Water chilling unit intelligent control method and system based on machine learning

Abstract

The embodiment of the invention relates to the technical field of refrigeration, and discloses a water chilling unit intelligent control method based on machine learning, which comprises the following steps: acquiring a series of infrared thermal image frames of a monitoring area through each infrared camera arranged in the monitoring area; calculating weighted center points of the grid units; performing a dynamic clustering algorithm on the weighted center points in the successive frames to identify one or more independent heat sources and track their trajectories; constructing a multi-dimensional feature vector for each tracked heat source; calculating a group of kinematics parameters of the tracked heat source based on the multi-dimensional feature vector; and based on the calculated kinematics parameters, assigning a motion intensity grade to the tracked heat source; and determining the exercise intensity grade and the kinematics parameter, generating a control command for the water chilling unit control system based on the received exercise intensity grade and the kinematics parameter, and transmitting the control command to the corresponding water chilling unit control system to regulate and control the operation of the water chilling unit control system. The temperature can be adjusted and controlled in a self-adaptive mode.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of refrigeration technology, in particular to a cold water unit intelligent control method and system based on machine learning. BACKGROUND

[0002] At present, the existing venues have a single way of refrigeration regulation and control, which generally adopts fixed regulation, cannot adapt to refrigeration regulation and control under changing conditions, and cannot provide suitable refrigeration temperature for personnel in the venue. Therefore, designing a scheme that can provide suitable refrigeration temperature and adapt to environmental changes has become a technical problem to be solved by the skilled in the art. SUMMARY

[0003] In view of the defects, the embodiment of the present application discloses a cold water unit intelligent control method based on machine learning, which can realize accurate venue area temperature regulation.

[0004] The first aspect of the embodiment of the present application discloses a cold water unit intelligent control method based on machine learning, comprising: A series of infrared thermal image frames of the monitoring area are obtained by each infrared camera arranged in the monitoring area; The thermal radiation data from each infrared thermal image frame is mapped to an N*N grid representing the physical coordinates of the monitoring area, and the weighted center points of the grid units are calculated, each weighted center point being represented by coordinates and a normalized temperature value; A dynamic clustering algorithm is performed on the weighted center points in the continuous frames to identify one or more independent heat sources and track their trajectories; For each tracked heat source, a multi-dimensional feature vector is constructed, which includes position information, monitoring temperature information, inter-frame displacement and temperature change rate; A set of kinematic parameters of the tracked heat source is calculated based on the multi-dimensional feature vector, the set of kinematic parameters at least including a speed value and a heat diffusion entropy value; and a motion intensity level is assigned to the tracked heat source based on the calculated kinematic parameters; The motion intensity level and kinematic parameters are determined, control commands for the cold water unit control system are generated based on the received motion intensity level and kinematic parameters, and the control commands are transmitted to the corresponding cold water unit control system to regulate its operation.

[0005] As an optional implementation, in the first aspect of the embodiment of the present application, the calculation of a set of kinematic parameters of the tracked heat source based on the multi-dimensional feature vector comprises: For each identified tracked heat source in the infrared thermal image frame, the temperature value of the grid where it is located is extracted, and the temperature range is divided into K temperature intervals; counting the number of radiation points of the tracked heat source in each temperature interval to obtain a frequency distribution vector wherein the frequency distribution vector is the ratio of the number of radiation points in the i th temperature interval to the total number of radiation points; calculating a corresponding heat diffusion entropy value according to the frequency distribution vector and a heat diffusion entropy value calculation formula, and taking the heat diffusion entropy value as a kinematic parameter of the tracked heat source, wherein the heat diffusion entropy value calculation formula is: H is the heat diffusion entropy value, is the frequency distribution vector, and K is the number of temperature intervals.

[0006] As an optional implementation, in the first aspect of the embodiment of the present application, after the corresponding heat diffusion entropy value is calculated according to the frequency distribution vector and the heat diffusion entropy value calculation formula, the method further comprises: obtaining a spatial weight matrix of the corresponding monitoring area, and giving a correction coefficient to the temperature of the radiation points in different areas according to the spatial weight matrix; and determining a corrected heat diffusion entropy value according to the heat diffusion entropy value, the spatial weight matrix and an entropy value correction formula; the entropy value correction formula is: wherein is the heat diffusion entropy value after spatial correction, is the heat diffusion entropy value before correction; is the spatial weight of the j th radiation point, M is the number of effective radiation points, is the frequency distribution vector.

[0007] As an optional implementation, in the first aspect of the embodiment of the present application, after the corresponding heat diffusion entropy value is calculated according to the frequency distribution vector and the heat diffusion entropy value calculation formula, the method further comprises: using a sliding window to perform sliding calculation on a plurality of infrared thermal image frames to obtain entropy value sequence data, and determining a smoothed entropy value parameter according to the entropy value sequence data and an exponential smoothing formula; the exponential smoothing formula is: wherein is the smoothed entropy value parameter at the current moment, is the smoothed entropy value parameter at the previous moment, is the entropy value parameter at the current moment, is a smoothing coefficient; calculating an entropy value change rate, matching the entropy value change rate and the smoothed entropy value parameter with a set condition to determine a current thermal state, and regulating and controlling the water chiller unit control system according to the current thermal state.

[0008] ​As an optional implementation, in the first aspect of the embodiment of the present application, the intelligent control method further comprises: determining current scene type information, and determining corresponding temperature interval information according to the scene type information, wherein the temperature interval information is obtained by training historical thermal radiation data of different scenes through a K-means clustering algorithm.

[0009] As an optional implementation, in the first aspect of the embodiment of the present application, the dynamic clustering algorithm is a DBSCAN algorithm, and the intelligent control method further comprises: if the displacement of a clustered heat source within a preset number of frames is equal to or greater than a first threshold value and the temperature of the clustered heat source is equal to or greater than a second threshold value, the clustered heat source is classified as an active human body heat source.

[0010] calculating a temperature decay rate at the previous position of the active human body heat source after the active human body heat source moves away, and taking the temperature decay rate as a corresponding kinematic parameter; The intelligent control method further comprises: monitoring a plurality of heat sources; when a predetermined number of heat sources simultaneously meet the conditions that the speed value exceeds a speed threshold value and the heat diffusion entropy value exceeds an entropy threshold value within a specified time window, a group high-intensity event is detected; generating a control instruction of a chiller unit control system according to the detected group high-intensity event.

