A group intelligence energy-saving coordination control method for a multi-connected central air conditioning system

Abstract

The application discloses a kind of multi-connected central air conditioning system group intelligence energy-saving coordination control methods, it is related to central air conditioning technical field, comprising: temperature sensor is laid in air conditioning area, forms the multi-point temperature measurement network of covering building space;Based on real-time temperature data, equipment characteristics and operating mode, a plurality of indoor units is divided into several control clusters with similar heat load variation mode by dynamic clustering algorithm;For each control cluster, calculate real-time temperature deviation;Determine its control priority;Comprehensively determine its current real-time operating control state;Generate the differentiated operating control signal for the cluster;Each operating control signal is respectively issued to each indoor unit in the corresponding control cluster.The application has the advantages that it can adapt to specific building structure, equipment characteristics and use mode, and achieve intelligent and flexible collaborative control of comfort and energy saving under the premise of prioritizing to ensure the temperature stability of high-priority areas.

Description

Technical Field

[0001] This invention relates to the field of central air conditioning technology, specifically to an intelligent energy-saving coordinated control method for multi-split central air conditioning system groups. Background Technology

[0002] Traditional multi-split central air conditioning systems rely heavily on manual experience for operation and control, or on simple temperature control strategies based on fixed schedules. Manual management struggles to accurately adapt to dynamically changing load demands, leading to low system efficiency and unstable comfort. Existing automated control systems typically employ uniform temperature setpoints or coarse-grained zone grouping control. This approach fails to effectively address the uneven distribution of heat loads across building areas, the dynamic changes caused by occupant movement and equipment start-ups and shutdowns, and the complexities of heat storage characteristics due to different building envelopes and orientations. It lacks the precision and dynamic control of the building's three-dimensional temperature field, easily causing localized overcooling, overheating, or temperature fluctuations, while simultaneously reducing overall system energy efficiency.

[0003] Current technologies lack real-time sensing and adaptive control mechanisms for the microscopic thermodynamics and load status during system operation. Significant changes occur in the building's thermal environment during critical events such as operating mode switching, personnel gathering in office areas, and the start-up and shutdown of large equipment. Fixed and lagging control strategies struggle to respond quickly and accurately, leading to significant temperature control deviations and potentially causing system oscillations. Furthermore, different operating modes, such as summer cooling, winter heating, and ventilation during transitional seasons, as well as different functional areas like conference rooms, offices, and corridors, have varying requirements for temperature control accuracy, stability, and response speed. Traditional systems often employ relatively fixed control logic and parameters, failing to provide differentiated and refined capacity output adjustments based on real-time operating status and regional importance. Therefore, there is an urgent need for a method and system capable of intelligently sensing the multi-dimensional thermal environment of the building space, dynamically dividing control clusters, and accurately and adaptively coordinating control based on the real-time thermodynamic behavior and control priorities of different clusters. This would optimize overall system energy efficiency while ensuring environmental comfort. Summary of the Invention

[0004] To address the aforementioned technical problems, an intelligent energy-saving coordinated control method for multi-split central air conditioning system groups is provided. This technical solution solves at least one of the technical problems mentioned in the background section.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for intelligent energy-saving coordinated control of multi-split central air conditioning system groups includes: Temperature sensors are deployed in each air-conditioned area covered by the multi-split central air conditioning system to form a multi-point temperature measurement network covering the building space. During system operation, temperature data of each air-conditioned area is collected in real time. Based on the temperature data, the operating characteristics of each indoor unit and the real-time operating mode, multiple indoor units are divided into several control clusters with similar heat load change patterns through a dynamic clustering algorithm. For each of the aforementioned control clusters, perform the following operations: Obtain the set temperature of the air-conditioned area within the cluster, and calculate the real-time temperature deviation based on the set temperature and the real-time average temperature of the cluster. Determine the control priority of the control cluster, wherein the control priority includes high priority or normal priority; Based on the magnitude of the real-time temperature deviation, the standard deviation of the temperature fluctuation of the cluster, and the determined control priority, the current real-time operation control status of the cluster is comprehensively determined. Based on the determined real-time operation control status, call the energy-saving control parameter set that matches the preset status, and generate a differentiated operation control signal for the cluster. Each of the aforementioned operation control signals is sent to the indoor units in the corresponding control cluster, driving them to perform cooling or heating operations with corresponding capabilities, thereby achieving precise, dynamic, and differentiated control of the temperature field in the three-dimensional space of the building, and optimizing the overall energy efficiency of the system while meeting comfort requirements.

[0006] Preferably, the step of dividing multiple indoor units into several control clusters with similar heat load change patterns using a dynamic clustering algorithm based on the temperature data, the operating characteristics of each indoor unit, and the real-time operating mode specifically includes: A multi-dimensional feature vector is constructed for each air-conditioned area corresponding to an indoor unit. The feature vector includes at least: the real-time average temperature of the area, the temperature uniformity index of the area, the thermal response coefficient of the indoor unit, and the spatial logical position weight in the current operating mode. Based on the multidimensional feature vectors corresponding to all indoor units, a density-based clustering algorithm is used for real-time dynamic grouping, and multiple indoor units whose feature vectors are close in distance in the feature space are grouped into the same control cluster. The clustering is recalculated in each control cycle or when the system detects a preset operating mode switching event, so as to achieve dynamic reorganization of the control cluster.

