A control method for constant temperature and current conversion in HVAC systems

Abstract

This application relates to the field of temperature control technology, and more particularly to a control method for HVAC thermostatic converters. The proposed scheme involves collecting circulating pump frequency commands and water temperature response data from multiple historical control cycles to construct the dynamic trajectory of the current water temperature in a temperature inertial field; combining this with the disturbance residual matrix to calculate the attractive potential energy state between the current water temperature and the target constant temperature value; and outputting the circulating pump frequency bandwidth based on this state to achieve dynamic adjustment. By introducing hydraulic connectivity to construct an adjacency matrix and a Laplace matrix, and combining this with a topological diffusion kernel to perform spatial convolution and weighted updates on the disturbance residual, accurate propagation and suppression of disturbances are achieved. This method can effectively address the nonlinear response of water temperature to pump frequency, delay effects, and disturbance coupling problems, improving the temperature stability and energy utilization efficiency of the heating system.

Description

Technical Field

[0001] This application relates to the field of temperature control technology, and in particular to a control method for constant temperature and current conversion in HVAC thermal systems. Background Technology

[0002] In existing HVAC systems, circulation pump frequency regulation largely relies on direct feedback based on the difference between real-time water temperature and target temperature. This lack of in-depth analysis and modeling of dynamic water temperature changes leads to lag and instability in the regulation results. Under conditions of load fluctuations or frequent external disturbances, system water temperature can easily deviate from the target value for extended periods, even causing continuous oscillations in local pipe networks, affecting heating quality. Furthermore, existing methods, when dealing with multi-node pipe networks, do not adequately consider the correlation of water temperatures at different locations. Disturbances can easily accumulate or propagate within the network, causing localized overheating or underheating, resulting in low energy efficiency. For switching control during operational phases, such as transitioning from initial heating to stable heating and then to low nighttime loads, existing technologies are insufficient in matching sampling frequency and regulation amplitude, leading to significant differences in temperature control performance at each stage and making it difficult to achieve stable, constant temperature throughout the entire cycle.

[0003] To address the above issues, this application presents a control method for HVAC thermostatic converter flow. Summary of the Invention

[0004] The technical problem this application aims to solve is to address the shortcomings of existing technologies by providing a control method for HVAC thermostatic converters. This method collects circulating pump frequency commands and water temperature response data from multiple historical control cycles to construct the dynamic trajectory of the current water temperature in a temperature inertial field. Combining this with the disturbance residual matrix, it calculates the attractive potential energy state between the current water temperature and the target isothermal value. Based on this state, it outputs the circulating pump frequency bandwidth range to achieve dynamic adjustment. By introducing hydraulic connectivity to construct an adjacency matrix and a Laplace matrix, and combining this with a topological diffusion kernel to perform spatial convolution and weighted updates on the disturbance residual, it achieves accurate propagation and suppression of disturbances. This method can effectively address the nonlinear response of water temperature to pump frequency, delay effects, and disturbance coupling problems, improving the temperature stability and energy utilization efficiency of the heating system.

[0005] To achieve the above objectives, this application provides the following technical solution:

[0006] A control method for constant temperature variable flow in HVAC systems, applied to a heating network, wherein the heating network includes a controller, a circulating pump, a water temperature sensor, and the network itself, the controller is equipped with a temperature inertial field and a disturbance residual matrix, wherein the temperature inertial field is simulated using virtual potential energy based on a preset target constant temperature value, and the disturbance residual matrix is ​​constructed based on the disturbance residual between water temperature changes and ideal changes within historical control cycles, the control method comprising:

[0007] The current water temperature is collected by the water temperature sensor, and the dynamic trajectory of the current water temperature in the temperature inertial field is constructed based on the pump frequency input and water temperature output changes of the circulating pump in multiple historical control cycles. The current water temperature is the water temperature sampling sequence collected in the current control cycle.

[0008] Based on the dynamic trajectory and the disturbance residual matrix, the attractive potential energy state between the current water temperature and the target isothermal value is calculated.

[0009] Based on the attractive potential energy state, the controller outputs a frequency bandwidth range to control the circulating pump to perform frequency adjustment within the frequency bandwidth range, thereby obtaining the pump frequency input of the circulating pump in the current control cycle. The width and center position of the frequency bandwidth range are jointly determined based on the attractive potential energy state and the disturbance residual matrix.

[0010] The construction of the dynamic trajectory of the current water temperature in the temperature inertial field includes:

[0011] Collect the circulation pump frequency control commands and corresponding water temperature response data within at least M control cycles before the acquisition, and construct a historical control state sequence arranged in chronological order. Each state unit in the historical control state sequence includes a timestamp, frequency increment value, water temperature change rate, and sampling interval duration.

[0012] Based on the historical control state sequence, the response delay distribution characteristics of water temperature in the heating network to frequency control actions are calculated, and the response offset function is calculated by fitting the residual curve.

[0013] The current water temperature is fitted with the response offset function to obtain a relative offset trajectory curve with the target constant temperature as the reference point, and a dynamic trajectory is output, wherein the relative offset trajectory curve is plotted with time on the horizontal axis and the output value of the response offset function on the vertical axis.

[0014] The disturbance residual matrix is ​​constructed based on historical measurement and prediction data from the previous k control cycles. The prediction data is a water temperature change sequence calculated by the controller based on the pump frequency input of the circulating pump and an ideal response model. The specific construction method is as follows:

[0015] After the kth control cycle ends, obtain the actual water temperature sampling sequence for the previous k control cycles. and the ideal water temperature sampling sequence predicted by the controller Where j represents a single control period, and the specific form of the sampling sequence is as follows:

[0016] ,

[0017] Where p represents the number of sampling points;

[0018] Based on the actual water temperature sampling sequence corresponding to each control cycle j and the ideal water temperature sampling sequence predicted by the controller, the residual vector is calculated by the difference.

[0019] Calculate based on the hydraulic connectivity of the heating pipe network. Given an adjacency matrix A of order A, calculate the corresponding Laplace matrix based on the adjacency matrix. , where D represents the degree matrix of the adjacency matrix;

[0020] Based on the preset diffusion coefficient Calculate the topological diffusion kernel , where e represents matrix exponentiation operation, and the residual vector is spatially convolved through the topological diffusion kernel to generate a residual field matrix;

[0021] The forgetting factor is calculated based on the pump frequency input corresponding to the control cycle. The residual field matrix is ​​then weighted using the forgetting factor, and the weighted residual field matrix is ​​summed to obtain the perturbation residual matrix.

[0022] The update of the disturbance residual matrix is ​​based on the pump frequency input of the previous control cycle and the current water temperature of the current control cycle. The specific update method is as follows:

[0023] Obtain the disturbance residual matrix of the output from the previous control cycle. ;

[0024] The forgetting factor of the disturbance residual matrix is ​​updated based on the pump frequency input of the previous control cycle to obtain the weighted value. ;

[0025] Calculate the residual field matrix for the current control cycle based on the current water temperature. The updated perturbation residual matrix is ​​calculated based on the perturbation residual matrix and weighting value of the previous control cycle and the residual field matrix of the current control cycle. The specific calculation method is as follows:

[0026] .

[0027] Based on the dynamic trajectory and the disturbance residual matrix, the attractive potential energy state between the current water temperature and the target isothermal value is calculated, including:

[0028] Based on the pump frequency input of the previous control cycle and the current water temperature of the current control cycle, the disturbance residual matrix output of the previous control cycle is updated to obtain the updated disturbance residual matrix.

[0029] Based on the dynamic trajectory, determine the deviation magnitude, rate of change, and trend direction of the current water temperature relative to the target constant temperature value, and calculate the corresponding trajectory feature vector;

[0030] The trajectory feature vector and the updated perturbation residual matrix are coupled to calculate the attractive potential energy state.

[0031] Couple the trajectory feature vector and the updated perturbation residual matrix, including:

[0032] Based on the trajectory feature vector, a disturbance-sensitive direction factor representing the current water temperature change trend is determined;

[0033] Extract the perturbation submatrix corresponding to the perturbation sensitive direction factor from the updated perturbation residual matrix;

[0034] The perturbation-sensitive direction factor and the perturbation submatrix are subjected to direction matching analysis to determine the influence relationship of residual perturbation on the current trajectory trend;

[0035] Based on the results of the directional matching analysis, the attraction stability index of the current water temperature on the control path is calculated to obtain the attraction potential energy state.

[0036] The width and center position of the frequency bandwidth are determined based on the attracted potential energy state and the perturbation residual matrix.