[0011] As an optional implementation, in the first aspect of the embodiment of the present application, the intelligent control method further comprises: constructing a trained LSTM model; inputting the constructed multi-dimensional feature vector into the LSTM model for recognition to determine the intensity of the next stage of movement; generating a control instruction of a chiller unit control system according to the intensity of the next stage of movement; or, the intelligent control method further comprises: generating a corresponding heat source map according to a series of infrared thermal image frames of the monitored area obtained by each infrared camera, and superimposing the heat source map and a blast opening map of the chiller unit to obtain a control map; the control command for the chiller unit control system is generated based on the received movement intensity level and kinematic parameters, and the control command is transmitted to the corresponding chiller unit control system to control the operation thereof, comprising: inputting the received movement intensity level, kinematic parameters, and control map into a pre-constructed air supply control model to determine the optimal air supply parameters of each refrigeration blast opening, wherein the optimal air supply parameters include air volume parameters, air direction parameters, and temperature parameters; Generate a control command for the chiller control system based on the optimal air supply parameters, and transmit the control command to the corresponding chiller control system to regulate its operation.

[0012] The second aspect of the embodiment of the application discloses a machine learning-based intelligent chiller control system, comprising: The acquisition module is configured to acquire a series of infrared thermal image frames of the monitoring area through the infrared cameras arranged in the monitoring area. The calculation module is configured to map the thermal radiation data from each infrared thermal image frame to an N×N grid representing the physical coordinates of the monitoring area, and calculate the weighted center points of the grid cells, each weighted center point being represented by a coordinate and a normalized temperature value. The clustering module is configured to perform a dynamic clustering algorithm on the weighted center points in consecutive frames to identify one or more independent heat sources and track their trajectories. The feature construction module is configured to construct a multi-dimensional feature vector for each tracked heat source, the multi-dimensional feature vector including position information, monitoring temperature information, inter-frame displacement, and temperature change rate. The motion calculation module is configured to calculate a set of kinematic parameters of the tracked heat source based on the multi-dimensional feature vector, the set of kinematic parameters including at least a velocity value and a thermal diffusion entropy value, and assign a motion intensity level to the tracked heat source based on the calculated kinematic parameters. The regulation module is configured to determine the motion intensity level and kinematic parameters, generate a control command for the chiller control system based on the received motion intensity level and kinematic parameters, and transmit the control command to the corresponding chiller control system to regulate its operation.

[0013] The third aspect of the embodiment of the application discloses an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor invokes the executable program code stored in the memory, and is configured to execute the machine learning-based intelligent chiller control method disclosed in the first aspect of the embodiment of the application.

[0014] The fourth aspect of the embodiment of the application discloses a computer-readable storage medium storing a computer program, wherein the computer program causes a computer to execute the machine learning-based intelligent chiller control method disclosed in the first aspect of the embodiment of the application.

[0015] Compared with the prior art, the embodiment of the application has the following beneficial effects: The method for intelligently controlling a water chiller based on machine learning in the embodiment of the present application divides the motion intensity level according to kinematic parameters, and the water chiller control system can formulate adaptive control instructions according to the heat source motion intensity. When the heat source motion intensity is high and the heat diffusion is fast, the refrigeration is accelerated; when the heat source distribution is stable and the intensity is low, the output is moderately reduced, thereby flexibly adapting to the real-time load requirements. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0017] Figure 1 is a flowchart of the method for intelligently controlling a water chiller based on machine learning disclosed in the embodiment of the present application; Figure 2 is a flowchart of the kinematic parameter calculation disclosed in the embodiment of the present application; Figure 3 is a structural diagram of the intelligent control system for a water chiller based on machine learning provided in the embodiment of the present application; Figure 4 is a structural diagram of an electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort belong to the scope of protection of the present application.

[0019] It should be noted that the terms first, second, third, fourth, etc. in the specification and claims of the present application are used to distinguish different objects, but not to describe a specific order. The terms of the embodiments of the present application include and have as well as any variations thereof, which are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0020] Embodiment one Please refer to Figure 1 , Figure 1is a flowchart of the method for intelligent control of a water chiller based on machine learning disclosed in embodiments of the present application. In the method described in embodiments of the present application, the execution subject is composed of software or / and hardware, which can receive relevant information through wired or / and wireless means and can send certain instructions. Of course, it can also have certain processing and storage functions. The execution subject can control multiple devices, such as remote physical servers or cloud servers and related software, or local hosts or servers and related software that perform relevant operations on devices placed somewhere. In some scenarios, it can also control multiple storage devices, which can be placed in the same place or different places. As shown in Figure 1 The method for intelligent control of a water chiller based on machine learning includes the following steps: S101: Obtain a series of infrared thermal image frames of the monitoring area through each infrared camera arranged in the monitoring area; S102: Map the thermal radiation data from each infrared thermal image frame to an N x N grid representing the physical coordinates of the monitoring area, and calculate the weighted center points of the grid cells, each weighted center point being represented by a coordinate and a normalized temperature value; S103: Perform a dynamic clustering algorithm on the weighted center points in consecutive frames to identify one or more independent heat sources and track their trajectories; S104: For each tracked heat source, construct a multi-dimensional feature vector containing position information, monitoring temperature information, inter-frame displacement, and temperature change rate; S105: Calculate a set of kinematic parameters for the tracked heat source based on the multi-dimensional feature vector, the set of kinematic parameters including at least a velocity value and a heat diffusion entropy value; and assign a motion intensity level to the tracked heat source based on the calculated kinematic parameters; S106: Determine the motion intensity level and kinematic parameters, generate control commands for the water chiller control system based on the received motion intensity level and kinematic parameters, and transmit the control commands to the corresponding water chiller control system to regulate its operation.

[0021] Embodiments of the present application use infrared cameras to collect thermal images and map thermal radiation data to physical coordinate grids and calculate weighted center points, which can accurately locate the physical position and corresponding temperature of heat sources in the monitoring area. At the same time, with the help of a dynamic clustering algorithm, multiple independent heat sources can be identified and their trajectories tracked, preventing misjudgment of heat sources, so that the water chiller can respond to the actual heat distribution situation.