[0007] Preferably, determining the control priority of the control cluster specifically includes: The control cluster is predefined based on the building function attributes of the air-conditioned area corresponding to the control cluster and the comfort requirements of the current operating mode. Among them, the clusters corresponding to the functional areas with high temperature stability requirements are preset as high priority. When the system detects or receives a signal that there is a temporary high load event in a specific area, the control priority of the control cluster corresponding to the area affected by the event is temporarily marked as the high priority; The temperature fluctuation standard deviation of clusters that are preset or marked as the normal priority is continuously monitored under steady state. If the temperature fluctuation standard deviation continues to exceed the preset adaptive adjustment threshold, the adjustment priority is automatically updated to the high priority.

[0008] Preferably, the step of comprehensively determining the current real-time operation control status of the cluster based on the magnitude of the real-time temperature deviation, the standard deviation of the temperature fluctuation of the cluster, and the determined control priority specifically includes: If the absolute value of the real-time temperature deviation is greater than the first deviation threshold, it is determined to be in a rapid control state. If the absolute value of the real-time temperature deviation is less than or equal to the first deviation threshold, then the standard deviation of the temperature fluctuation of the cluster is further calculated based on historical temperature data: If the standard deviation of the temperature fluctuation is greater than the fluctuation threshold, it is determined to be a fluctuation suppression state; If the control priority of the cluster is high, it is determined to be in high-precision monitoring state.

[0009] Preferably, in the step of calling a set of energy-saving control parameters that pre-matches the determined real-time operation control state and generating a differentiated operation control signal for the cluster: The preset energy-saving control parameter set for the rapid adjustment state focuses on enhancing the proportional control effect; The energy-saving control parameter set preset for the fluctuation suppression state focuses on enhancing the differential control effect; Furthermore, under the same real-time operation and control conditions, the energy-saving control parameter set configured for clusters with high control priority has a stronger differential control effect compared to the energy-saving control parameter set preset for clusters with normal priority.

[0010] Preferably, the energy-saving control parameter set is obtained in the following way: During the initial system debugging phase, a preset parameter library based on experimental and simulation calibration is obtained; During the normal operation phase of the system, the parameters are dynamically fine-tuned based on the preset parameter library and a rule-based self-tuning mechanism. When the cumulative runtime of the system reaches the model training threshold, machine learning methods based on historical runtime data are used for optimization to obtain an optimization parameter library that is compatible with the current building and system characteristics.

[0011] Preferably, obtaining the preset parameter library based on experimental and simulation calibration specifically includes: On experimental platforms or high-precision system simulation models, various real-time operation control states are manually set or simulated for each typical system operation mode. Distinguish between high-priority and regular-priority regions through testing; Under each simulated operating condition combination, through experiments or simulation optimization, a set of energy-saving control parameters that balance response speed, temperature stability and optimal system energy efficiency are obtained. The optimized parameters are stored in the form of a multidimensional mapping table or configuration file to form an initial preset parameter library. The index keys of the preset parameter library include at least: running mode, real-time running control status, and control priority.

[0012] Preferably, the step of dynamically fine-tuning the parameters based on the preset parameter library and a rule-based self-tuning mechanism specifically includes: When the system determines that a control cluster is in a certain real-time operating control state for a period of time, if the temperature control effect in that area does not meet expectations, it triggers self-tuning of the energy-saving control parameters used by the current cluster, using a set of heuristic rules for dynamic fine-tuning. The specific steps are as follows: If the temperature approaches the set point too slowly, the proportional coefficient Kp will be increased by the preset step size. If the temperature overshoot is too large, decrease the proportional coefficient Kp by the preset step size or increase the differential coefficient Kd by the preset step size. If the steady-state deviation is eliminated slowly, the integral coefficient Ki is increased, but at the same time, it is monitored whether this will cause new oscillations. For regions where regulation is of high priority, increasing the differential coefficient Kd to suppress fluctuations takes precedence over increasing the proportional coefficient Kp.

[0013] Preferably, when the cumulative runtime of the system reaches the model training threshold, the optimization is performed using a machine learning method based on historical runtime data to obtain an optimization parameter library adapted to the current building and system characteristics. This specifically includes: Machine learning is performed using historical data collected during system operation to build a parameter optimization knowledge base; For a given set of system operating states, a similarity search is performed based on the parameter optimization knowledge base to find the historical operating record that is closest to the current features, and the set of energy-saving control parameters in that record that has been proven to have the best control effect in practice is output.

[0014] Preferably, the method further includes an adaptive monitoring and control mechanism, wherein the adaptive monitoring and control mechanism specifically comprises: For clusters under high-precision monitoring, increase the frequency of temperature monitoring and acquisition for their corresponding air-conditioned areas.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention constructs a multi-point temperature measurement network within a building space and uses a dynamic clustering algorithm to divide control clusters with similar heat load change patterns in real time, achieving refined perception and zoning of the three-dimensional spatial temperature field within the multi-split air conditioning system. By comprehensively determining the real-time operating control status and control priority of each cluster and invoking corresponding differentiated energy-saving control strategies, it can optimize the overall system operating energy efficiency while meeting the personalized comfort needs of each area, effectively avoiding localized overcooling, overheating, and large temperature fluctuations. Simultaneously, the system possesses online parameter self-tuning and machine learning optimization capabilities based on historical data, enabling it to adapt to the thermal characteristics of specific buildings, equipment performance, and user habits. While prioritizing temperature stability in high-priority areas, it improves the overall system operating energy efficiency, achieving intelligent and flexible synergy between comfort and energy saving. Attached Figure Description