[0037] Based on the attractive potential energy state, the controller outputs a frequency bandwidth range, including:

[0038] Taking the frequency center value of the previous control cycle as a reference, the frequency center offset of the current control cycle is determined according to the state of the attracting potential energy, wherein the frequency center offset is inversely proportional to the state of the attracting potential energy.

[0039] Based on the disturbance intensity on the current control path in the disturbance residual matrix, the adjustment coefficient of the frequency bandwidth range is determined, wherein the adjustment coefficient is used to control the width convergence rate of the frequency bandwidth.

[0040] Based on the frequency center offset and the adjustment coefficient, combined with the frequency bandwidth range of the previous cycle, the corresponding frequency bandwidth range is output.

[0041] The method further includes:

[0042] The sampling period of the water temperature sensor is determined based on the operating period of the heating network.

[0043] The operating phases include an initial heating phase, a stable heating phase, and a nighttime load phase, with different target constant temperatures for each phase.

[0044] The sampling period of the water temperature sensor is determined based on the operating period of the heating network, including:

[0045] If the initial heating phase is underway, a first sampling period is set.

[0046] If the system is in a stable heating phase, a second sampling period is set, wherein the first sampling period is shorter than the second sampling period.

[0047] If the nighttime load phase is in progress, a third sampling period is set, wherein the second sampling period is shorter than the third sampling period.

[0048] The first sampling period, the second sampling period, and the third sampling period are adjusted based on the perturbation convergence rate in the perturbation residual matrix, the dynamic trajectory of the current water temperature in the temperature inertial field, and the offset of the historical adjustment of the circulation pump frequency.

[0049] Compared with the prior art, the beneficial effects of this application are:

[0050] This application achieves global perception and precise control of the water temperature status at multiple nodes in the heating network by introducing a temperature inertial field and a disturbance residual matrix into the controller, combined with dynamic water temperature trajectory analysis. It not only considers the water temperature deviation amplitude, rate of change, and trend direction simultaneously in frequency regulation, but also utilizes the disturbance residual matrix to suppress the propagation and accumulation of local disturbances, thereby significantly improving the stability of constant temperature control. Under load fluctuations or external disturbances, the water temperature can quickly return to the target constant temperature value, reducing local oscillations. Attached Figure Description

[0051] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0052] Figure 1 This is a flowchart illustrating a control method for HVAC thermostatic converter according to an embodiment of this application;

[0053] Figure 2 This is a schematic diagram illustrating an exemplary application scenario of an embodiment of this application;

[0054] Figure 3 This is a schematic diagram of the controller module in an embodiment of this application;

[0055] Figure 4 This is a schematic diagram of the matrix calculation process in an embodiment of this application;

[0056] Figure 5 This is a flowchart illustrating the dynamic trajectory generation method according to an embodiment of this application;

[0057] Figure 6 This is a schematic diagram of the dynamic trajectory of an embodiment of this application. Detailed Implementation

[0058] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0059] The term "embodiment" as used herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0060] To more clearly illustrate the control method of constant temperature and current conversion in HVAC thermal power according to this application, the following will elaborate on an optional specific implementation method of this application in conjunction with actual engineering scenarios.

[0061] In this exemplary technology, the method is applied to HVAC systems, primarily in the scenario of secondary pipe network loop regulation in centrally heated buildings. The system structure includes: a digital circulating pump, an outlet water temperature sensor, an automatic controller, and the heating pipe network. The system's operational objective is to adjust the operating frequency of the digital circulating pump in real time during different operating periods, ensuring that the temperature of the hot water in the pipe network is maintained as stably as possible within the target constant temperature range, while simultaneously reducing energy waste and the risk of overshoot.

[0062] The specific adjustment logic of the above example technology is as follows:

[0063] When the system starts up for the first time, the controller first completes the following initialization steps:

[0064] Frequency setting initialization: The default operating frequency of the digital circulating pump is set to 50Hz, which serves as the reference value for the controller's initial judgment logic. This frequency value is calculated by the system operating condition designer based on the building's heat load, return water flow rate, and initial heat balance, and can be adjusted appropriately based on system commissioning parameters.

[0065] The system divides 24 hours into several fixed time periods, preferably one hour, and sets the temperature sampling frequency within each period, for example, collecting the current outlet water temperature every 5 minutes. This time granularity allows the controller to determine the temperature change trend within each time period and provides data support for subsequent frequency adjustment strategies.

[0066] Logical variables are initialized, including the water temperature sampling record queue, frequency record array, and historical target frequency cache table for each time period, to support the feedforward prediction logic and mutation compensation mechanism for frequency updates.

[0067] After the system enters a stable operating state, the overall operation flow of the control method is as follows:

[0068] The controller continuously collects the actual water temperature during the current period through a water temperature sensor and records the sampled values ​​into a local cache according to a set sampling period (e.g., every 5 minutes) to form a water temperature change sequence for the current period. Simultaneously, it records the actual frequency of the digital circulation pump corresponding to each sampling point, providing input reference for subsequent frequency adjustment decisions.

[0069] Before the start of each new time period, the controller uses the last sampled water temperature data of the current time period to predict the target frequency of the circulation pump required for the next time period by invoking a set of preset adjustment rules. The core idea of ​​the prediction logic is: if the current water temperature is lower than the set value, the frequency should be appropriately increased to increase the hot water flow rate and enhance heating; conversely, if the current water temperature is too high, the pump speed should be appropriately reduced to prevent overshoot.

[0070] Understandably, the adjustment rule here is a simplified engineering experience-based adjustment logic. Essentially, it utilizes the deviation between the current water temperature and the reference value to calculate the frequency target by adjusting a multiplicative factor. This calculation does not use complex mathematical modeling; instead, it obtains the controller's frequency output strategy for the next stage through linear approximation and adjustment factor correction.

[0071] In addition to feedforward predictive control, the system also has an instant response strategy based on threshold judgment. That is, if the system detects a change in water temperature of more than 3°C compared to the initial temperature of the same period during operation, it immediately triggers a frequency update mechanism, recalculates the target pump speed frequency, and instructs the circulating pump to adjust its frequency.

[0072] In one optional implementation, the system automatically distinguishes operating phases based on different times of the 24 hours of the day and adjusts the control strategy in conjunction with heat load patterns.

[0073] During the initial heating phase, the system's heat load rapidly increases from its overnight low, requiring quick indoor heating. The controller is initially set to a higher frequency range (e.g., 50Hz) and the water temperature sampling cycle is shortened to 2 minutes to enhance system response sensitivity and ensure rapid establishment of thermal equilibrium.

[0074] During the stable heating phase, when heat load fluctuations are relatively mild, the controller can employ a predictive frequency strategy, combined with historical frequency operating trajectories, to optimize pump speed and reduce energy consumption while maintaining a stable target temperature. The sampling period is restored to 5 minutes.

[0075] During the nighttime load phase, as the user's heat load decreases, the water temperature changes tend to stabilize. The controller will set a lower frequency target value (such as below 40Hz) and can extend the sampling period to 10 minutes to reduce system energy consumption and wear caused by frequent adjustments.

[0076] like Figure 1 As shown, to address the issue of nonlinear response of water temperature to pump frequency, this application provides a control method for HVAC thermostatic converters, applied to heating pipe networks. The heating pipe network includes a controller, a circulating pump, a water temperature sensor, and the pipe network itself. The controller is equipped with a temperature inertial field and a disturbance residual matrix, including:

[0077] S1: The current water temperature is collected by the water temperature sensor, and the dynamic trajectory of the current water temperature in the temperature inertial field is constructed based on the pump frequency input and water temperature output changes of the circulating pump in multiple historical control cycles.

[0078] In this embodiment, a water temperature sampling sequence within the current control cycle is acquired using a water temperature sensor. This sampling sequence covers representative measuring points at different locations within the pipe network and is combined with historical data on pump frequency input and water temperature output changes from multiple control cycles to generate the dynamic trajectory of the current water temperature in a temperature inertial field. The temperature inertial field is a virtual potential energy distribution model, constructed considering factors such as the target isothermal value, response delay distribution, and water temperature change rate, used to describe the inertial offset trend of water temperature changes to the control input.

[0079] Understandably, the water temperature sampling sequence avoids the drawbacks of adjusting solely based on instantaneous deviations, and can reflect the overall trend and potential stable path of water temperature changes while taking into account the system's thermal inertia and delay effects.

[0080] Those skilled in the art will understand that the type of operational data collection can be selected according to the actual working conditions. For example, it can be single-point distribution measurement, multi-point distribution measurement, or include additional flow and pressure information, as long as it can be guaranteed to the minimum that the dynamic trajectory of water temperature can be derived from it. This application does not impose any restrictions on this.