[0022] Specifically, each N x N grid cell corresponds to a block of actual physical area in the monitoring area, such as a square area of 1 square meter. The coordinates of the weighted center point, which are usually two-dimensional plane coordinates such as x and y values, are the barycentric positions of all thermal radiation points in the grid cell, that is, the average positions calculated according to the temperature weights of the radiation points. The higher the temperature of the radiation point, the greater the contribution to the center point coordinates. This coordinate accurately corresponds to a specific position in the physical space, establishing a one-to-one correspondence between abstract thermal radiation data and the spatial layout of the actual monitoring area. The normalized temperature value of the weighted center point is a quantitative description of the overall thermal intensity in the grid cell. It is obtained by weighting and averaging the temperature values of the radiation points in the grid according to their weights (such as higher temperature, greater weight), and then normalized to a specific range (such as 0~1). This value reflects the average heat level of the grid area, and eliminates the absolute temperature differences caused by different infrared cameras or environmental light, making the thermal states of different grids and different time points comparable. The weighted center point is essentially a feature point that combines spatial position and thermal intensity information, which represents the overall thermal distribution state in the corresponding grid cell, and provides spatial anchoring basis data for subsequent heat source identification, trajectory tracking and control decision-making.

[0023] In the embodiments of the present application, the raw data of the infrared thermal image is a pixel-level temperature matrix (such as 1024 x 768 pixels), each pixel corresponds to a temperature value and coordinates. If these raw pixel points are directly clustered, the following problems will occur: data explosion, high computational cost; and because the original pixel points are discrete, adjacent pixels may be divided into different clusters due to small temperature differences. For example, the pixel temperature at the edge of the human heat source is slightly lower than that in the core area, and direct clustering may divide a complete human thermal radiation into multiple small clusters, which cannot be identified as "an independent heat source". The infrared image is easily disturbed by the environment (such as device reflection, air convection), and there are isolated high-temperature pixels (noise points). Direct clustering may misjudge these noise points as "small heat sources", resulting in redundant clustering results and interfering with subsequent heat source tracking and control decisions.

[0024] And the way of dividing the grid first and then clustering in the embodiment of the present application can achieve the following effects: first, greatly reduce the data volume and improve the computing efficiency; assuming that the monitoring area is divided into 100x100 grids (a total of 10,000 grid units), regardless of the number of pixels in the original image, only 10,000 weighted center points are finally output. The data volume is compressed from the million level to the ten-thousand level, and the computing pressure of the clustering algorithm is reduced by more than 100 times, which can meet the real-time requirement. Second, it can strengthen the spatial correlation. Each grid unit corresponds to a continuous physical area (such as 1m*1m), and the weighted center point integrates the temperature and position information of all pixels in the area, which is equivalent to "concentrating" the overall thermal characteristics in the area. For example, multiple adjacent grids covered by a human heat source, the weighted center point of each grid will be biased towards the human core, and the temperature characteristics are similar, making it easier to be classified as a complete cluster during clustering, avoiding being split. Third, filter local noise and improve clustering stability. The weighted calculation (temperature weighted average) in the grid will naturally smooth the local noise: a single high-temperature noise pixel has little effect on the weighted center point of the entire grid, while the continuous heat radiation of the real heat source will continuously affect multiple grids, and its characteristics are more prominent at the grid level. For example, a single high-temperature pixel caused by sunlight reflection will be "diluted" by the average calculation in the grid, while the continuous heat dissipation of the human body will cause the weighted center points of multiple grids to present stable high-temperature characteristics.

[0025] Specifically, after identifying the heat source, a multi-dimensional feature vector covering position, temperature, displacement, and temperature change rate can be constructed for the specific heat source, and kinematic parameters such as speed and heat diffusion entropy can be calculated, which can fully reflect the dynamic thermal characteristics of the heat source. This multi-dimensional description gives the control of the water chiller more abundant decision basis, avoiding the problem of untimely or inaccurate control caused by relying on only a single temperature factor.

[0026] According to the kinematic parameter, the motion intensity level is divided, and the water chiller control system can formulate adaptive control instructions according to the motion intensity of the heat source. When the heat source motion intensity is high and the heat diffusion is fast, the refrigeration is accelerated; when the heat source distribution is stable and the intensity is low, the output is appropriately reduced, thereby flexibly adapting to the real-time load requirement. The entire process relies on the automatic operation of data acquisition, algorithm analysis, and instruction generation, reducing the need for manual intervention. Through machine learning algorithm to continuously analyze the thermal image data, the water chiller can optimize the operation mode according to the dynamic changes of the monitoring area thermal environment, and effectively improve the intelligent control level. Because the water chiller can accurately adjust the operation according to the actual dynamic situation of the monitoring area heat source, it can prevent excessive refrigeration or insufficient refrigeration, ensure the environmental temperature control effect, and help reduce unnecessary energy consumption, achieving energy saving and consumption reduction.

[0027] More preferably, as shown in Figure 2 the multi-dimensional feature vector is calculated to obtain a set of kinematic parameters of the tracked heat source, including; S1051: Extract the temperature value of the grid where each identified tracked heat source in the infrared thermal image frame is located, and divide the temperature range into K temperature intervals; S1052: Count the number of radiation points of the tracked heat source in each temperature interval to obtain a frequency distribution vector , where the frequency distribution vector is the ratio of the number of radiation points in the i-th temperature interval to the total number of radiation points; S1053: Calculate the corresponding heat diffusion entropy value according to the frequency distribution vector and the heat diffusion entropy value calculation formula, and take the heat diffusion entropy value as the kinematic parameter of the tracked heat source, where the heat diffusion entropy value calculation formula is: , H is the heat diffusion entropy value, is the frequency distribution vector, and K is the number of temperature intervals.

[0028] In the embodiment of the present application, by dividing the temperature range into K intervals and counting the number of radiation points in each interval, the heat distribution characteristics of the heat source are converted into a quantifiable frequency distribution vector, making the originally abstract heat diffusion state specific and calculable. This quantization method provides a data basis for subsequent analysis of the heat diffusion law of the heat source.