[0016] Figure 1 This is a flowchart of the intelligent energy-saving coordinated control method for multi-split central air conditioning system groups proposed in this solution; Figure 2 This is a flowchart illustrating the method proposed in this solution for determining the current real-time operational control status of the cluster. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] Reference Figure 1 As shown, a method for intelligent energy-saving coordinated control of a multi-split central air conditioning system group includes: Temperature sensors are deployed in each air-conditioned area covered by the multi-split central air conditioning system to form a multi-point temperature measurement network covering the building space. Through the distributed and networked deployment method, the limitations of traditional single-point or simple room-by-room temperature measurement are overcome, realizing the ability to perceive the temperature field distribution in the three-dimensional space of the building in a refined and gridded manner, and providing a data basis for subsequent dynamic and differentiated capacity control. During system operation, temperature data for each air-conditioned zone is collected in real time. Based on this temperature data, the operating characteristics of each indoor unit, and its real-time operating mode, a dynamic clustering algorithm divides multiple indoor units into several control clusters with similar heat load variation patterns. Instead of pre-fixing control zones, the system dynamically clusters numerous indoor units into clusters with similar behavior based on real-time collected temperature field data, the operating characteristics of each indoor unit (such as thermal response coefficient), and its current operating mode. This allows the system to adapt to dynamic and uneven heat loads caused by factors such as occupant distribution, equipment start-up and shutdown, and solar radiation within the building. It achieves dynamic reorganization of control zones, overcoming the rigidity of fixed zone or floor-based grouping control strategies. The specific steps include: A multi-dimensional feature vector is constructed for each indoor unit corresponding to its air-conditioned area. The feature vector includes at least: the real-time average temperature of the area, the area's temperature uniformity index, the indoor unit's thermal response coefficient, and the spatial logical location weight under the current operating mode. The specific steps are as follows: At the start of each control cycle, the system acquires raw data from the temperature sensor network and the system controller. First, it processes multiple temperature readings collected within the most recent sampling window for the air-conditioned area corresponding to each indoor unit, calculating the real-time average temperature of that area. Simultaneously, it calculates the standard deviation of the temperature readings within this window, serving as an indicator of the area's temperature uniformity. Next, it reads the pre-calibrated or calculated thermal response coefficient for each indoor unit from the system's stored device attribute library. Finally, based on real-time operating mode information received from the system, such as summer office cooling, winter nighttime heating, and transitional season ventilation, it queries a predefined mode-location weight mapping table to obtain the spatial logical location weight of the area under the current operating mode. For a system with N indoor units, this process constructs a four-dimensional feature vector for the area corresponding to the i-th indoor unit: ,in, Let be the feature vector of the i-th region. Let be the real-time average temperature of the i-th region. Let be the standard deviation of temperature in the i-th region within the sampling window. The thermal response coefficient of the indoor unit serving the i-th region, The spatial logical position weight of the i-th region in the current operating mode; Based on the multidimensional feature vectors corresponding to all indoor units, a density-based clustering algorithm is used for real-time dynamic grouping, classifying multiple indoor units whose feature vectors are close in distance in the feature space into the same control cluster. The specific steps are as follows: The four-dimensional feature vectors of all N regions The features are aggregated to form an N×4 feature matrix and then placed into a feature space. Subsequently, a pre-defined density-based clustering algorithm is used to group each feature vector in this feature space. The core operations of the algorithm are: setting the neighborhood radius (Eps) and minimum number of points (MinPts) parameters; for each feature vector, calculating its Euclidean distance to other points in the four-dimensional feature space; grouping points whose distances are within Eps and whose density around the core point reaches MinPts into the same cluster; and marking discrete points that cannot be grouped into any high-density cluster as a separate cluster. The output of this step is: dividing all N indoor units into K clusters, i.e., K control clusters, and generating a mapping table of indoor unit IDs and cluster IDs. The specific method for calculating the Euclidean distance of each feature vector to other points in the four-dimensional feature space is as follows:

[0019] in, Let be the Euclidean distance between the feature vectors of the i-th region and the feature vectors of the j-th region. Let be the real-time average temperature of the j-th region. Let be the temperature standard deviation of the j-th region. Let be the thermal response coefficient of the indoor unit corresponding to the j-th region. The spatial logical position weight of the j-th region in the current operating mode; During clustering, the density around each feature vector is determined by: determining the number of all other feature vectors whose Euclidean distance to the current feature vector is less than the set neighborhood radius Eps, and denoted as the density around that point. Simultaneously, during clustering, priority is given to selecting clusters that are closer to the core point; Let's illustrate the above clustering process with a specific simulation example: Now assume there are 4 indoor units corresponding to the areas, denoted as area 1, area 2, area 3, and area 4, with a neighborhood radius of 0.3 and a minimum number of points of 2; After generating a corresponding four-dimensional feature vector for each region, the Euclidean distance between each feature vector is calculated according to the formula. Assume the calculation result is: d(1,2)=0.185, d(1,3)=0.205, d(1,4)=0.326, d(2,3)=0.315, d(2,4)=0.307, d(3,4)=0.175; At this point, the distance between the feature vector of region 1 and regions 2 and 3 is less than the set neighborhood radius of 0.3, and the number exceeds 2, so region 1 is the core point; The feature vector of region 2 is less than 0.3 away from that of region 1, and the number of features does not exceed 2, so region 2 is not a core point; The feature vector of region 3 is less than 0.3 away from regions 1 and 4, and the number of features exceeds 2, indicating that region 3 is the core point. The feature vector of region 4 is less than 0.3 away from that of region 3, and the number of features does not exceed 2, so region 4 is not a core point; In this scheme, clusters require similar control parameters, therefore the density connectivity criterion is not used. Based on the above results, the achievable clustering results are as follows: Using region 1 as the core point, regions 1, 2, and 3 are grouped into the same cluster, while region 4 is grouped into a separate cluster. In this case, the sum of the minimum cluster distances of regions 1, 2, and 3 as the same cluster is 0.185 + 0.205 = 0.39, while the distance of region 4 as a separate cluster is 0. The total distance for this clustering method is 0.39. Using region 1 as the core point, regions 1 and 2 are in the same cluster. Using region 3 as the core point, regions 3 and 4 are in the same cluster. At this time, the sum of the minimum cluster distances of regions 1 and 2 as the same cluster is 0.185, and the sum of the minimum cluster distances of regions 3 and 4 as the same cluster is 0.175. The total distance of this clustering method is 0.185 + 0.175 = 0.36. With region 3 as the core point, regions 1, 3, and 4 are in the same cluster, and region 2 is in a separate cluster. At this time, the sum of the minimum cluster distances of regions 1, 3, and 4 as the same cluster is 0.205 + 0.175 = 0.38, and the distance of region 2 as a separate cluster is 0. The total distance of this clustering method is 0.38. According to the criterion of prioritizing the clusters with smaller distances to the core points during clustering, the final clustering results are: region 1 and region 2 are in the same cluster, and region 3 and region 4 are in another cluster. Clustering is recalculated in each control cycle or when the system detects a preset operating mode switching event to achieve dynamic reorganization of the control cluster. The specific steps are as follows: The system incorporates a re-aggregation trigger mechanism with triggering conditions including periodic triggering: automatically triggering a new round of clustering calculations at the end of each fixed control cycle, such as every 5 minutes; and event-driven triggering: when the system detects or receives a specific preset system event, such as an operating mode switching command or a signal of people gathering in a specific area, it immediately interrupts the current cycle and triggers an emergency clustering recalculation. Each time triggered, the system re-executes the above steps, generating new feature vectors and performing clustering based on the latest temperature data and operating status, thereby updating the indoor unit ID—cluster ID mapping table. This mechanism ensures that the control cluster division can dynamically respond to changes in building heat load distribution, personnel activity, and system operating requirements, achieving true adaptive zoning control.

[0020] By constructing a multi-dimensional feature vector integrating real-time temperature, uniformity, thermal response characteristics, and spatial weights for each control region, the real-time thermodynamic behavior of each region can be comprehensively and quantitatively characterized from multiple dimensions. This provides accurate and rich evidence for intelligent clustering, solving the problem of inaccurate control zoning caused by traditional methods relying solely on a single temperature setpoint or simple location grouping. Employing a density-based clustering algorithm for real-time dynamic grouping automatically discovers and aggregates regions with similar behaviors in the feature space, forming truly clustered control clusters. This achieves a fundamental shift in zoning from a preset fixed pattern to data-driven, dynamically adaptive zoning, significantly improving the matching degree between zoning and the actual heat load state. Furthermore, a dual-trigger mechanism of periodicity and events drives dynamic cluster reorganization, ensuring not only periodic optimization of zoning but also instantaneous response to key events such as operating mode switching and personnel movement. This allows control zoning to follow changes in the building's internal thermal environment and usage demands in real time, greatly enhancing the system's adaptability and control timeliness. The combined effect of these multiple mechanisms lays a solid and reliable foundation for achieving refined and differentiated control of the three-dimensional spatial temperature field and overall energy efficiency optimization in a dynamically changing building environment. For each of the aforementioned control clusters, perform the following operations: The set temperature of the air-conditioned area within the cluster is obtained, and the real-time temperature deviation is calculated based on the set temperature and the real-time average temperature of the cluster. By comparing the user's set temperature with the measured average temperature of the cluster, a key control input signal—temperature deviation—is generated, providing a quantitative basis for generating accurate capacity control commands in the future. The control cluster's control priority is determined, including high priority or normal priority. Based on the functional attributes or specific operational events of the corresponding area of ​​the cluster, a high or normal priority label is assigned. This provides a logical basis for implementing differentiated control precision and response speed strategies, ensuring focused monitoring and control of key functional areas, and optimizing overall system energy consumption while guaranteeing core comfort needs. Specifically, the method for determining the control priority is as follows: The control cluster is predefined based on the building function attributes of the air-conditioned area corresponding to the control cluster and the comfort requirements of the current operating mode. Among them, functional areas with high requirements for temperature stability, such as conference rooms and precision instrument rooms, are pre-defined as high priority clusters. When the system detects or receives a signal of a temporary high-load event in a specific area, such as the start-up of large equipment or a temporary large-scale gathering of people, the control priority of the control cluster corresponding to the area affected by the event is temporarily marked as high priority. The temperature fluctuation standard deviation of clusters preset or marked as having the conventional priority is continuously monitored under steady state. If the temperature fluctuation standard deviation continuously exceeds the preset adaptive adjustment threshold, the control priority is automatically updated to the high priority. The temperature fluctuation standard deviation is calculated by collecting a real-time temperature sequence within a period and calculating the standard deviation of all temperature data in the real-time temperature sequence. In some preferred embodiments, the period duration is set to 30 minutes. The preset adaptive adjustment threshold is set to 5%-10% of the steady-state temperature setpoint in some preferred embodiments. The specific percentage is determined based on the control accuracy requirements of the region. The higher the control accuracy requirements of the region, the lower the percentage. The threshold is adaptively calculated by combining the preset percentage and the real-time temperature setpoint.