[0081] S2: Based on the dynamic trajectory and the disturbance residual matrix, calculate the attraction potential energy state between the current water temperature and the target constant temperature value;

[0082] In this embodiment, the disturbance residual matrix reflects the deviation distribution between water temperature changes and the ideal response within historical control cycles. It is processed through topological diffusion of the hydraulic connectivity of the heating network to reveal the spatial propagation characteristics of disturbances at different locations. In the calculation of the attractive potential energy state, the direction, amplitude, and rate of change of the current water temperature offset are described by the trajectory feature vector, while the disturbance residual matrix is ​​used to correct the spatial and historical stability predictions of this trajectory, thereby obtaining a comprehensive and quantifiable attractive potential energy state index.

[0083] Understandably, this application considers not only the current deviation value but also historical disturbance patterns and spatial correlations, making it particularly suitable for complex scenarios with uneven pipeline loads and mutual temperature influences among nodes. The specific construction method of the disturbance residual matrix can be adjusted according to the connectivity structure of the pipeline and computational resources, as long as it can reflect the spatial and temporal coupling effects of the disturbance.

[0084] S3: Based on the state of the attraction potential energy, the controller outputs a frequency bandwidth range to control the circulating pump to perform frequency adjustment within the frequency bandwidth range, thereby obtaining the pump frequency input of the circulating pump in the current control cycle.

[0085] In this embodiment, the center position of the frequency bandwidth is determined jointly by the attractive potential energy state and the perturbation residual matrix, while its width considers the balance between perturbation intensity and attractive stability index. When the attractive potential energy state indicates that the current temperature change has high stability, the bandwidth can be appropriately narrowed to reduce frequency fluctuations; conversely, the bandwidth is expanded to improve response speed. Finally, the circulating pump performs real-time frequency adjustment within this bandwidth range to achieve rapid and stable approximation of the target isothermal value.

[0086] Those skilled in the art will understand that the bandwidth range can be achieved by the controller through software parameter configuration, or by combining it with a hardware limiting device; this application does not impose any limitations on this.

[0087] This application is applicable to HVAC systems with dynamic heat load fluctuations and nonlinear thermal response characteristics, and is especially applicable to water circulation heating networks with backflow path coupling, significant spatial temperature inertia, frequency control lag, and distributed disturbance conduction characteristics.

[0088] Understandably, when such systems are used for constant temperature control, traditional single-point feedback strategies and static mapping models cannot effectively respond to the control errors caused by the inertia of the pipeline structure, residual load disturbances, and historical response delays, which can easily lead to water temperature fluctuations, overshoot, or energy consumption redundancy.

[0089] Application scenarios include, but are not limited to:

[0090] A district heating system with complex pipe layout and multi-point heat source distribution;

[0091] A floor radiant heating system with a dendritic or ring-shaped hydraulic topology and a significant time delay in the return water path feedback.

[0092] A smart building thermal control system that maintains high responsiveness and stability to small disturbances during low-load operation (such as constant temperature at night);

[0093] Office buildings or factory heating networks that require concurrent frequency control of multiple circulating pumps and have uneven local adjustment feedback.

[0094] The control method proposed in this application addresses common system characteristics that do not rely on a single control strategy, nor are they based on a fixed thermal load model or static pipeline parameter configuration. Instead, it constructs a dynamic temperature inertial field to spatially attract and model the target isothermal state, and uses the disturbance residual matrix to characterize the residual propagation pattern of the control deviation in the topology, thereby achieving variable frequency regulation control.

[0095] Applicable application scenarios must meet at least one of the following structural or response characteristics:

[0096] Temperature control response has a significant hysteresis characteristic, making it difficult to achieve precise closed-loop control in real time;

[0097] The heat load changes show short-term, drastic disturbances, but the overall trend is gradual.

[0098] Water temperature sampling signals are affected by the location distribution of the pipeline network or the configuration of the sampling cycle, resulting in unstable or delayed feedback.

[0099] The frequency regulation of the controller indirectly drives changes in water temperature, and there is a significant intermediate thermal inertia link.

[0100] In system operation scenarios, it is not advisable to use models with large-frequency mutations or high-speed responses to ensure operational stability and pump life.

[0101] It should be noted that the control method proposed in this application does not rely on strong assumptions about the topology of the heating network, nor does it require absolute decoupling or feedback consistency between circulating pumps. Instead, it achieves stable guidance of the target water temperature of the system by attracting potential energy fields and suppressing residual disturbances, while ensuring the dynamic flexibility of regulation.

[0102] It should be noted that the disturbance residual matrix involved in this application is not a weight or correction matrix used for static compensation in the traditional sense, but a dynamic disturbance characteristic field with self-updating and spatiotemporal diffusion characteristics. It can adapt to the migration law of disturbance behavior in different control cycles, thereby realizing the continuous convergence control of nonlinear thermal systems under disturbance.

[0103] Similarly, the frequency bandwidth adjustment described in this application is not a simple frequency conversion algorithm based on single-point frequency modulation, but rather avoids the overshoot or oscillation problems caused by traditional frequency up or down adjustment strategies when facing inertial lag, based on the analysis of the attractive potential energy state and the disturbance residual quantity.

[0104] See Figure 2 This figure is a schematic diagram of an exemplary application scenario provided by an embodiment of this application.

[0105] Figure 2 The application scenarios shown include a heat source 100, a circulating pump 101, multiple water temperature sensors 102 and a controller 103, and a pipeline body 104.

[0106] In this embodiment, the heat source 100 can be a device such as a boiler, heat exchange unit, or centralized heating station that can continuously provide heat medium, used to deliver hot water or other heat transfer medium with a preset temperature into the pipe network body 104. A circulating pump 101 is installed between the heat source 100 and the pipe network body 104, used to realize the continuous delivery and flow regulation of the heat medium in the pipe network body 104 under the regulation of the controller 103. Multiple water temperature sensors 102 are respectively deployed at key node locations in the pipe network body 104, such as on each user branch or return water pipe, used to collect water temperature data at different locations in real time to reflect the operating thermal distribution of the entire heating system.

[0107] Figure 2 The controller 103 is further shown to be electrically connected to the circulating pump 101 and multiple water temperature sensors 102. It is used to receive water temperature data collected by each water temperature sensor 102, and calculate the attraction potential energy state between the current water temperature and the target constant temperature value based on the data and a preset temperature inertial field model and disturbance residual matrix. Then, according to the state, the controller outputs the corresponding frequency bandwidth range of the circulating pump 101 to realize dynamic adjustment of the flow rate.

[0108] It is understood that the pipeline network 104 may include main pipelines, branch pipelines, user-end heat exchange devices, etc., and its pipeline layout can be a parallel structure, a ring structure, or a hybrid structure to adapt to different heating load demands. The control method of this application can adapt to various pipeline network topologies and can achieve rapid heating, stable constant temperature, and energy-saving operation at night by adjusting the operating status and frequency range of the circulating pump 101 under different operating periods, thereby effectively reducing energy consumption and improving heating quality.

[0109] See Figure 3 The figure is a schematic diagram of the controller module provided in an embodiment of this application.

[0110] Figure 3The controller 103 is shown to include a water temperature data acquisition module 1031, a dynamic trajectory calculation module 1032, a disturbance residual matrix processing module 1033, and a frequency adjustment control module 1034.

[0111] The water temperature data acquisition module 1031 is used to interact with the water temperature sensors 102 deployed at key nodes of the heating pipe network to complete the real-time acquisition and caching of water temperature sampling data within the current control cycle.

[0112] In this embodiment, the water temperature data acquisition module 1031 and the water temperature sensor 102 can use a multi-channel high-speed analog-to-digital converter interface in conjunction with an isolated signal conditioning circuit to achieve synchronous sampling and anti-interference processing of temperature signals at different measuring points, ensuring the consistency of acquisition accuracy and time.

[0113] The dynamic trajectory calculation module 1032 is used to construct the dynamic trajectory of the current water temperature in the temperature inertial field based on the water temperature sampling data and the preset temperature inertial field model.

[0114] In this embodiment, the dynamic trajectory calculation module 1032 can perform computational tasks such as delay response modeling, residual fitting and trajectory curve generation through an embedded digital signal processor or floating-point arithmetic unit to obtain the trajectory of the current control cycle.