[0029] The heat diffusion entropy value obtained based on the frequency distribution vector and the entropy value calculation formula can effectively reflect the degree of disorder and complexity of the heat distribution of the heat source. When the entropy value is high, it indicates that the heat radiation of the heat source is distributed more dispersedly in different temperature intervals, and the heat diffusion state is more complex; when the entropy value is low, it indicates that the heat radiation is concentrated in a few temperature intervals, and the heat diffusion is relatively stable. This accurate characterization helps to better understand the thermal behavior of the heat source. Including the heat diffusion entropy value in the kinematic parameter system enriches the description dimension of the state of the tracked heat source. Compared with relying only on parameters such as speed, the heat diffusion entropy value can provide additional information from the perspective of heat distribution, making the kinematic parameters more comprehensively reflect the dynamic characteristics of the heat source, and providing a more reliable basis for subsequent classification of motion intensity levels.

[0030] Since the heat diffusion entropy value can accurately reflect the heat diffusion of the heat source, the water chiller control system can more accurately judge the degree of influence of the heat source on the environment temperature based on this. For example, for a heat source with high heat diffusion entropy value, complex heat distribution and fast diffusion, a more aggressive refrigeration strategy can be adopted; for a heat source with low entropy value and stable heat distribution, the refrigeration intensity can be adjusted appropriately, thereby further improving the pertinence and effectiveness of the control.

[0031] By accurately grasping the heat diffusion characteristics of the heat source, the water chiller can more reasonably match the refrigeration output, avoiding energy waste caused by inaccurate judgment of the heat diffusion state. Under the premise of ensuring the temperature of the monitored area to be appropriate, further optimize the energy consumption and improve the energy saving performance of the system.

[0032] More preferably, after the corresponding heat diffusion entropy value is calculated according to the frequency distribution vector and the heat diffusion entropy value calculation formula, the method further comprises: S1054: obtaining a spatial weight matrix of the corresponding monitoring area, and giving a correction coefficient to the radiation point temperature of different areas according to the spatial weight matrix; S1055: and determining a corrected heat diffusion entropy value according to the heat diffusion entropy value, the spatial weight matrix and the entropy value correction formula; the entropy value correction formula is: wherein, is the spatially corrected heat diffusion entropy value, is the uncorrected heat diffusion entropy value; is the spatial weight of the jth radiation point, and M is the number of effective radiation points, is the frequency distribution vector.

[0033] The scheme of the embodiment of the application By giving a correction coefficient to the radiation point temperature of different areas through the spatial weight matrix, the calculation of the heat diffusion entropy value is no longer limited to simple temperature distribution statistics, but is combined with the spatial characteristics of the monitoring area (such as the heat dissipation demand of different positions, the importance of equipment layout, etc.). For example, a higher weight can be given to a densely populated area or a key equipment area, so that the entropy value is more in line with the actual space temperature control priority, and the accuracy of the entropy value in characterizing the real thermal environment is enhanced.

[0034] The corrected heat diffusion entropy value (H') distinguishes the contribution degree of the radiation points in different areas through the spatial weight, and can more sensitively capture the heat diffusion changes in the key area. Even if the original heat diffusion entropy values (H) of two heat sources are the same, if the high-temperature radiation points thereof are distributed in areas with different weights (such as core areas and edge areas), the corrected entropy values will have significant differences, thereby more accurately reflecting the actual influence of the heat source on the environmental temperature control.

[0035] Based on the spatially corrected heat diffusion entropy value, the chiller control system can realize more refined regionalized control. For example, when the corrected entropy value shows that the heat diffusion in the high-weight area is complex, the refrigeration in the area can be preferentially strengthened; and the heat diffusion changes in the low-weight area have less influence on the control decision, avoiding the one-size-fits-all regulation and control, and improving the spatial adaptability of the control strategy.

[0036] The scheme of the embodiment of the application corrects through the spatial weight, so that the system can focus on the key area which has a greater influence on the environmental temperature, meets the temperature control demand of the core area, and reduces the excessive regulation and control of the secondary area. This differentiated control can further optimize the energy distribution, reduce unnecessary energy consumption, and realize the dual improvement of energy saving and temperature control effect on the premise of ensuring the overall environmental comfort. ​

[0037] The spatial layout of the actual monitoring area is often complex (such as the presence of partitions, uneven distribution of equipment, etc.), and the introduction of the spatial weight matrix enables the calculation of the heat diffusion entropy value to adapt to the spatial characteristics of different scenarios. By adjusting the weight coefficient, this method can be flexibly applied to offices, shopping malls, industrial workshops and other places, improving the versatility and practicality of the technical solution.

[0038] More preferably, after the corresponding heat diffusion entropy value is calculated according to the frequency distribution vector and the heat diffusion entropy value calculation formula, it further comprises: The sliding window is used to slide and calculate a plurality of infrared thermal image frames to obtain entropy value sequence data, and the entropy value sequence data and an exponential smoothing formula are used to determine a smoothed entropy value parameter, and the exponential smoothing formula is: wherein, is the smoothed entropy value parameter at the current time, is the smoothed entropy value parameter at the previous time, is the entropy value parameter at the current time, is a smoothing coefficient; The entropy value change rate is calculated, the entropy value change rate and the smoothed entropy value parameter are matched with the set condition to determine the current thermal state, and the chiller control system is regulated and controlled according to the current thermal state.

[0039] Specifically, the infrared thermal image frame may generate noise due to environmental interference (such as transient heating of equipment, fluctuation of light), causing abnormal fluctuations in the entropy value of a single frame. After the entropy value sequence is calculated by the sliding window and processed by the exponential smoothing formula, the smoothed entropy value parameter can effectively filter transient noise, weaken the influence of accidental factors on the heat diffusion characteristics, and make the entropy value more stably reflect the long-term trend of the heat source, providing a more reliable data basis for subsequent judgment.

[0040] In the embodiment of the present application, the entropy value change rate is calculated, which can quantify the evolution speed and direction of the heat diffusion state of the heat source. For example, if the entropy value rises rapidly and the change rate is positive, it indicates that the heat diffusion complexity is increasing; if the entropy value slowly decreases and the change rate is negative, it indicates that the heat distribution tends to be stable. The capture of this dynamic trend makes up for the limitation that the entropy value at a single time can only reflect the static state, enabling the system to more sensitively perceive changes in the thermal environment.