[0021] The proposed control priority determination mechanism is based on a multi-layered, adaptive, and dynamic-static integrated intelligent decision-making logic. Predefined functional attributes and comfort requirements provide the system with initial, a priori recognition of high-priority areas, ensuring focused monitoring of known critical areas from the system's initial operation. The introduction of dynamic event-based tagging enables the system to respond instantly to sudden events that significantly alter local heat load, such as personnel gatherings and equipment start-ups / shutdowns, temporarily elevating the control priority of relevant areas. This demonstrates the dynamic adaptability of the control strategy to uncertainties during building use. Most innovatively, the continuous monitoring-based adaptive update mechanism endows the system with learning and self-diagnostic capabilities: even areas preset to normal priority can be automatically identified and upgraded to higher priority if their temperature stability continues to deteriorate, thus enabling proactive detection and handling of unforeseen abnormal conditions. These three layers of judgment logic are interconnected, jointly ensuring the identification of the key concept of "control priority". It can dynamically evolve with changes in the operation process and building status, so that subsequent differentiated control can always focus on the area that needs the most precise control at present, fundamentally improving the robustness and intelligence level of the entire control system.

[0022] Reference Figure 2 As shown, based on the magnitude of the real-time temperature deviation, the standard deviation of the cluster's temperature fluctuation, and the determined control priority, the current real-time operation control status of the cluster is comprehensively determined; the specific steps include: If the absolute value of the real-time temperature deviation is greater than the first deviation threshold, it is determined to be a rapid control state. In some preferred embodiments, the first deviation threshold is set to 5% of the temperature setpoint. For example, if the temperature setpoint is 24°C, the first deviation threshold is 1.2°C. The higher the stability control accuracy, the smaller the first deviation threshold is set. If the absolute value of the real-time temperature deviation is less than or equal to the first deviation threshold, then the standard deviation of the temperature fluctuation of the cluster is further calculated based on historical temperature data. Here, the standard deviation of temperature fluctuation is obtained by calculating the standard deviation of the temperature sequence collected in the past 10 minutes. If the standard deviation of temperature fluctuation is greater than the fluctuation threshold, it is determined to be a fluctuation suppression state. In some preferred embodiments, the value of the fluctuation threshold is set to 3% of the temperature setpoint. For example, if the temperature setpoint is 24°C, the fluctuation threshold is 0.72. For control areas with higher stability requirements, the fluctuation threshold is set smaller. If the control priority of the cluster is high priority, it is determined to be in high-precision monitoring state. For areas in high-precision monitoring state, the temperature monitoring and acquisition frequency of the corresponding air-conditioning area is increased. In this embodiment, the temperature monitoring and acquisition frequency of high-precision monitoring state is increased to twice the temperature sampling frequency.

[0023] The proposed real-time operation control state determination mechanism implements a hierarchical, condition-triggered intelligent decision-making logic. First, using the absolute value of the real-time temperature deviation exceeding a first deviation threshold as the sole criterion, clusters with temperatures severely deviating from the setpoint are directly and quickly identified as being in a rapid adjustment state. This ensures the system can prioritize control resources and correct significant deviations with maximum response speed, demonstrating the control system's rapidity. When the temperature has entered the allowable deviation range, a second-level judgment is initiated. By comparing the standard deviation of temperature fluctuation with the fluctuation threshold, areas where the average temperature meets the standard but fluctuates drastically and affects comfort are identified and classified as fluctuation suppression states. This reflects the control system's high-order pursuit of temperature stability and comfort consistency. Clusters marked as high priority are assigned a high-precision monitoring state. They are maintained with the highest level of monitoring and control precision to preventatively resist any possible disturbances and ensure extreme temperature stability in critical areas. This progressive determination mechanism enables the control strategy to accurately match the instantaneous dynamic needs of the clusters, making it a key decision-making link for achieving intelligent and refined differential control.