[0115] The disturbance residual matrix processing module 1033 is used to construct a residual vector based on the ideal response prediction value of the circulating pump frequency input and water temperature sampling sequence after each control cycle, generate a topological diffusion kernel in combination with the hydraulic connectivity of the heating pipe network, perform spatial convolution on the residual vector to form a disturbance residual matrix, and perform weighted updates based on the forgetting factor.

[0116] In this embodiment, the perturbation residual matrix processing module 1033 can run on a multi-core microprocessor or an FPGA platform with a matrix operation optimized instruction set to achieve efficient matrix calculation and real-time updates.

[0117] The frequency adjustment control module 1034 is used to calculate the attraction potential energy state between the current water temperature and the target constant temperature value based on the trajectory feature vector output by the dynamic trajectory calculation module 1032 and the updated disturbance residual matrix output by the disturbance residual matrix processing module 1033, and generate frequency bandwidth range control commands for the circulating pump accordingly.

[0118] In this embodiment, the frequency regulation control module 1034 can schedule and execute the closed-loop control algorithm through the real-time operating system, and output control signals to the circulating pump actuator through the pulse width modulation or frequency conversion drive interface.

[0119] It is understood that the controller 103 in this embodiment can be an embedded control platform with industrial protection level, such as using an ARM Cortex-A series multi-core processor, equipped with large-capacity DDR memory and non-volatile memory, to support data buffering and multi-round iterative calculation of historical control cycles during long-term operation. At the same time, the controller 103 can also interact with the host computer monitoring system through Ethernet, RS-485 or wireless communication modules to facilitate remote monitoring and optimization adjustment of operating parameters. This application does not make specific limitations.

[0120] Before detailing the specific technical aspects of the steps, this application must reiterate:

[0121] In typical heating network operation, the response of water temperature to the frequency adjustment of circulating pumps is not linear, but is affected by multiple factors such as network thermal inertia, fluid transport delay, and the superposition of load disturbances. Especially in operating conditions with long pipelines, complex branch structures, or frequent load fluctuations at the user end, the trend of water temperature change often exhibits significant lag and overshoot. This makes traditional control strategies based on real-time deviation feedback prone to over-response or hysteresis during frequency adjustment, thus making it difficult to maintain a stable constant temperature for extended periods.

[0122] Furthermore, due to the hydraulic connectivity between nodes in the pipeline network, load disturbances at one location may propagate along the topological path to other nodes, causing a cumulative effect of local temperature fluctuations. If the control strategy fails to identify and dynamically correct the spatial diffusion behavior of these disturbances, a situation often arises where frequent adjustments are made to a single control point, yet the overall temperature distribution remains uneven.

[0123] Based on the aforementioned problems, this application is not limited to directly adjusting the PID control algorithm for a single temperature measurement point, but rather adopts a dual modeling approach using the temperature inertial field and the disturbance residual matrix:

[0124] The temperature inertial field is used to characterize the time-delay characteristics of water temperature and the attraction trend of the target constant temperature value, i.e., the dynamic trajectory; the disturbance residual matrix, by analyzing the difference between the actual water temperature and the ideal response in historical control cycles and combining it with the hydraulic connectivity topology of the pipe network, forms a dynamic data structure that can reflect the spatial diffusion and attenuation law of disturbance.

[0125] Understandably, during control decision-making, the dynamic trajectory and the residual disturbance matrix participate in the calculation in parallel. By calculating the potential energy state, it can reflect the trend information of the current temperature change and superimpose the spatial influence of historical disturbances, thereby achieving higher adaptability and predictability in the setting of frequency bandwidth.

[0126] It is important to note that this application does not rely on a fixed network zoning structure or a single mathematical control model, but rather addresses heating systems with significant response delays, strong spatial coupling, and high randomness in load disturbances. The advantage of this application lies in its ability to identify potential trends that could lead to overshoot or hysteresis during control calculations by coupling the spatial gradient direction of the disturbance with the dynamic trajectory feature vector. This allows for proactive compensation and limiting in frequency bandwidth allocation. This approach avoids the limitations of relying solely on the magnitude of temperature deviations, enabling high temperature stability and regulation accuracy even in complex scenarios involving multiple users, long pipelines, and multiple branches.

[0127] Next, the principle of the method in this application regarding the perturbation residual matrix will be further elaborated.

[0128] It is easy to understand that the perturbation residual matrix in this application includes two parts: construction and updating.

[0129] In the construction phase, the operational data from the first k historical control cycles are first comprehensively processed. This data includes the pump frequency input of the circulating pump and the corresponding actual water temperature sampling sequence for each cycle. Simultaneously, this is combined with the water temperature change sequence calculated by the controller based on the ideal response model. The difference between these two data points yields a residual vector reflecting the magnitude and direction of the deviation. This residual vector not only records the temperature deviation over time but is also mapped into a residual field matrix reflecting the propagation relationship of disturbances along spatial paths in subsequent processing, taking into account the hydraulic connectivity topology of the heating network. A topological diffusion kernel operation is introduced during the construction of the residual field matrix, allowing the diffusion and attenuation patterns of disturbance effects to be observed between adjacent or hydraulically highly coupled nodes. This gives the disturbance residual matrix the ability to express spatial correlation.

[0130] The ideal response model is constructed based on the factory settings of the heating network.

[0131] Understandably, when this application is initially put into operation or when the control method is first activated, the controller does not contain a disturbance residual matrix and a temperature inertial field formed through historical data accumulation and calculation. This is because these two core data structures rely on the collection, processing, and modeling of real operating data of the heating network in different control cycles to form a representative mathematical description. The disturbance residual matrix needs to be calculated based on the difference between the actual water temperature change and the ideal predicted water temperature change in multiple historical control cycles, while the temperature inertial field needs to be established by fitting the delay characteristics, changing trends, and potential energy distribution between the target constant temperature value and the actual water temperature response. In the initial stage without historical operating records, this information cannot be obtained directly. Therefore, it is necessary to gradually generate an initial version of the disturbance residual matrix and temperature inertial field through data accumulation, calculation, and iterative updates in the first few control cycles to provide a reliable reference basis for subsequent attraction potential energy calculation and frequency adjustment.

[0132] In this application, the specific value of k can be determined by those skilled in the art through experiments, but at least three control cycles are required, and this application does not impose too many restrictions here.

[0133] In one example, the construction of the perturbation residual matrix includes the following steps:

[0134] After the kth control cycle ends, obtain the actual water temperature sampling sequence for the previous k control cycles. and the ideal water temperature sampling sequence predicted by the controller Where j represents a single control period, and the specific form of the sampling sequence is as follows:

[0135] ,

[0136] Where p represents the number of sampling points;

[0137] Specifically, after the k-th regulation cycle ends, it is necessary to simultaneously acquire two types of data from the previous k regulation cycles: one is the actual water temperature sampling sequence, and the other is the water temperature change sequence calculated based on the ideal response model. By comparing the actual operating results with the calculation results of the ideal response model, the deviation characteristics in the regulation process can be quantitatively reflected, thereby providing basic data for subsequent disturbance analysis.

[0138] In this embodiment, the sampling sequence is recorded sequentially according to the time order within the control period, and each sampling point is accompanied by a timestamp to ensure consistency between subsequent residual calculation and time-series response analysis. The predicted value of the ideal response model is calculated based on the pump frequency input of the circulating pump within the corresponding control period, using a predetermined thermodynamic transfer function and hydraulic characteristic curve, thereby ensuring consistency between the predicted data and the actual data under the input conditions.

[0139] Based on the actual water temperature sampling sequence corresponding to each control cycle j and the ideal water temperature sampling sequence predicted by the controller, the residual vector is calculated by the difference.

[0140] In this embodiment, a residual vector is formed by calculating the difference between the two types of sampling sequences at each time point. The residual vector not only reflects the amplitude deviation between the water temperature change and the ideal state, but also retains the phase information of the fluctuation, which is convenient for subsequent judgment of whether the disturbance is lagging or leading.

[0141] In one optional implementation, to reduce sampling noise interference when forming the residual vector, a window width smaller than the control period length can be set. The residual vector is then weighted and averaged, with the weight distribution exhibiting a centrally symmetric structure, centered on the current sampling point and decreasing in weight for neighboring points. This preserves the main trend of the disturbance while effectively suppressing high-frequency noise caused by sensor accuracy issues, communication delays, or transient external disturbances.