[0041] Combining the smoothed entropy value parameter with the entropy value change rate can more carefully divide the current thermal state, such as stable low entropy, rapid increase in entropy, slow decrease in entropy, etc. Different thermal states correspond to different degrees of influence and change speed of the heat source on the environment, and the system can match more accurate regulation strategies accordingly - for example, starting the pre-cooling mechanism for the rapid increase in entropy state, and maintaining the basic refrigeration for the stable low entropy state, thereby improving the timeliness and foresight of the control decision.

[0042] The entropy value parameter after smoothing processing reduces misjudgment caused by instantaneous fluctuations. Combined with the setting of the trigger condition (such as adjusting only when the change rate exceeds the threshold), the frequent start-stop or large output adjustment of the water chiller caused by small disturbances can be avoided. This buffering mechanism reduces the mechanical wear and tear of the equipment, prolongs the service life, and reduces the additional energy consumption caused by frequent regulation.

[0043] In dynamic scenarios such as frequent personnel flow and intermittent equipment operation, the heat diffusion state of the heat source often presents nonlinear changes. The combination of sliding window and exponential smoothing can adapt to this dynamic characteristic, both retaining the influence of recent data (adjusting sensitivity through the alpha coefficient) and considering historical trends, so that the system can still maintain stable and accurate control performance in complex scenarios.

[0044] More preferably, the intelligent control method further comprises: determining current scene type information, and determining corresponding temperature interval information according to the scene type information, wherein the temperature interval information is obtained by training historical heat radiation data of different scenes through a K-means clustering algorithm.

[0045] The thermal environment characteristics of different scenes (such as offices, conference rooms, sports venues, and production workshops) differ significantly (such as normal temperature ranges and heat source distribution patterns). By training historical heat radiation data of different scenes through a K-means clustering algorithm, temperature interval information that matches the scene type is obtained, so that the division of temperature intervals no longer uses a unified standard, but rather fits the thermal characteristics of specific scenes, improving the rationality and relevance of temperature interval division.

[0046] Temperature intervals are the basis for calculating frequency distribution vectors and heat diffusion entropy values. Scene-based temperature interval information enables entropy value calculation to better match the actual heat distribution patterns of the current scene. For example, the temperature interval division for a conference room scenario takes into account the heat dissipation characteristics of personnel gathering, while a device room focuses on the temperature range of equipment operation, allowing the heat diffusion entropy value to more accurately reflect the heat source state in the scenario.

[0047] By identifying the current scene type and calling the corresponding temperature interval parameters, the system can flexibly respond to temperature control requirements in different application scenarios. When the scene changes (such as a conference room changing from empty to full), the system can automatically adapt to the temperature interval standards of the new scene, avoiding misjudgments of the heat state caused by changes in the scene, and ensuring that the control strategy of the water chiller always matches the characteristics of the scene.

[0048] The K-means clustering is used for training the historical heat radiation data, which is to make the system learn the heat feature rules of different scenes from the historical operation, and reduces the subjectivity and error of manually setting the temperature interval. With the accumulation of historical data, the temperature interval information can be continuously optimized, so that the system can better understand the thermal environment of various scenes, and the robustness and adaptability of the overall control algorithm are improved.

[0049] The association of scene type information and temperature interval makes the control of the water chiller not only rely on real-time heat data, but also combine the inherent properties of the scene (such as the comfortable temperature preference of the office scene, the device protection temperature threshold of the industrial scene). This combination makes the generation of control commands more in line with the functional requirements of the scene, ensuring the core temperature control target (such as personnel comfort, device safety), while further optimizing energy consumption and operation efficiency.

[0050] More preferably, the dynamic clustering algorithm is DBSCAN algorithm, and the intelligent control method further comprises: If the displacement of a clustered heat source within a preset frame number is equal to or greater than a first threshold value and its temperature is equal to or greater than a second threshold value, it is classified as an active human body heat source.

[0051] The temperature decay rate at the previous position of the active human body heat source after moving away is calculated, and the temperature decay rate is also used as the corresponding kinematic parameter; The intelligent control method further comprises: Monitoring a plurality of heat sources; When a predetermined number of heat sources simultaneously meet the conditions of speed value exceeding a speed threshold value and heat diffusion entropy value exceeding an entropy threshold value within a specified time window, a group high-intensity event is detected; Generating a regulation and control instruction of the water chiller control system according to the detected group high-intensity event.

[0052] The embodiment of the application classifies specific heat sources as active human body heat sources by using the DBSCAN algorithm for dynamic clustering of heat sources and combining displacement and temperature thresholds (first threshold value, second threshold value), which can effectively distinguish between human bodies and other static heat sources (such as devices). As a dynamic heat source, the human body has essential differences in heat dissipation characteristics from devices, and can optimize the refrigeration strategy for the human activity area after being identified separately. At the same time, the temperature decay rate after the human body moves away is calculated, which can reflect the heat dissipation speed of the area, and provide a more detailed basis for whether to maintain refrigeration or adjust the intensity, avoiding excessive refrigeration in unoccupied areas.

[0053] By monitoring multiple heat sources and setting a time window, when a predetermined number of heat sources simultaneously meet the speed and heat diffusion entropy threshold values, it is determined to be a group high-intensity event. Such events usually cause rapid temperature rise and complex heat diffusion, and the system can start the intensive cooling mode in advance to avoid the hysteresis caused by temperature accumulation after regulation, and improve the temperature control effect on dense crowd scenes.

[0054] As a density-based clustering method, DBSCAN algorithm does not need to pre-set the number of clusters, and can adaptively identify heat source groups of different densities, especially suitable for mixed environments of human and equipment heat sources in complex scenes. Compared with traditional clustering algorithms, it has more advantages in processing noise points (such as transient heat sources) and irregularly shaped heat sources, reducing the possibility of misjudging non-human heat sources as active humans, and improving the accuracy of heat source classification.

[0055] According to the individual characteristics (such as moving track, temperature decay) of active human heat sources and the overall characteristics of group high-intensity events, the system can develop a hierarchical regulation strategy: local precise cooling for individual active human areas; overall intensive cooling for group high-intensity event areas; and reducing cooling output for areas without active humans and with temperature decay to a reasonable range. This differentiated control maximizes the reduction of invalid energy consumption while ensuring the comfort of key areas.