[0024] The aforementioned real-time operation control status judgment process comprehensively considers temperature deviation, regional stability, and control priority, thereby intelligently identifying the specific operating condition of the cluster. This multi-dimensional judgment enables the control system to more accurately understand the current situation, laying the foundation for selecting the most suitable control law; Based on the determined real-time operation control state, a preset energy-saving control parameter set matching that state is invoked, and a differentiated operation control signal optimized for the cluster is generated. Based on the condition determination result from the previous step, the most suitable parameter set is invoked, thereby generating a differentiated capacity control signal that fits the actual needs of the current cluster. This achieves precise matching and output of the control strategy. The specific steps include: The energy-saving control parameter set preset for the fast adjustment state focuses on enhancing the proportional control effect, and the parameter set configured for the fast adjustment state focuses on enhancing the proportional (P) control effect. The purpose is to give the control system the ability to respond quickly and strongly to the current significant temperature deviation, pull the temperature back to the set range at the maximum speed, and prioritize solving the significant deviation problem. The energy-saving control parameter set preset for the fluctuation suppression state focuses on enhancing the differential control effect, and the parameter set configured for the fluctuation suppression state focuses on enhancing the differential (D) control effect. This is to utilize the predictive and damping characteristics of the differential action on the temperature change trend to effectively suppress frequent temperature fluctuations and improve the stability of the regional temperature. Furthermore, under the same real-time operating control state, the energy-saving control parameter set configured for high-priority clusters exhibits stronger differential control compared to the preset energy-saving control parameter set for regular-priority clusters. Under the same operating conditions, a secondary distinction is made between the parameters in the high-priority and regular-priority regions: a parameter set with stronger differential action is configured for the high-priority region. This secondary distinction is achieved through the differential calibration of the preset parameter library and the priority mechanism of subsequent system operation fine-tuning. During the construction of the preset parameter library, experiments are conducted to differentiate between the high-priority and regular-priority regions. In the experiments in the high-priority region, the control priority for increasing the differential coefficient is increased to enhance the differential coefficient during the initial optimization process of the high-priority region. In the subsequent dynamic fine-tuning rules, the priority for enhancing the differential coefficient is set higher than other operations. The combination of these two approaches achieves the adjustment of the parameter set with stronger differential action for the high-priority region. This is equivalent to adding a forward-looking buffer and damping layer to the high-priority region, making it less sensitive to external disturbances, thereby achieving higher control stability and accuracy at the micro level than the regular region, ensuring absolute reliability of comfort in critical areas.

[0025] The control parameter regulation mechanism proposed in this method intelligently links different control actions in the classic PID control principle with specific operation control objectives and regional importance. It is the core strategy library for realizing differentiated operation control signal generation and ultimately achieving synergistic optimization of comfort and energy saving.

[0026] Specifically, the energy-saving control parameter set is obtained as follows: During the initial system debugging phase, a preset parameter library based on experimental and simulation calibration is obtained. The specific steps are as follows: On experimental platforms or high-precision system simulation models, for each typical system operating mode, such as the summer office day cooling mode, various real-time operating control states are manually set or simulated, such as rapid regulation state and fluctuation suppression state. Simultaneously, high-priority and regular-priority regions are differentiated for testing. Under each simulated operating condition combination, a set of energy-saving control parameters that balances response speed, temperature stability, and optimal system energy efficiency is obtained through experimental or simulation optimization. In the experimental parameter optimization of the high-priority region, an emphasis is placed on temperature stability to meet the control requirement of increasing the differential coefficient in the initial parameters of the high-priority region. The optimized parameters are stored in the form of a multi-dimensional mapping table or configuration file, forming an initial preset parameter library. The index keys of this library include at least: operating mode, real-time operating control state, and regulation priority. During system operation, the corresponding preset parameter set is directly called through the current operating state combination index. During the normal operation phase of the system, based on the preset parameter library, the parameters are dynamically fine-tuned using a rule-based self-tuning mechanism. The specific steps are as follows: When the system determines that a certain control cluster is in a real-time operating control state for a period of time, if the temperature control effect in that area does not meet expectations, for example, if the decay rate of the temperature deviation e(t) is too slow or constant amplitude oscillations occur, then self-tuning of the energy-saving control parameters used by the current cluster is triggered, and a set of heuristic rules are used for dynamic fine-tuning: if the temperature approaches the set point too slowly, the proportional coefficient Kp is increased by a preset step size; if the temperature overshoot is too large, the proportional coefficient Kp is decreased by a preset step size or the derivative coefficient Kd is increased by a preset step size; if the steady-state deviation is eliminated slowly, the integral coefficient Ki is increased, but at the same time, it is monitored whether this will cause new oscillations; for areas with high control priority, increasing the derivative coefficient Kd has a higher priority than other coefficient adjustment operations, and decreasing the derivative coefficient Kd has a lower priority than other coefficient adjustment operations. For example, if the overshoot in a high-priority area is too large, decreasing the proportional coefficient Kp by a certain step size has a lower adjustment priority than increasing the derivative coefficient Kd by a certain step size, so as to prioritize suppressing overshoot. The fine-tuned parameters are immediately applied to the current control cycle, and this adjustment is recorded in the system operation log as raw data for subsequent optimization learning; When the system's cumulative runtime reaches the model training threshold, machine learning methods based on historical runtime data are used for optimization to obtain an optimization parameter library that is compatible with the current building and system characteristics. The specific steps are as follows: After collecting sufficient data during daily operation, a self-learning optimization mechanism is activated. Specifically, machine learning is performed using historical data collected during system operation to build a parameter optimization knowledge base. For a given combination of system operating states, a similarity search is performed based on this knowledge base to find the historical operating record that most closely matches the current characteristics, and the energy-saving control parameters from that record, proven to have the best control effect in practice, are output. This self-learning optimization mechanism can update or replace the original preset parameter library, or serve as a high-performance supplement. When the system encounters the same or similar operating conditions again, it can call upon this optimization knowledge base to obtain a better set of parameters that better fits the actual site conditions. This approach allows the control system to gradually adapt to the unique thermodynamic and load characteristics brought about by specific building structures, air conditioning equipment characteristics, and user usage patterns.