[0142] Calculate based on the hydraulic connectivity of the heating pipe network. Given an adjacency matrix A of order A, calculate the corresponding Laplace matrix based on the adjacency matrix. , where D represents the degree matrix of the adjacency matrix;

[0143] Specifically, the adjacency matrix A is used to fix the concept of which measuring points in the pipeline network are directly hydraulically connected to each other in a square matrix, which facilitates the subsequent quantitative description of the spatial propagation relationship of disturbances. During construction, the set of nodes is first determined, that is, the measuring points or calculation nodes participating in the analysis (such as main pipe nodes, branch inlets, terminal heat exchange points, etc.). Then, based on the as-built drawings, 3D model, valve opening, one-way valve direction, and on-site flow direction markings, it is determined whether there is a reachable direct hydraulic connection between two nodes: if the two points are directly connected through the same pipe section, or form a continuous path through the shell side or pipe side of the same equipment, then the corresponding row and column position in the adjacency list is marked as connected; otherwise, it is marked as not connected.

[0144] Furthermore, for branches with a clear master-slave flow direction, the connectivity can be set to be direction-sensitive: upstream to downstream is marked as connected, and the reverse direction is marked as disconnected if it is blocked or throttled.

[0145] Understandably, the adjacency matrix A is essentially a computable connected list of the pipeline network skeleton, providing a foundation for subsequent diffusion, coupling, and weight allocation. In the specific implementation, the disturbance residue does not remain only at its generation location but also diffuses or attenuates along the actual hydraulic path. Only by explicitly defining the connectivity in the matrix can subsequent spatial processing steps follow physically reachable paths, avoiding spurious propagation across pipes and sections, and improving the reliability and interpretability of subsequent calculations.

[0146] Furthermore, the degree matrix D is used to record the number of direct connections between each node and the outside world, or the overall connectivity strength. In actual generation, only the statistical values ​​of each row (or column) of the adjacency matrix are taken and filled into the main diagonal, while the remaining positions are left empty. Intuitively, this can be understood as:

[0147] If a node has a certain number of pipes connected to it and a certain number of direct external entrances and exits, then write that number in the corresponding dimension; if weighted connectivity is used (e.g., thick pipes are weighted more, long distances are weighted less, and valves that are half open are weighted less), then the weights are accumulated and written together.

[0148] Furthermore, the Laplace matrix can be viewed as a constraint table:

[0149] Each node's own cell is filled with its total connectivity, while for those nodes directly connected to it, a connectivity marker with the opposite meaning is filled in the corresponding position (indicating that there is a coupling or constraint relationship between the two), and the remaining unconnected pairs are left empty.

[0150] Understandably, the purpose of the constraint table is to solidify the laws of energy conservation and proximity coupling when performing subsequent calculations on perturbation diffusion and coupling.

[0151] In layman's terms, the extent to which a node can distribute residual disturbances to its surroundings, and to which nodes they will be distributed, is entirely determined by the structure of this table.

[0152] refer to Figure 4 , Figure 4 This is a schematic diagram of the matrix calculation process in an embodiment of this application.

[0153] Figure 4 Assume there is a secondary main pipe that connects to three residential branches in sequence, labeled N1 (upstream measuring point of the main pipe), N2 (first household branch entrance), N3 (second household branch entrance), N4 (third household branch entrance), and N5 (main pipe return side measuring point). N1 and N2 are directly connected through the same pipe, N2 and N3 are connected similarly, N3 and N4 are connected similarly, and N4 and N5 are connected similarly. Each household branch entrance is only connected to the main pipe, and the internal parts of the branch are not included in this connectivity determination.

[0154] Figure 4 The adjacency matrix is ​​further illustrated, and its meaning can be defined as follows: N1 is connected to N2, but not connected to N3 / N4 / N5; N2 is connected to N1 and N3, but not connected to N4 / N5; N3 is connected to N2 and N4, but not connected to N1 / N5; N4 is connected to N3 and N5, but not connected to N1 / N2; N5 is connected to N4, but not connected to N1 / N2 / N3.

[0155] In embodiments not shown in the figure, if the main pipe has a clear unidirectional flow direction (e.g., from N1 to N5), the upstream to downstream can be marked as strong connection, and the reverse connection can be set as weak or none depending on whether there is a check valve or throttling device on site.

[0156] Figure 4 The degree matrix is ​​further shown. The main diagonal of the degree matrix only needs to be written with the number of nodes that are connected: N1 is 1, N2 is 2, N3 is 2, N4 is 2, and N5 is 1.

[0157] In embodiments not shown in the figure, if weighted connectivity is used, for example, the connectivity weight of a large-diameter main pipe is set to 1, while a certain segment is only taken as 0.5 because the valve is half open, then the accumulated weight can be written to the main diagonal of the corresponding node. This application will not elaborate on this.

[0158] Figure 4 The Laplace matrix is ​​further illustrated, and its meaning can be understood as follows:

[0159] The corresponding degree is filled in the cell corresponding to each node, while the positions directly connected to it are filled with a connectivity marker with the opposite meaning (indicating that the two are mutually constrained), and the unconnected positions are left blank.

[0160] It is easy to understand that if there is overheating residue at N2, it will first affect N1 and N3; if the valve at N5 is temporarily closed, after the connection is updated, N4 will no longer release the disturbance to N5, and the residue will be transmitted back and forth between N3 and N4 more.

[0161] Based on the preset diffusion coefficient Calculate the topological diffusion kernel , where e represents matrix exponentiation operation, and the residual vector is spatially convolved through the topological diffusion kernel to generate a residual field matrix;

[0162] Specifically, the so-called topological diffusion kernel is used to propagate and attenuate the temperature residuals generated at one or more nodes during the previous control cycle along the actually reachable hydraulic connectivity path. This ensures that the residual effects are not confined to the source point but are gradually distributed among neighboring nodes according to their proximity, accessibility, and resistance magnitude based on the pipe network connections. Considering the significant directionality and resistance differences in the secondary heating network, the diffusion kernel calculation is based on weighted connectivity. The weights can comprehensively consider factors such as the equivalent hydraulic resistance of the pipe section, the design flow rate ratio, valve opening, and historical throughput. For branches with unidirectional checkpoints or obvious mainstream directions, a larger reachability weight is given in the direction from upstream to downstream, while a smaller weight or direct shielding is given in the reverse direction, thus making the diffusion behavior consistent with actual hydraulics.

[0163] In this embodiment, the numerical calculation of the topological diffusion kernel is performed using scaling and rational approximation, which have good numerical stability. This method is suitable for sparse, large-scale connected matrices, and specifically includes the following two aspects:

[0164] In the first aspect, the connectivity matrix, i.e., the Laplacian matrix, is scaled by the largest eigenvalue and decomposed into multiple small steps of equivalent diffusion. Then, within each step, a rational approximation is used to calculate the diffusion operator, ultimately yielding a symmetric or near-symmetric kernel matrix. To reduce online overhead, the topology diffusion kernel can be cached based on the topology version number: when the valve position, branch opening / closing, or flow direction configuration remains unchanged, the diffusion kernel is reused across multiple control cycles; when a change in connectivity is detected or the diffusion coefficient is adjusted by the strategy, incremental recalculation is triggered, updating only within the affected rows and columns.

[0165] Secondly, the value of the diffusion coefficient is determined by two types of information: one is the runtime period, which provides a basic range; the other is online state variables (the rate of water temperature change and the convergence trend of frequency bandwidth in the current cycle), which are used to make further corrections within the basic range. For example, in the initial heating stage when a high rate of temperature change and rapid convergence of frequency bandwidth are detected, the diffusion coefficient is shifted upward, allowing the residual effects to be mapped to downstream nodes more quickly, facilitating early convergence of the control strategy; in the nighttime load stage when the residual disturbance decays slowly, the diffusion coefficient is shifted downward, making the diffusion more localized and avoiding excessive smoothing that masks small but persistent local deviations.

[0166] It is understandable that by performing a topological diffusion on the kernel matrix through the diffusion coefficient, a set of diffusion residual weights can be obtained. This set of weights represents how the previous deviation will affect the surrounding nodes along the network if no new control action is applied in this cycle.

[0167] Furthermore, based on the aforementioned residual weights, a residual field matrix is ​​generated by combining the residual vectors. The residual field matrix reflects the residual influence strength of each node on itself at the diagonal position, and reflects the coupling strength between any two nodes due to connectivity and proximity at the off-diagonal position. The residual field matrix obtained in this way has clear characteristics: on the one hand, it maintains the correspondence with the source residual amplitude; on the other hand, it explicitly encodes the spatial coupling, providing a direct quantitative basis for subsequent attraction potential energy assessment and frequency bandwidth contraction. It can identify disturbances that will be amplified or backflowed at neighboring nodes earlier, reducing over-adjustment or repeated adjustment in the next cycle.