[0056] Whether it is the movement of a single person or large-scale group activities, the system can quickly respond through heat source clustering, feature threshold judgment and other mechanisms. This method is suitable for office areas, shopping malls, conference rooms, production workshops and other scenes with frequent personnel flow, solving the problem of insufficient response of traditional control methods to dynamic crowd cooling, and improving the universality and practicality of intelligent control methods.

[0057] More preferably, the intelligent control method further comprises: constructing a trained LSTM model; inputting the constructed multi-dimensional feature vector into the LSTM model for identification to determine the motion intensity of the next stage; generating a regulation instruction of the chiller unit control system according to the motion intensity of the next stage; Alternatively, the intelligent control method further comprises: generating a corresponding heat source map according to a series of infrared thermal image frames of the monitored area obtained by each infrared camera, and superimposing the heat source map and the air outlet map of the chiller unit to obtain a regulation map; The control command for the chiller unit control system is generated based on the received motion intensity level and kinematic parameters, and the control command is transmitted to the corresponding chiller unit control system to regulate its operation, comprising: The received motion intensity level, kinematic parameters and regulation map are input into the pre-constructed air supply control model to determine the optimal air supply parameters of each refrigeration air outlet, the optimal air supply parameters including air volume parameters, air direction parameters and temperature parameters; A control command for the water chiller control system is generated according to the optimal air supply parameters, and the control command is transmitted to the corresponding water chiller control system to regulate the operation thereof.

[0058] The scheme of the embodiment of the present application can analyze a multi-dimensional feature vector through a trained LSTM model to predict the heat source motion intensity in the next stage. This prediction capability enables the water chiller control to change from passive response to the current state to active prediction of future demand, for example, to increase refrigeration preparation in advance when high motion intensity heat sources are predicted to appear soon, thereby avoiding discomfort caused by temperature fluctuations and significantly improving the timeliness and predictability of temperature control. That is, the heat production distribution of the entire venue is visualized as a heat source map, which is superimposed with an air outlet map.

[0059] The heat source map and the air outlet map are superimposed to generate a regulation map, which intuitively presents the spatial correspondence between heat source distribution and refrigeration equipment. Based on this, the air supply control model can calculate the optimal air supply parameters (air volume, air direction and temperature) for each refrigeration air outlet to achieve fine control of where refrigeration is needed and where precise air supply is provided. For example, the air volume is increased and the air direction is adjusted to directly blow away heat points in high-density heat source areas, and the output is reduced in low-temperature areas to avoid energy waste while improving local temperature control effect.

[0060] The LSTM model is good at capturing long-term dependencies in time series data and can effectively predict the trend of heat source motion intensity, especially suitable for scenarios with periodicity or trend such as personnel flow and equipment operation. The spatial matching of the heat source-air outlet map solves the problem of uneven regional refrigeration in traditional control, and the combination of the two enables the system to maintain stable and efficient control performance in a complex and dynamic environment.

[0061] The prediction output of the LSTM model and the parameter calculation of the air supply control model form a closed-loop decision chain, reducing the need for manual intervention. The system can autonomously complete the whole process of state perception, trend prediction, spatial matching, parameter optimization and instruction generation, not only improving control efficiency, but also continuously optimizing decision logic through continuous learning (such as iterative training of the LSTM model) to achieve self-improvement of intelligent control level.

[0062] Through the advance regulation of the LSTM model, the energy peak of temporary high-intensity refrigeration is avoided; through the precise optimization of the air outlet parameters, the energy loss of invalid air supply is reduced. Both of them work together to minimize the overall energy consumption of the water chiller on the premise of ensuring the temperature control accuracy of each region, achieving the dual goals of precise refrigeration and energy-saving operation.

[0063] Specifically, the core components and parameter design of the LSTM model Input layer: time series processing of multi-dimensional feature vectors, input data format: the model input is a historical time series sequence, not a single frame feature. For example, take the past 10 frames (10 seconds apart) of multi-dimensional feature vectors to form an input sequence with a shape of (10, D) (D is the feature dimension, such as position, temperature, displacement, entropy value, etc., assuming D=8). Effect: standardize the feature vector (e.g., normalize to the range 0-1) and convert it to a tensor format that LSTM can handle.

[0064] LSTM layer: capture temporal dependencies, number of layers and units: usually use 2-layer LSTM (adjustable according to data volume), the first layer has 64 hidden units (capture local temporal features), and the second layer has 32 hidden units (extract long-term trends).

[0065] Key parameters: return_sequences=True (first layer): return the output of each time step for the second LSTM layer to process; return_sequences=False (second layer): only return the output of the last time step, representing the compressed features of the entire sequence; add a Dropout layer (e.g., dropout=0.2): prevent overfitting and enhance the model's generalization ability.

[0066] Fully connected layer: feature integration and dimension compression, set 1-2 fully connected layers, for example: first layer: 16 neurons, activation function is ReLU (enhance non-linear fitting ability); second layer: 8 neurons, further compress the feature dimension. Effect: convert the high-dimensional feature vector output by LSTM into compact features related to the output target.

[0067] Output layer: predict exercise intensity, output form: choose according to the definition of exercise intensity (e.g., discrete levels 0-5 or continuous values): if discrete levels (classification task): the number of output layer neurons = the number of levels (e.g., 6), use Softmax as the activation function, output the probability of each level; if continuous values (regression task): output layer has 1 neuron, no activation function (or use Linear), directly output the predicted intensity value. Loss function: cross-entropy loss for classification tasks, mean squared error (MSE) for regression tasks. Then input historical training data to continuously optimize the model parameters to obtain the final model parameters.

[0068] Whether it is a dynamic scenario that relies on time series prediction or a complex layout that requires spatial precision control, this method can adapt to the needs by flexibly invoking the prediction model or spatial matching mechanism, further expanding the application scenarios of the water chiller control system and making it more suitable.

[0069] The intelligent control method of the water chiller based on machine learning in the embodiment of the present application divides the motion intensity level according to the kinematic parameters, and the water chiller control system can formulate adaptive control instructions according to the motion intensity of the heat source. When the motion intensity of the heat source is high and the heat diffusion is fast, the refrigeration is accelerated; when the heat source distribution is stable and the intensity is low, the output is moderately reduced, thereby flexibly adapting to the real-time load requirements.

