indicators for the next cycle and calculate the weighted total loss. Take the optimal

[0068]

[0069] in To predict the expected value within the interval, To optimize the window, the final output will be the optimal start / stop strategy. The system determines the next start-up and shutdown timing, mitigating safety risks while also considering equipment lifespan and energy efficiency. To achieve optimal performance throughout the entire lifecycle, the system continuously tracks the cumulative losses and maintenance costs of each strategy, dynamically fine-tuning weights and strategies to achieve intelligent, self-evolving start-up and shutdown behavior.

[0070] Step 5: Adaptive easing start / stop and fault closed-loop scheduling

[0071] Based on intelligent decision-making instructions, an adaptive slow-control start-stop program is initiated, which innovatively prevents the risks of water hammer and icing during start-stop by automatically adjusting the fluid flow and pressure in multiple stages.

[0072] If an abnormality or fault is detected in the air-cooled unit, the closed-loop self-healing scheduling mechanism is automatically activated to seamlessly switch to the water-cooled unit and automatically complete the load migration. At the same time, all data and switching trajectories are archived in real time for easy maintenance and traceability.

[0073] Based on the optimal start / stop command issued by the intelligent decision-making module in step 4, the adaptive slow-control start / stop program is first initiated. This program employs a multi-stage progressive control method for fluid flow and pipeline pressure. Specifically, after the start / stop command for the water-cooling unit is issued, it does not directly perform full-power start / stop. Instead, it uses an intelligent valve control system to gradually increase or decrease the flow rate and pressure in stages. For example, upon startup, low-speed pumping and slow valve opening are used to gradually and smoothly increase the hydraulic pressure, while monitoring pressure fluctuations. The system effectively suppresses water hammer by ensuring that the temperature does not exceed a pre-set safety threshold. Simultaneously, it analyzes the temperature distribution in the pipelines and heat exchangers in real time and dynamically adjusts the refrigerant flow rate to ensure the temperature drop rate remains within a controllable range, preventing the risk of icing due to localized overcooling. For rapid cooling needs, multiple pumps are activated in batches using a progressive, tiered approach to achieve controllable increases in flow rate and pressure.

[0074] During operation, if the system detects an anomaly in the air-cooled unit (such as a sudden drop in efficiency, abnormal vibration, or communication failure), it will automatically activate a closed-loop self-healing scheduling mechanism. This mechanism first quickly identifies the nature and scope of the anomaly using built-in criteria, then immediately issues a contingency plan to activate redundant water-cooled units, and automatically completes the cooling load migration according to a load ratio algorithm. Specifically, the system calculates the actual power load of air-cooled and water-cooled units in real time, dynamically allocates the target load for each unit, and flexibly schedules water-cooled backup resources based on load priority and equipment health factors. The entire switching process is fully automatic and seamless, ensuring not only continuous cooling but also avoiding equipment shocks caused by sudden start-up and shutdown.

[0075] Furthermore, the system utilizes high-frequency data acquisition and intelligent log archiving to record all key data, adjustment parameters, and switching trajectories during the start-up and shutdown process in real time. Each start-up, shutdown, and fault switching is associated with a unique identifier and timestamp, supporting subsequent equipment health analysis, operational decision tracing, and scheduling strategy optimization and upgrades. This mechanism significantly enhances the system's adaptive security, fault self-healing capabilities, and data traceability.

[0076] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0077] It should be understood that the above detailed description of the technical solutions of the present invention with reference to preferred embodiments is illustrative and not restrictive. Those skilled in the art can modify the technical solutions described in the embodiments or make equivalent substitutions for some of the technical features based on reading this specification; however, these modifications or substitutions do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent start-stop strategy for water-cooled units in an air-cooled main cooling system with water-cooled backup cooling, characterized in that: The strategy includes: A self-growing sensing network and data integration are constructed, and a self-growing sensing network activation algorithm is adopted to dynamically deploy and activate monitoring nodes according to changes in equipment operating heat load, so as to achieve multi-point high-density sensing. Multi-stage progressive temperature control trend prediction: Using a multi-stage progressive temperature control trend extrapolation model, first, through short-term local fluctuation analysis, potential temperature rise sensitive areas are captured, and then gradually extended to long-term prediction of the entire system. Dynamic safety margin adaptive judgment is based on progressive temperature control trend results. A dynamic safety margin threshold judgment mechanism is introduced to automatically adjust the single / multi-level alarm and redundant intervention thresholds in real time according to near-future trends, rather than fixed thresholds. Multi-objective self-balancing start-up and shutdown decision-making: Construct a three-dimensional trade-off self-balancing decision-maker, combine the safety margin range, equipment health status and energy consumption assessment results output in step three, realize automatic transfer of multi-objective weights—dynamic inter-adjustment of energy consumption, safety and equipment health, and calculate the optimal start-up and shutdown timing and the required combination of water-cooling units; Adaptive start-stop and fault scheduling: Based on intelligent decision-making instructions, the adaptive start-stop program is activated, and the automatic fluid flow and pressure are adjusted in multiple steps to prevent water hammer and icing risks during start-stop.