[0168] For example, the N1 to N5 cascaded trunking in the aforementioned example illustrates the intuitive results after diffusion and convolution.

[0169] Assuming an overheating residual occurs at position N2, and the valve remains fully open with the main current direction pointing from N1 to N5, after diffusion kernel processing, the remaining influence of N2 is still primarily its own, but a portion will be allocated to N1 and N3. Due to the larger weight of the main current direction, the share allocated to N3 is slightly higher than that to N1; N4 and N5 receive only a very small share or are essentially unaffected. Using this set of diffused weights to generate the residual field matrix reveals a high coupling strength between N2 and N3, suggesting that the adjustment strategy should prioritize tightening the frequency bandwidth and controlling the adjustment slope within the N2 to N3 region to prevent the disturbance from being further propagated downstream.

[0170] Furthermore, if valve N5 is temporarily closed at this time, the diffusion nucleus changes accordingly after the connection relationship is updated, and N4 no longer releases residuals to N5. The high value region of the residual field matrix then gathers between N3 and N4. Based on this, the control strategy shifts the focus of amplitude limiting and slow adjustment forward to avoid pushing the disturbance to the closed end.

[0171] The forgetting factor is calculated based on the pump frequency input corresponding to the control cycle. The residual field matrix is ​​weighted and assigned values ​​through the forgetting factor. The perturbation residual matrix is ​​obtained by summing the weighted residual field matrix.

[0172] Specifically, the forgetting factor is introduced to reflect the attenuation pattern of the impact of historical disturbances on current control decisions. Since the operating environment and load characteristics of the heating network change over time, the residuals of different control cycles have different reference values ​​for the current cycle: the earlier the cycle, the weaker the direct impact of its disturbances on the current state; conversely, the disturbances of the most recent cycle are better able to reflect the inertia and trend of the current operating conditions.

[0173] In this embodiment, the weighting process employs element-wise matrix multiplication, multiplying the k residual field matrices by their corresponding forgetting factors to calculate the decayed residual distribution. Subsequently, all weighted residual field matrices are superimposed in chronological order to obtain a perturbation residual matrix that comprehensively reflects the distribution characteristics of multi-period perturbations. This perturbation residual matrix not only retains the perturbation pattern of the latest period but also incorporates the residual influences of previous periods. Furthermore, the forgetting factor automatically suppresses the weight of outdated information, resulting in a comprehensive description that is both timely and continuous. When assessing the attractive potential energy state, a perturbation map that includes both historical trends and emphasizes recent conditions can be used to more accurately determine whether a perturbation in a certain region is an occasional fluctuation or a persistent problem, avoiding over-adjustment due to short-term anomalies and preventing the neglect of long-term local deviations.

[0174] In the update section, the disturbance residual matrix is ​​not simply replaced. Instead, the existing disturbance residual matrix is ​​weighted and attenuated by updating the forgetting factor related to the pump frequency input of the control cycle, so that the influence of historical disturbances gradually weakens over time. At the same time, combined with the actual water temperature sampling data of the current control cycle, a new residual field is calculated and superimposed on the attenuated matrix to form an update result that can dynamically reflect the recent disturbance state.

[0175] In one example, the update of the disturbance residual matrix is ​​based on the pump frequency input of the previous control cycle and the current water temperature of the current control cycle, and the specific update method is as follows:

[0176] Obtain the disturbance residual matrix of the output from the previous control cycle. ;

[0177] The forgetting factor of the disturbance residual matrix is ​​updated based on the pump frequency input of the previous control cycle to obtain the weighted value. ;

[0178] Calculate the residual field matrix for the current control cycle based on the current water temperature. The updated perturbation residual matrix is ​​calculated based on the perturbation residual matrix and weighting value of the previous control cycle and the residual field matrix of the current control cycle. The specific calculation method is as follows:

[0179] .

[0180] The calculation of the residual field matrix, forgetting factor, and perturbation residual matrix has been described above and will not be repeated here. It should be noted that the residual vector of the current water temperature can be calculated from the target isothermal value, without the need for additional water temperature prediction steps.

[0181] Next, we will further elaborate on the dynamic trajectory part of the method in this application.

[0182] For reference Figure 5 , Figure 5 This is a flowchart illustrating the dynamic trajectory generation method according to an embodiment of this application.

[0183] In one example, constructing the dynamic trajectory of the current water temperature in a temperature inertial field includes:

[0184] S1.1: Collect the circulation pump frequency control commands and corresponding water temperature response data within at least M control cycles, and construct a historical control state sequence arranged in chronological order. Each state unit in the historical control state sequence includes a timestamp, frequency increment value, water temperature change rate, and sampling interval duration.

[0185] It should be noted that the value of M here can be set by experiment, but it must satisfy that M is less than k+1;

[0186] Specifically, the purpose of constructing a historical control state sequence is to quantify and record the dynamic relationship between the regulation behavior of the circulating pump and water temperature changes over multiple control cycles, thereby providing calculable input for subsequent response delay analysis.

[0187] In this embodiment, the acquisition of historical control state sequences not only records the absolute frequency value of the circulating pump, but also uses the frequency change in each cycle as the main indicator to eliminate fixed deviations under different base loads, focusing more on the impact of the control action itself. Simultaneously, the water temperature change rate is obtained by dividing the temperature difference between adjacent sampling points by the duration of the sampling interval, which can intuitively reflect the rate of temperature rise or fall caused by the control action. The introduction of timestamps ensures the time alignment accuracy of multi-cycle data when calculating the delay distribution, avoiding offset errors caused by data asynchrony.

[0188] S1.2: Based on the historical control state sequence, calculate the response delay distribution characteristics of water temperature in the heating network to frequency control actions, and calculate the response offset function by fitting the residual curve;

[0189] Specifically, the water temperature response in heating pipe networks exhibits a significant delay characteristic. This means that after the circulation pump frequency is adjusted, the water temperature change does not occur instantly but rather after a certain period. This response delay depends not only on the water flow transmission time but also on factors such as the pipe's heat capacity and the heat dissipation rate at the user end. Therefore, the delay distribution may vary at different nodes and during different operating stages. Without quantitative analysis of the delay distribution, the matching degree between control commands and water temperature response will decrease, resulting in delayed or overshooting adjustments.

[0190] In this embodiment, the delay distribution is calculated by performing cross-correlation analysis on the historical frequency increment sequence and the corresponding water temperature change rate sequence, thereby determining the degree of correlation between the two at different delay times. To eliminate temperature disturbances caused by measurement noise and non-frequency control factors, a residual curve fitting method is used to compare the observed actual water temperature change with the ideal change based on the zero-delay assumption. The distribution of the residual over time reflects the shape and amplitude of the delay, and the corresponding response offset function is calculated based on the distribution of the residual over time.

[0191] S1.3: Fit the current water temperature with the response offset function to obtain a relative offset trajectory curve with the target constant temperature value as the reference point, and output the dynamic trajectory, wherein the relative offset trajectory curve has the sampling marker point as the horizontal axis and the output value of the response offset function as the vertical axis.

[0192] Specifically, the core of dynamic trajectory lies in depicting the motion of the current water temperature in the temperature inertial field, while the response offset function ensures that the trajectory not only reflects changes in temperature but also takes into account relative position corrections caused by delays. By using the target isothermal value as a reference point, the vertical axis of the trajectory can be converted into a relative offset.

[0193] In this embodiment, the fitting process first matches the water temperature sampling sequence within the current control cycle point by point with the sampling identifier and the response offset function, so that each temperature sampling value corresponds to a delay correction amount. Then, the corrected temperature values ​​are converted into relative offsets (i.e., the difference between the current value and the target constant temperature value), and a two-dimensional curve is formed with the sampling identifier axis. The direction of the two-dimensional curve in the sampling sequence, and the fluctuation amplitude and trend on the vertical axis, constitute all the information of the dynamic trajectory. The shape of the trajectory directly reflects whether the current water temperature tends to stabilize in the sampling dimension, or whether there is continuous deviation or oscillation.

[0194] refer to Figure 6 , Figure 6 This is a schematic diagram of the dynamic trajectory of an embodiment of this application.

[0195] Figure 6 The horizontal axis represents sampling points N1 to N5, indicating the sampling positions set sequentially along the water flow path of the heating network within the current control cycle. The vertical axis represents the relative offset after correction based on the response offset function, with the target constant temperature value taken as the zero reference point.

[0196] Figure 6 The diagram further illustrates that hollow markers represent the current temperature of each sampling point, while solid markers represent the corrected offset value, which is based on the current temperature and incorporates the response delay characteristics of water temperature to frequency control commands over multiple historical control cycles, after correction by the response offset function.