[0070] Embodiment two Please refer to Figure 3 , Figure 3 is a structural schematic diagram of the intelligent control system of the water chiller based on machine learning disclosed in the embodiment of the present application. As shown in Figure 3 , the intelligent control system of the water chiller based on machine learning can include: The acquisition module 21 is configured to acquire a series of infrared thermal image frames of the monitoring area through each infrared camera arranged in the monitoring area. The calculation module 22 is configured to map the thermal radiation data from each infrared thermal image frame to an N×N grid representing the physical coordinates of the monitoring area, and calculate the weighted center points of the grid units, each weighted center point being represented by a coordinate and a normalized temperature value. The clustering module 23 is configured to perform a dynamic clustering algorithm on the weighted center points in the continuous frames to identify one or more independent heat sources and track their trajectories. The feature construction module 24 is configured to construct a multi-dimensional feature vector for each tracked heat source, the multi-dimensional feature vector including position information, monitoring temperature information, inter-frame displacement, and temperature change rate. The motion calculation module 25 is configured to calculate a set of kinematic parameters of the tracked heat source based on the multi-dimensional feature vector, the set of kinematic parameters including at least a velocity value and a heat diffusion entropy value; and assign a motion intensity level to the tracked heat source based on the calculated kinematic parameters. The regulation module 26 is configured to determine the motion intensity level and the kinematic parameters, generate control commands for the water chiller control system based on the received motion intensity level and kinematic parameters, and transmit the control commands to the corresponding water chiller control system to regulate its operation.

[0071] The intelligent control method of the water chiller based on machine learning in the embodiment of the present application divides the motion intensity level according to the kinematic parameters, and the water chiller control system can formulate adaptive control instructions according to the motion intensity of the heat source. When the motion intensity of the heat source is high and the heat diffusion is fast, the refrigeration is accelerated; when the heat source distribution is stable and the intensity is low, the output is moderately reduced, thereby flexibly adapting to the real-time load requirements.

[0072] Embodiment three Please refer to Figure 4 ,Figure 4 is a structural schematic diagram of an electronic device disclosed by the embodiment of the present application. The electronic device can be a computer, a server, and the like, and of course, under certain circumstances, can also be a smart device such as a mobile phone, a tablet computer, and a monitoring terminal, and an image acquisition device with processing function. As shown in the figure, the electronic device can include: Figure 4 a memory 510 storing executable program codes; a processor 520 coupled with the memory 510; The processor 520 calls the executable program codes stored in the memory 510 to execute part or all of the steps in the machine learning-based intelligent control method of the water chiller in the embodiment one.

[0073] The embodiment of the present application discloses a computer readable storage medium storing a computer program, wherein the computer program causes a computer to execute part or all of the steps in the machine learning-based intelligent control method of the water chiller in the embodiment one.

[0074] The embodiment of the present application also discloses a computer program product, wherein when the computer program product runs on a computer, it causes the computer to execute part or all of the steps in the machine learning-based intelligent control method of the water chiller in the embodiment one.

[0075] The embodiment of the present application also discloses an application publishing platform, wherein the application publishing platform is used to publish a computer program product, wherein when the computer program product runs on a computer, it causes the computer to execute part or all of the steps in the machine learning-based intelligent control method of the water chiller in the embodiment one.

[0076] In various embodiments of the present application, it should be understood that the size of the serial number of the processes does not mean the inevitable sequence of execution, and the execution sequence of the processes should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiment of the present application.

[0077] The units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. According to actual needs, part or all of the units can be selected to achieve the purpose of the embodiment of the present application.

[0078] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically independently, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0079] ​The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer accessible memory. Based on such understanding, the technical solutions of the present application essentially or the part of the prior art that contributes to the present application or the whole or part of the technical solutions can be embodied in the form of a software product, which is stored in a memory and includes a part or all steps of the method executed by a computer device (which can be a personal computer, a server or a network device, and specifically can be a processor in the computer device) to perform the embodiments of the present application.

[0080] In the embodiments provided by the present application, it should be understood that B corresponding to A means that B is associated with A, and B can be determined according to A. However, it should also be understood that the determination of B according to A does not mean that B is determined only according to A, but B can also be determined according to A and / or other information.

[0081] A person of ordinary skill in the art can understand that part or all of the steps in the various methods of the embodiments can be completed by a program instructing relevant hardware, and the program can be stored in a computer readable storage medium, including a Read-Only Memory (ROM), a Random Access Memory (RAM), a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), a One-time Programmable Read-Only Memory (OTPROM), an Electrically-Erasable Programmable Read-Only Memory (EEPROM), a Compact Disc Read-Only Memory (CD-ROM) or other optical disk storage, a magnetic disk storage, a magnetic tape storage, or any other medium that can be used to carry or store data which can be read by a computer.

[0082] The above describes in detail the machine learning-based intelligent control method and system of a water chiller unit, an electronic device and a storage medium according to the embodiments of the present application. The principles and implementation manners of the present application are described by applying specific examples in this paper. The above description of the embodiments is only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation manner and application range can be changed. In summary, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A machine learning-based intelligent control method for a water chiller unit, characterized in that, The method comprises: acquiring a series of infrared thermal image frames of a monitoring area through each infrared camera arranged in the monitoring area; mapping thermal radiation data from each infrared thermal image frame onto an N×N grid representing physical coordinates of the monitoring area, and calculating weighted center points of grid cells, each weighted center point being represented by a coordinate and a normalized temperature value; performing a dynamic clustering algorithm on the weighted center points in consecutive frames to identify one or more independent heat sources and track their trajectories; for each tracked heat source, constructing a multi-dimensional feature vector containing position information, monitoring temperature information, inter-frame displacement, and temperature change rate; calculating a set of kinematic parameters of the tracked heat source based on the multi-dimensional feature vector, the set of kinematic parameters including at least a velocity value and a heat diffusion entropy value; and assigning a motion intensity level to the tracked heat source based on the calculated kinematic parameters; determining the motion intensity level and kinematic parameters, generating control commands for the chiller control system based on the received motion intensity level and kinematic parameters, and transmitting the control commands to the corresponding chiller control system to regulate its operation.