[0027] The proposed progressive, multi-stage energy-saving control parameter set acquisition and optimization scheme provides reliable and reproducible startup knowledge for the system based on a preset parameter library calibrated through experiments and simulations. This ensures the system possesses basic and robust control capabilities from initial commissioning and lays a high-quality benchmark for subsequent optimization. The rule-based self-tuning mechanism during operation endows the system with online and dynamic fine-tuning capabilities, enabling it to adaptively correct preset parameters based on real-time control effects. This effectively addresses subtle deviations between the preset model and actual conditions, as well as slow time-varying factors during operation, improving the system's real-time adaptability and robustness. In particular, its rule of prioritizing the adjustment of differential action in high-priority areas demonstrates proactive protection for the stability of critical regions. Machine learning optimization based on historical data represents a significant advancement in system intelligence. By learning from massive amounts of operational data, it can uncover optimal parameter combinations that surpass preset rules and human experience, allowing the system parameter library to continuously evolve. Ultimately, this achieves a high degree of fit with specific buildings, equipment, and usage patterns, resulting in truly customized, efficient, and comfortable control. These three stages—from preset to self-tuning to self-learning—form a complete parameter lifecycle closed loop, enabling the control system to not only possess expert experience but also the ability to learn and grow in practice, thereby continuously improving the comfort, stability, and energy efficiency of temperature control. Each of the aforementioned operational control signals is sent to the indoor units in the corresponding control cluster, driving them to perform cooling or heating operations with the appropriate capabilities. This achieves precise, dynamic, and differentiated regulation of the building's three-dimensional temperature field, optimizing overall system energy efficiency while meeting comfort requirements. The calculated differentiated operational control signals are then distributed to specific indoor units within each dynamic cluster, driving them to operate at different capabilities. Ultimately, through the coordinated action of all clusters, precise and dynamic adjustment of the entire building's three-dimensional temperature field is achieved macroscopically, optimizing overall system energy efficiency while meeting the individual comfort needs of each area.

[0028] In summary, the advantages of this invention are as follows: By constructing a multi-point temperature measurement network in the building space and using a dynamic clustering algorithm to divide control clusters with similar heat load change patterns in real time, it achieves refined perception and zoning of the three-dimensional spatial temperature field under the jurisdiction of the multi-split air conditioning system. By comprehensively determining the real-time operation control status and regulation priority of each cluster and calling the corresponding differentiated energy-saving control strategies, it can optimize the overall operating energy efficiency of the system while meeting the personalized comfort needs of each area, effectively avoiding local overcooling, overheating, and large temperature fluctuations. Simultaneously, the system possesses online parameter self-tuning and machine learning optimization capabilities based on historical data, enabling it to adapt to the thermal characteristics of specific buildings, equipment performance, and user habits. While prioritizing the temperature stability of high-priority areas, it improves the overall operating energy efficiency of the system, achieving intelligent and flexible synergy between comfort and energy saving.

[0029] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for intelligent energy-saving coordinated control of multi-split central air conditioning system groups, characterized in that, include: Temperature sensors are deployed in each air-conditioned area covered by the multi-split central air conditioning system to form a multi-point temperature measurement network covering the building space. During system operation, temperature data of each air-conditioned area is collected in real time. Based on the temperature data, the operating characteristics of each indoor unit and the real-time operating mode, multiple indoor units are divided into several control clusters with similar heat load change patterns through a dynamic clustering algorithm. For each of the aforementioned control clusters, perform the following operations: Obtain the set temperature of the air-conditioned area within the cluster, and calculate the real-time temperature deviation based on the set temperature and the real-time average temperature of the cluster. Determine the control priority of the control cluster, wherein the control priority includes high priority or normal priority; Based on the magnitude of the real-time temperature deviation, the standard deviation of the temperature fluctuation of the cluster, and the determined control priority, the current real-time operation control status of the cluster is comprehensively determined. Based on the determined real-time operation control status, call the energy-saving control parameter set that matches the preset status, and generate a differentiated operation control signal for the cluster. Each of the aforementioned operation control signals is sent to the indoor unit in the corresponding control cluster to drive it to perform cooling or heating operations with corresponding capabilities, thereby achieving precise, dynamic and differentiated control of the temperature field in the three-dimensional space of the building, and optimizing the overall energy efficiency of the system while meeting comfort requirements. The step of comprehensively determining the current real-time operation control status of the cluster based on the magnitude of the real-time temperature deviation, the standard deviation of the cluster's temperature fluctuation, and the determined control priority specifically includes: If the absolute value of the real-time temperature deviation is greater than the first deviation threshold, it is determined to be in a rapid control state. If the absolute value of the real-time temperature deviation is less than or equal to the first deviation threshold, then the standard deviation of the temperature fluctuation of the cluster is further calculated based on historical temperature data: If the standard deviation of the temperature fluctuation is greater than the fluctuation threshold, it is determined to be a fluctuation suppression state; If the control priority of this cluster is high, it is determined to be in high-precision monitoring state; The process involves, based on the determined real-time operation control state, calling upon a preset set of energy-saving control parameters that matches that state, and generating a differentiated operation control signal for the cluster. The preset energy-saving control parameter set for the rapid adjustment state focuses on enhancing the proportional control effect; The energy-saving control parameter set preset for the fluctuation suppression state focuses on enhancing the differential control effect; Furthermore, under the same real-time operation and control conditions, the energy-saving control parameter set configured for clusters with high control priority has a stronger differential control effect compared to the energy-saving control parameter set preset for clusters with normal priority. The method for obtaining the energy-saving control parameter set is as follows: During the initial system debugging phase, a preset parameter library based on experimental and simulation calibration is obtained; During the normal operation phase of the system, the parameters are dynamically fine-tuned based on the preset parameter library and a rule-based self-tuning mechanism. When the cumulative runtime of the system reaches the model training threshold, machine learning methods based on historical runtime data are used for optimization to obtain an optimization parameter library that is compatible with the current building and system characteristics.