[0197] Understandably, by comparing the differences between hollow and solid markers along the vertical axis, the influence of temperature inertia on water temperature distribution can be visually reflected: when the corrected offset value converges towards zero compared to the current offset value, it indicates that the inertial effect helps the temperature stabilize and return to the target constant temperature; conversely, if the corrected offset value is far from zero, it suggests that the current inertial trend may push the temperature further away from the target constant temperature. The overall trend of the curve can reveal the spatial trend of water temperature change and the result of inertial correction on the trajectory shape.

[0198] It should be noted that the temperature correction here is not for the accuracy of the current water temperature sample value itself. When the water temperature sensor collects the current temperature, its measurement value already has an engineering-acceptable accuracy and does not require error correction.

[0199] The correction mentioned in this embodiment refers to the introduction of a response offset based on the deviation relationship between frequency control actions and water temperature changes in historical control cycles when constructing the dynamic trajectory. This offset is then superimposed on the current water temperature sampling sequence as an inertial correction value to reflect the true response position of the heating network in the temperature inertial field.

[0200] Specifically, in historical control cycles, the impact of changes in the circulation pump frequency on water temperature has a delayed effect; that is, after the control action is executed, the water temperature does not immediately reach the new equilibrium value, but rather undergoes a transition period. In the current control cycle, the actual temperature value of each sampled point is superimposed with the predicted offset, and the target constant temperature value is used as the reference zero point. The result is then converted into a relative offset, thus obtaining... Figure 6 The relative offset trajectory curve shown not only reflects the relative position of each sampling point to the target constant temperature value, but also reflects the trend direction and inertial strength of temperature change. For example, at point N5, the current temperature is close to the target value, but the curve still maintains an upward trend, indicating that the position is still in the stage of heating inertia.

[0201] Next, we will further elaborate on the part of the method of this application regarding the attractive potential energy state.

[0202] It is understandable that the attraction potential energy state is used in this application to characterize the stability trend and regression ability of the current water temperature under the action of the target constant temperature value. Its essence is to organically combine the dynamic trajectory information of water temperature with the spatial distribution information of the disturbance residual matrix in order to determine whether the water temperature change tends to be attracted by the target constant temperature value and gradually stabilize, or whether it shows a trend of deviation or even divergence under the influence of external disturbances.

[0203] In one example, based on the dynamic trajectory and the disturbance residual matrix, the attractive potential energy state between the current water temperature and the target isothermal value is calculated, including:

[0204] S2.1: Based on the pump frequency input of the previous control cycle and the current water temperature of the current control cycle, update the disturbance residual matrix output of the previous control cycle to obtain the updated disturbance residual matrix.

[0205] The update of the perturbation residual matrix has been described in detail above, and will not be repeated here.

[0206] S2.2: Based on the dynamic trajectory, determine the deviation range, rate of change, and trend direction of the current water temperature relative to the target constant temperature value, and calculate the corresponding trajectory feature vector;

[0207] Specifically, the purpose of this step is to extract the two-dimensional information of the dynamic trajectory, which describes the state of water temperature change, into a feature vector that can be numerically calculated with the disturbance residual matrix. The dynamic trajectory itself can only reflect the relative offset and trend change of the sampling points, but cannot directly participate in the spatial distribution analysis of the disturbance. Therefore, it is necessary to quantify the offset amplitude, rate of change, and trend direction of each sampling position in the trajectory to form a trajectory feature vector.

[0208] In this embodiment, the construction of the trajectory feature vector includes the following three dimensions:

[0209] In the first dimension, the correction offset of each sampling marker point within the current control cycle is normalized to make the values ​​between different locations comparable.

[0210] In the second dimension, the rate of change of offset between adjacent sampling points is calculated to obtain the local rate of change, and the direction of the change trend (approaching the target constant temperature value or moving away from the target constant temperature value) is determined by the sign.

[0211] In the third dimension, the offset magnitude, rate of change, and trend direction are encoded as triplet vectors, which are then arranged in the order of sampling points to obtain the trajectory feature vector. This constructed vector reflects both the overall trend and preserves details of local changes, making its matching analysis with the perturbation residual matrix more targeted.

[0212] S2.3: Couple the trajectory feature vector and the updated perturbation residual matrix to calculate the attractive potential energy state;

[0213] Specifically, the core of this step lies in achieving coupled calculation of the trajectory feature vector and the updated perturbation residual matrix. This is because simple trajectory changes may be caused by external perturbations or by internal inertial adjustments, while the perturbation residual matrix records the spatial accumulation and propagation characteristics of historical perturbations. Through coupled calculation, it can be determined whether the trajectory change is consistent with the distribution trend of the residual perturbation, thus distinguishing between perturbation-driven deviations and spontaneous stabilization processes.

[0214] Understandably, the attractive potential energy state can comprehensively reflect the water temperature's ability to return to the target constant temperature value and the degree of influence of external disturbances. It has a dual criterion: on the one hand, if the trajectory change direction is consistent with the propagation direction of high-intensity disturbances, the potential energy state decreases, indicating that the control strategy needs to strengthen the driving force; on the other hand, if the trajectory change direction is opposite to the propagation direction of disturbances, the potential energy state increases, and the control strategy can appropriately narrow the frequency bandwidth to reduce unnecessary adjustments.

[0215] It should be noted that the attractive potential energy state here is the aggregated overall potential energy state. In some alternative implementations, the circulating pumps can be distributed to achieve precise zone control.

[0216] In one example, coupling the trajectory feature vector and the updated perturbation residual matrix includes:

[0217] S2.3.1: Based on the trajectory feature vector, determine the disturbance-sensitive direction factor representing the current water temperature change trend;

[0218] In this embodiment, the direction factor is extracted by statistically analyzing the rate of change and trend direction components in the trajectory feature vector to form a local direction component for each sampling location. These local components are then weighted and averaged, with the weights determined based on the hydraulic importance or thermal weight allocation of each location within the pipe network structure. For example, upstream nodes near the circulating pump can be assigned a higher weight due to their greater impact on the overall water temperature.

[0219] S2.3.2: Extract the perturbation submatrix corresponding to the perturbation sensitive direction factor from the updated perturbation residual matrix;

[0220] Specifically, since the residual perturbation matrix fully records the residual perturbation transmission relationship between each node, while the directional factor only involves nodes on the current trend path, there is no need to perform calculations on the entire matrix. This can reduce the amount of computation and allow us to focus on the perturbation region that is directly related to the current trend.

[0221] In this embodiment, the method for extracting the perturbation submatrix is ​​as follows: based on the sampling node numbers involved in the direction factor, the corresponding rows and columns are selected from the perturbation residual matrix to form a smaller matrix. This submatrix specifically describes the perturbation propagation between these nodes and between them and their surrounding nodes.

[0222] S2.3.3: Perform directional matching analysis between the disturbance-sensitive direction factor and the disturbance submatrix to determine the influence relationship of residual disturbance on the current trajectory trend;

[0223] Understandably, the core logic of matching analysis lies in:

[0224] If the main direction of the disturbance propagation is the same as the current trend of water temperature change, it can be inferred that the disturbance is promoting the temperature to change in that direction; otherwise, it can be inferred that the disturbance is hindering the development of the current trend.

[0225] In this embodiment, the matching analysis is achieved by calculating the angle between the direction factor and the perturbation propagation vector in the submatrix. The smaller the angle, the more consistent the two are. An angle close to 180 degrees indicates that the trends are opposite.

[0226] S2.3.4: Based on the results of the directional matching analysis, calculate the attraction stability index of the current water temperature on the control path to obtain the attraction potential energy state;

[0227] In one alternative implementation, the numerical range of the stability index can be set based on engineering experience. For example, a value closer to 1 indicates that it is more conducive to the temperature returning to the target value, while a value closer to 0 indicates that it is more susceptible to disturbances and difficult to stabilize near the target value.

[0228] In this embodiment, the stability index calculation integrates directional consistency weight, disturbance intensity weight, and node position weight within the pipe network's thermal structure, forming a comprehensive evaluation value coupled with multiple factors. Finally, this evaluation value is mapped to an attractive potential energy state for direct use in regulation and control.

[0229] The width and center position of the frequency bandwidth are determined based on the attracted potential energy state and the perturbation residual matrix.

[0230] In one example, based on the attractive potential energy state, the controller outputs a frequency bandwidth range, including:

[0231] S3.1: Using the frequency center value of the previous control cycle as a reference, determine the frequency center offset of the current control cycle based on the state of the attracting potential energy, wherein the frequency center offset is inversely proportional to the state of the attracting potential energy.