2. The machine learning-based intelligent control method for a water chiller unit according to claim 1, wherein, The method further comprises: for each identified tracked heat source in the infrared thermal image frames, extracting the temperature value of the grid where the tracked heat source is located, and dividing the temperature range into K temperature intervals; counting the number of radiation points of the tracked heat sources in each temperature interval to obtain a frequency distribution vector wherein the frequency distribution vector is the ratio of the number of radiation points of the i-th temperature interval to the total number of radiation points. According to the frequency distribution vector and a heat diffusion entropy value calculation formula, a corresponding heat diffusion entropy value is calculated, and the heat diffusion entropy value is taken as a kinematic parameter of the tracked heat source, wherein the heat diffusion entropy value calculation formula is: , H is a heat diffusion entropy value, is a frequency distribution vector, and K is a temperature interval number. 3.The method of claim 2, wherein, after calculating the corresponding heat diffusion entropy value according to the frequency distribution vector and the heat diffusion entropy value calculation formula, the method further comprises: obtaining a spatial weight matrix of the corresponding monitoring area, and assigning a correction coefficient to the radiation point temperature of different regions according to the spatial weight matrix; And according to the heat diffusion entropy value, the spatial weight matrix and the entropy value correction formula, a corrected heat diffusion entropy value is determined. ; wherein, is the spatially modified heat diffusion entropy value, is the unmodified heat diffusion entropy value; is the spatial weight of the jth radiation point, M is the number of effective radiation points, is the frequency distribution vector. 4.The method of claim 2, wherein, after calculating the corresponding heat diffusion entropy value according to the frequency distribution vector and the heat diffusion entropy value calculation formula, the method further comprises: The entropy value sequence data is obtained by using a sliding window to slide and calculate a plurality of infrared thermal image frames, and a smoothed entropy value parameter is determined according to the entropy value sequence data and an exponential smoothing formula, the exponential smoothing formula being: wherein, is the smoothed entropy value parameter at a current moment, is the smoothed entropy value parameter at a previous moment, is the entropy value parameter at the current moment, is a smoothing coefficient. calculating an entropy change rate, matching the entropy change rate and the smoothed entropy value parameter with a set condition to determine the current heat state, and regulating the chiller control system according to the current heat state. 5.The method of claim 2, wherein, The intelligent control method further comprises: determining current scene type information, and determining corresponding temperature interval information according to the scene type information, the temperature interval information being obtained by training historical thermal radiation data of different scenes through a K-means clustering algorithm.

6. The machine learning-based intelligent control method for a water chiller unit as claimed in claim 1, wherein, The dynamic clustering algorithm is a DBSCAN algorithm, and the intelligent control method further comprises: if a clustered heat source has a displacement equal to or greater than a first threshold value within a preset number of frames and has a temperature equal to or greater than a second threshold value, it is classified as an active human heat source. calculating a temperature decay rate at the previous position of the active human heat source after it moves away, and taking the temperature decay rate as a corresponding kinematic parameter; The intelligent control method further comprises: monitoring a plurality of heat sources; detecting a group high-intensity event when a predetermined number of heat sources simultaneously meet the conditions of a velocity value exceeding a velocity threshold value and a heat diffusion entropy value exceeding an entropy threshold value within a specified time window; generating a regulation instruction for the chiller control system according to the detected group high-intensity event.

7. The machine learning-based intelligent control method for a water chiller unit as claimed in claim 1, wherein, The intelligent control method further comprises: constructing a trained LSTM model; The constructed multi-dimensional feature vector is input into the LSTM model for recognition to determine the movement intensity of the next stage; According to the movement intensity of the next stage, a control instruction of the water chilling unit control system is generated; Or, the intelligent control method further comprises: According to a series of infrared thermal image frames of the monitoring area obtained by each infrared camera, a corresponding heat source map is generated, and the heat source map is superimposed with the air port map of the water chilling unit to obtain a control map; The control command for the water chilling unit control system is generated based on the received movement intensity level and kinematic parameters, and the control command is transmitted to the corresponding water chilling unit control system to regulate its operation, comprising: The received movement intensity level, kinematic parameters and control map are input into the pre-constructed air supply control model to determine the optimal air supply parameters of each refrigeration air port, and the optimal air supply parameters include air volume parameters, air direction parameters and temperature parameters; According to the optimal air supply parameters, a control command for the water chilling unit control system is generated, and the control command is transmitted to the corresponding water chilling unit control system to regulate its operation.

8. A machine learning based intelligent control system for a water chiller unit, the system comprising: Comprise: An acquisition module for acquiring a series of infrared thermal image frames of a monitoring area through each infrared camera arranged in the monitoring area; A calculation module for mapping thermal radiation data from each infrared thermal image frame to an N×N grid representing the physical coordinates of the monitoring area, and calculating the weighted center points of the grid cells, each weighted center point being represented by coordinates and a normalized temperature value; A clustering module for performing a dynamic clustering algorithm on the weighted center points in consecutive frames to identify one or more independent heat sources and track their trajectories; A feature construction module for constructing a multi-dimensional feature vector for each tracked heat source, the multi-dimensional feature vector including position information, monitoring temperature information, inter-frame displacement and temperature change rate; A movement calculation module for calculating a set of kinematic parameters of the tracked heat source based on the multi-dimensional feature vector, the set of kinematic parameters including at least a velocity value and a heat diffusion entropy value; And assigning a movement intensity level to the tracked heat source based on the calculated kinematic parameters; A control module for determining the movement intensity level and kinematic parameters, generating a control command for the water chilling unit control system based on the received movement intensity level and kinematic parameters, and transmitting the control command to the corresponding water chilling unit control system to regulate its operation.

9. An electronic device, comprising: Comprise: A memory storing executable program code; A processor coupled with the memory; The processor invokes the executable program code stored in the memory for executing the machine learning-based intelligent control method of the water chilling unit according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, wherein the computer program causes the computer to execute the machine learning-based intelligent control method of the water chilling unit according to any one of claims 1 to 7. The computer readable storage medium stores a computer program, wherein the computer program causes the computer to execute the machine learning-based intelligent control method of the water chilling unit according to any one of claims 1 to 7.

Images