2. The intelligent start / stop strategy for the water-cooled unit in a wind-cooled main water-cooled backup cooling system according to claim 1, characterized in that: The self-growing sensing network and data integration include: dynamically deploying and activating monitoring nodes based on changes in equipment operating heat load to achieve multi-point high-density sensing, and ensuring the flexibility and integrity of the monitoring network through density adaptive activation rules; data integration standardizes, filters, and corrects real-time errors in the heterogeneous temperature, humidity, and wind speed data collected by each monitoring node, adopts an active data integration method based on weighted fuzzy consistency criteria to eliminate outliers and anomalies, and forms a high-confidence environmental profile matrix through spatiotemporal interpolation.

3. The intelligent start / stop strategy for the water-cooled unit in a wind-cooled main water-cooled backup cooling system according to claim 1, characterized in that: The aforementioned multi-stage progressive temperature control trend prediction includes: a multi-stage progressive temperature control trend extrapolation model, which takes a high-confidence environmental profile matrix as input, firstly uses a progressive interval perturbation detection method to perform short-term series fluctuation analysis on local areas or key equipment nodes, locates temperature rise sensitive areas through a sensitivity weighting function, automatically expands the influence domain when the sensitivity continuously exceeds the adaptive threshold, and enters the medium- and long-term trend stage, and adopts a time-progressive convolution-regression joint modeling strategy to perform multi-scale modeling and weighted aggregation of the temperature control state.

4. The intelligent start / stop strategy for the water-cooled unit in a wind-cooled main water-cooled backup cooling system according to claim 1, characterized in that: The aforementioned dynamic safety margin adaptive judgment includes: a dynamic safety margin thresholding mechanism, which is based on the multi-stage temperature control trend prediction results and the prediction values ​​of each local node, analyzes the temperature rise critical rate, temperature prediction extreme value, and real-time performance parameters of the air-cooled unit in real time, and comprehensively forms a dynamic margin envelope interval to replace the traditional fixed temperature threshold. By normalizing and aggregating key indicators, alarm and redundancy intervention thresholds are dynamically set. When the comprehensive threshold meets the water cooling intervention conditions, the water cooling unit preparation or startup process is automatically initiated, realizing real-time adaptive safety margin management and alarm hierarchical redundancy decision-making.

5. The intelligent start / stop strategy for the water-cooled unit in a wind-cooled main water-cooled backup cooling system according to claim 1, characterized in that: The multi-objective self-balancing start-up and shutdown decision-making includes a three-dimensional trade-off self-balancing decision-maker. The decision-maker integrates the safety margin range, equipment health status, and energy consumption assessment results as decision inputs. Through a multi-objective weight self-transfer mechanism, it realizes dynamic weight inter-adjustment of energy consumption, safety, and equipment health, comprehensively evaluates the risk, lifespan, and energy efficiency performance of each water-cooling unit combination scheme in the future start-up and shutdown cycle, adaptively outputs the optimal start-up and shutdown timing and water-cooling unit configuration, and continuously optimizes the weights and decision-making strategies based on the cumulative losses and maintenance expenditures throughout the entire cycle, realizing full life-cycle self-evolutionary management of start-up and shutdown behavior.

6. The intelligent start / stop strategy for the water-cooled unit in a wind-cooled main water-cooled backup cooling system according to claim 1, characterized in that: The adaptive slow start-stop and fault scheduling include: after the start-stop command of the water cooling unit is issued, a multi-step progressive control method of liquid flow and pipeline pressure is adopted. The flow rate and pressure are slowly adjusted in stages through the intelligent valve control system to suppress water hammer and freezing risks. Combined with real-time adjustment of refrigerant flow rate and pipeline temperature distribution, the temperature drop rate is kept stable. When an abnormality is detected in the air-cooled unit, the system automatically identifies the nature and scope of the abnormality, activates the redundant water-cooled unit activation plan, completes the cold load migration based on the load ratio algorithm, and allocates the target load and schedules water-cooled backup resources in real time. The entire process automatically archives data and switching trajectories to achieve seamless connection and continuous cooling guarantee.

7. The intelligent start / stop strategy for the water-cooled unit in a wind-cooled main water-cooled standby cooling system according to claim 3, characterized in that: The multi-stage progressive temperature control trend prediction model also includes an adaptive influence factor adjustment unit, which uses a feedback correction mechanism to update the weights of each node and model parameters in real time after each stage of calculation, so as to adapt to the thermal inertia of the equipment and changes in the external environment, and achieve sensitive and robust temperature control trend prediction.

Images