[0232] S3.2: Determine the adjustment coefficient of the frequency bandwidth range based on the disturbance intensity on the current control path in the disturbance residual matrix, wherein the adjustment coefficient is used to control the width convergence rate of the frequency bandwidth;

[0233] S3.3: Based on the frequency center offset and the adjustment coefficient, and combined with the frequency bandwidth range of the previous cycle, output the corresponding frequency bandwidth range.

[0234] In one example, the method further includes:

[0235] The sampling period of the water temperature sensor is determined based on the operating period of the heating network.

[0236] The operating phases include an initial heating phase, a stable heating phase, and a nighttime load phase, with different target constant temperatures for each phase.

[0237] In one example, determining the sampling period of the water temperature sensor based on the operating period of the heating network includes:

[0238] If the initial heating phase is underway, a first sampling period is set.

[0239] If the system is in a stable heating phase, a second sampling period is set, wherein the first sampling period is shorter than the second sampling period.

[0240] If the nighttime load phase is in progress, a third sampling period is set, wherein the second sampling period is shorter than the third sampling period.

[0241] The first sampling period, the second sampling period, and the third sampling period are adjusted based on the perturbation convergence rate in the perturbation residual matrix, the dynamic trajectory of the current water temperature in the temperature inertial field, and the offset of the historical adjustment of the circulation pump frequency.

[0242] Understandably, the disturbance convergence rate is calculated based on the load changes in the disturbance residual matrix, and the offset is obtained by summing and averaging the historical circulating pump frequency adjustments.

[0243] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A control method for constant temperature and variable current in heating and ventilation systems, applied to heating pipe networks, characterized in that, The heating network includes a controller, a circulating pump, a water temperature sensor, and the network itself. The controller is equipped with a temperature inertial field and a disturbance residual matrix. The temperature inertial field is simulated using virtual potential energy based on a preset target constant temperature value. The disturbance residual matrix is ​​constructed based on the disturbance residual between water temperature changes and ideal changes within historical control cycles. The control method includes: The current water temperature is collected by the water temperature sensor, and the dynamic trajectory of the current water temperature in the temperature inertial field is constructed based on the pump frequency input and water temperature output changes of the circulating pump in multiple historical control cycles. The current water temperature is the water temperature sampling sequence collected in the current control cycle. Based on the dynamic trajectory and the disturbance residual matrix, the attractive potential energy state between the current water temperature and the target isothermal value is calculated. Based on the state of the attraction potential energy, the controller outputs a frequency bandwidth range to control the circulating pump to perform frequency adjustment within the frequency bandwidth range, thereby obtaining the pump frequency input of the circulating pump in the current control cycle. The construction of the dynamic trajectory of the current water temperature in the temperature inertial field includes: Collect the circulation pump frequency control commands and corresponding water temperature response data within at least M control cycles before the acquisition, and construct a historical control state sequence arranged in chronological order. Each state unit in the historical control state sequence includes a timestamp, frequency increment value, water temperature change rate, and sampling interval duration. Based on the historical control state sequence, the response delay distribution characteristics of water temperature in the heating network to frequency control actions are calculated, and the response offset function is calculated by fitting the residual curve. The current water temperature is fitted with the response offset function to obtain a relative offset trajectory curve with the target constant temperature value as the reference point, and a dynamic trajectory is output. The relative offset trajectory curve has the sampling marker point as the horizontal axis and the output value of the response offset function as the vertical axis. The disturbance residual matrix is ​​constructed based on historical measurement and prediction data from the previous k control cycles. The prediction data is a water temperature change sequence calculated by the controller based on the pump frequency input of the circulating pump and an ideal response model. The specific construction method is as follows: After the kth control cycle ends, obtain the actual water temperature sampling sequence for the previous k control cycles. and the ideal water temperature sampling sequence predicted by the controller Where j represents a single control period, and the specific form of the sampling sequence is as follows: ， Where p represents the number of sampling points; Based on the actual water temperature sampling sequence corresponding to each control cycle j and the ideal water temperature sampling sequence predicted by the controller, the residual vector is calculated by the difference. Calculate based on the hydraulic connectivity of the heating pipe network. Given an adjacency matrix A of order A, calculate the corresponding Laplace matrix based on the adjacency matrix. , where D represents the degree matrix of the adjacency matrix; Based on the preset diffusion coefficient Calculate the topological diffusion kernel , where e represents matrix exponentiation operation, and the residual vector is spatially convolved through the topological diffusion kernel to generate a residual field matrix; The forgetting factor is calculated based on the pump frequency input corresponding to the control cycle. The residual field matrix is ​​weighted and assigned values ​​through the forgetting factor. The perturbation residual matrix is ​​obtained by summing the weighted residual field matrix. Based on the dynamic trajectory and the disturbance residual matrix, the attractive potential energy state between the current water temperature and the target isothermal value is calculated, including: Based on the pump frequency input of the previous control cycle and the current water temperature of the current control cycle, the disturbance residual matrix output of the previous control cycle is updated to obtain the updated disturbance residual matrix. Based on the dynamic trajectory, determine the deviation magnitude, rate of change, and trend direction of the current water temperature relative to the target constant temperature value, and calculate the corresponding trajectory feature vector; The trajectory feature vector and the updated perturbation residual matrix are coupled to calculate the attractive potential energy state; Couple the trajectory feature vector and the updated perturbation residual matrix, including: Based on the trajectory feature vector, a disturbance-sensitive direction factor representing the current water temperature change trend is determined; Extract the perturbation submatrix corresponding to the perturbation sensitive direction factor from the updated perturbation residual matrix; The perturbation-sensitive direction factor and the perturbation submatrix are subjected to direction matching analysis to determine the influence relationship of residual perturbation on the current trajectory trend; Based on the results of the directional matching analysis, the attraction stability index of the current water temperature on the control path is calculated to obtain the attraction potential energy state.

2. The control method for constant temperature and current conversion in HVAC thermal systems according to claim 1, characterized in that, The update of the disturbance residual matrix is ​​based on the pump frequency input of the previous control cycle and the current water temperature of the current control cycle. The specific update method is as follows: Obtain the disturbance residual matrix of the output from the previous control cycle. ; The forgetting factor of the disturbance residual matrix is ​​updated based on the pump frequency input of the previous control cycle to obtain the weighted value. ; Calculate the residual field matrix for the current control cycle based on the current water temperature. The updated perturbation residual matrix is ​​calculated based on the perturbation residual matrix and weighting value of the previous control cycle and the residual field matrix of the current control cycle. The specific calculation method is as follows: 。 3. The control method for constant temperature and current conversion in HVAC thermal systems according to claim 1, characterized in that, The width and center position of the frequency bandwidth are determined based on the attracted potential energy state and the disturbance residual matrix.

4. The control method for constant temperature and current conversion in HVAC thermal systems according to claim 3, characterized in that, Based on the attractive potential energy state, the controller outputs a frequency bandwidth range, including: Taking the frequency center value of the previous control cycle as a reference, the frequency center offset of the current control cycle is determined according to the state of the attracting potential energy, wherein the frequency center offset is inversely proportional to the state of the attracting potential energy. Based on the disturbance intensity on the current control path in the disturbance residual matrix, the adjustment coefficient of the frequency bandwidth range is determined, wherein the adjustment coefficient is used to control the width convergence rate of the frequency bandwidth. Based on the frequency center offset and the adjustment coefficient, combined with the frequency bandwidth range of the previous cycle, the corresponding frequency bandwidth range is output.

5. The control method for constant temperature and current conversion in HVAC thermal systems according to claim 1, characterized in that, The method further includes: The sampling period of the water temperature sensor is determined based on the operating period of the heating network. The operating phases include an initial heating phase, a stable heating phase, and a nighttime load phase, with different target constant temperatures for each phase.

6. The control method for constant temperature and current conversion in HVAC thermal systems according to claim 5, characterized in that, The sampling period of the water temperature sensor is determined based on the operating period of the heating network, including: If the initial heating phase is underway, a first sampling period is set. If the system is in a stable heating phase, a second sampling period is set, wherein the first sampling period is shorter than the second sampling period. If the nighttime load phase is in progress, a third sampling period is set, wherein the second sampling period is shorter than the third sampling period. The first sampling period, the second sampling period, and the third sampling period are adjusted based on the perturbation convergence rate in the perturbation residual matrix, the dynamic trajectory of the current water temperature in the temperature inertial field, and the offset of the historical adjustment of the circulation pump frequency.

Images