2. The intelligent energy-saving coordinated control method for a multi-split central air conditioning system group according to claim 1, characterized in that, The process of dividing multiple indoor units into several control clusters with similar heat load change patterns based on the temperature data, the operating characteristics of each indoor unit, and the real-time operating mode, using a dynamic clustering algorithm, specifically includes: A multi-dimensional feature vector is constructed for each air-conditioned area corresponding to an indoor unit. The feature vector includes at least: the real-time average temperature of the area, the temperature uniformity index of the area, the thermal response coefficient of the indoor unit, and the spatial logical position weight in the current operating mode. Based on the multidimensional feature vectors corresponding to all indoor units, a density-based clustering algorithm is used for real-time dynamic grouping, and multiple indoor units whose feature vectors are close in distance in the feature space are grouped into the same control cluster. The clustering is recalculated in each control cycle or when the system detects a preset operating mode switching event, so as to achieve dynamic reorganization of the control cluster.

3. The intelligent energy-saving coordinated control method for a multi-split central air conditioning system group according to claim 2, characterized in that, The determination of the control cluster's regulation priority specifically includes: The control cluster is predefined based on the building function attributes of the air-conditioned area corresponding to the control cluster and the comfort requirements of the current operating mode. Among them, the clusters corresponding to the functional areas with high temperature stability requirements are preset as high priority. When the system detects or receives a signal that there is a temporary high load event in a specific area, the control priority of the control cluster corresponding to the area affected by the event is temporarily marked as the high priority; The temperature fluctuation standard deviation of clusters that are preset or marked as the normal priority is continuously monitored under steady state. If the temperature fluctuation standard deviation continues to exceed the preset adaptive adjustment threshold, the adjustment priority is automatically updated to the high priority.

4. The intelligent energy-saving coordinated control method for a multi-split central air conditioning system group according to claim 3, characterized in that, The acquisition of the preset parameter library based on experimental and simulation calibration specifically includes: On experimental platforms or high-precision system simulation models, various real-time operation control states are manually set or simulated for each typical system operation mode. Distinguish between high-priority and regular-priority regions through testing; Under each simulated operating condition combination, through experiments or simulation optimization, a set of energy-saving control parameters that balance response speed, temperature stability and optimal system energy efficiency are obtained. The optimized parameters are stored in the form of a multidimensional mapping table or configuration file to form an initial preset parameter library. The index keys of the preset parameter library include at least: running mode, real-time running control status, and control priority.

5. The intelligent energy-saving coordinated control method for a multi-split central air conditioning system group according to claim 4, characterized in that, The step of dynamically fine-tuning the parameters based on the preset parameter library and a rule-based self-tuning mechanism specifically includes: When the system determines that a control cluster is in a certain real-time operating control state for a period of time, if the temperature control effect in that area does not meet expectations, it triggers self-tuning of the energy-saving control parameters used by the current cluster, using a set of heuristic rules for dynamic fine-tuning. The specific steps are as follows: If the temperature approaches the set point too slowly, the proportional coefficient Kp will be increased by the preset step size. If the temperature overshoot is too large, decrease the proportional coefficient Kp by the preset step size or increase the differential coefficient Kd by the preset step size. If the steady-state deviation is eliminated slowly, the integral coefficient Ki is increased, but at the same time, it is monitored whether this will cause new oscillations. For regions where regulation is of high priority, increasing the differential coefficient Kd to suppress fluctuations takes precedence over increasing the proportional coefficient Kp.

6. The intelligent energy-saving coordinated control method for a multi-split central air conditioning system group according to claim 5, characterized in that, When the cumulative runtime of the system reaches the model training threshold, optimization is performed using machine learning methods based on historical runtime data to obtain an optimization parameter library adapted to the current building and system characteristics. Specifically, this includes: Machine learning is performed using historical data collected during system operation to build a parameter optimization knowledge base; For a given set of system operating states, a similarity search is performed based on the parameter optimization knowledge base to find the historical operating record that is closest to the current features, and the set of energy-saving control parameters in that record that has been proven to have the best control effect in practice is output.

7. The intelligent energy-saving coordinated control method for multi-split central air conditioning system groups according to claim 6, characterized in that, It also includes an adaptive monitoring and control mechanism, which specifically includes: For clusters under high-precision monitoring, increase the frequency of temperature monitoring and acquisition for their corresponding air-conditioned areas.

Images