An intelligent energy-saving control method for temperature regulation equipment

By establishing a unified model in temperature control equipment and coordinating the allocation of flow and pressure difference, the problems of rising energy consumption and frequent equipment start-ups and shutdowns in complex coupled pipe networks were solved, achieving energy saving and stable operation at the system level.

CN123113379APending Publication Date: 2026-11-17THE SECOND CONSTR OF CHINA CONSTR EIGHTH ENG DIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611486701.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-23
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing temperature control equipment in complex coupled pipe networks suffers from problems such as increased overall energy consumption, frequent equipment start-ups and shutdowns, and temperature quality fluctuations because it fails to take into account hydraulic factors, source-side equipment efficiency characteristics, engineering boundaries, and total system power.

Method used

A unified model is adopted that combines the relationship between thermal, hydraulic and equipment efficiency to generate an objective function with the minimum total power of the system as the main objective. Through rolling optimization and tracking control, flow rate and pressure difference are allocated in a coordinated manner, safe operating boundaries are set, and adjustments are made in real time to achieve energy saving of the system.

Benefits of technology

In complex coupled pipe networks, overall energy saving at the system level is achieved, avoiding energy consumption increases, equipment oscillations and temperature fluctuations, maintaining the accuracy and reliability of optimization solutions, and taking into account both energy saving effect and temperature quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113379A_ABST
    Figure CN123113379A_ABST
Patent Text Reader

Abstract

This invention relates to the field of temperature control equipment technology, specifically an intelligent energy-saving control method for temperature control equipment. The method includes: acquiring the topology and parameters of the temperature control pipeline network; generating a unified model encompassing thermal, hydraulic, and equipment efficiency relationships; generating an objective function with the primary objective of minimizing total system power and balancing temperature deviation and actuator changes; setting temperature range, flow rate and pressure difference range, and safe operating boundaries for actuators and equipment; and implementing a rolling optimization controller based on the unified model and external disturbance prediction. The beneficial effects are: in complex coupled pipeline networks, the method coordinates the allocation of flow rate and pressure difference, ensuring that source-side equipment such as pumps, fans, and heat pumps operate in high-efficiency zones for extended periods. This helps avoid problems in existing solutions, such as increased total system energy consumption, multi-branch flow competition and oscillation, frequent start-stop cycles, and temperature quality fluctuations caused by unilaterally reducing heat loss, thus achieving overall energy saving at the system level.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of temperature control equipment technology, specifically to an intelligent energy-saving control method for temperature control equipment. Background Technology

[0002] Temperature control equipment typically delivers cooling or heating energy to energy-consuming objects via pipelines, and losses during the delivery process directly affect energy supply efficiency. To reduce delivery losses, existing technologies mostly rely on temperature monitoring to build pipeline heat loss prediction models, set preset heat loss indicators to screen or plan control pipelines, and then implement segmented control of the pipelines.

[0003] However, such solutions have a relatively singular objective function, using only achieving or minimizing heat loss as the criterion. They fail to incorporate hydraulic factors such as flow rate, pressure difference, and local resistance, as well as the efficiency characteristics of source-side equipment such as pumps, fans, compressors, or heat pumps (including the relationship between performance coefficient and load, minimum stable frequency, and shortest start-up and shutdown time), and the total system power into a unified optimization. In multi-branch coupled pipe networks, reducing heat loss often requires increasing flow velocity or tightening the supply and return temperature difference, leading to an increase in pumping power. After flow redistribution and pressure difference reconfiguration, source-side equipment is pushed out of the high-efficiency operating zone, ultimately resulting in a decrease in heat loss data but a higher actual power consumption for the entire system. Furthermore, process boundaries and peak power constraints are often not simultaneously incorporated into the optimization, easily leading to frequent equipment start-ups and shutdowns, and fluctuations in temperature quality on-site.

[0004] Therefore, there is a need for an intelligent energy-saving control method that can take into account hydraulic factors, source-side equipment efficiency characteristics, engineering boundaries and total system power, to achieve overall energy saving at the system level in complex coupled pipe networks. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent energy-saving control method for temperature regulation equipment to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent energy-saving control method for temperature regulation equipment. This method first obtains the topology and operating parameters of the temperature regulation pipeline network, and then establishes a unified model that simultaneously covers three types of relationships: thermal, hydraulic, and equipment efficiency.

[0007] Among them, the thermal relationship includes the coupling relationship between the pipeline network diagram structure and the supply and return temperature difference and flow rate of the terminal heat exchanger. The pipeline network diagram structure records the nodes, branches, and the length, diameter, local resistance and insulation layer parameters of the branches; the hydraulic relationship includes the correspondence between the flow rate and pressure difference of the branches and the continuity constraints of the nodes; the equipment efficiency relationship includes the correspondence between the efficiency, speed and flow rate of the water pump and the fan, as well as the correspondence between the performance coefficient of the compressor or heat pump and the load and supply and return temperature.

[0008] Based on this, an objective function is generated, with the primary objective of minimizing the total system power, while also balancing temperature difference deviation and changes in the actuated quantity. The total system power consists of pumping power, fan power, and source-side power. Subsequently, the temperature range, flow rate and differential pressure range, as well as the safe operating boundaries of the actuators and equipment are set.

[0009] Under the premise that the unified model is effective, the rolling optimization controller is run based on external disturbance prediction to solve the reference setpoint and send it down to the execution layer. The reference setpoint covers pump speed, fan speed, valve opening, source side frequency and supply and return temperature difference. Finally, the tracking controller runs in the second sampling period to output control signals to the water pump, fan, valve and compressor or heat pump to drive the measured state to approach the reference setpoint.

[0010] This method collects temperature, flow rate, pressure difference and power data in the first sampling period, and determines whether the model residual of the unified model exceeds the model residual threshold. Once it exceeds the threshold, the unified model parameters are updated, and the prediction, comparison and verification are repeated until the residual returns to within the threshold. The first sampling period is 1 to 10 seconds, and the model residual threshold is calibrated during the debugging phase.

[0011] Optionally, for multi-branch coupled networks, rolling optimization adopts distributed coordinated solution, that is, the entire network is decomposed into multiple branch local subproblems that can be solved in parallel, and consistency constraints are established for shared variables such as total flow and trunk pressure difference; when the change of the reference set obtained in two adjacent iterations is lower than the consistency threshold, a coordinated reference set covering the entire network is output.

[0012] Among them, the temperature difference deviation is calculated as the weighted sum of the squares of the deviations between the set temperature and the measured temperature in each control area, and the execution quantity change is calculated as the weighted sum of the changes in valve opening and rotation speed at adjacent times. The power weight, temperature difference deviation weight and execution change weight are all set as configurable parameters.

[0013] Furthermore, the temperature range includes the upper and lower limits of the temperature of each area and the supply and return temperature difference, as well as the maximum rate of rise and fall; the flow rate and differential pressure range includes the upper and lower limits of flow rate, differential pressure, and valve opening, and the minimum stable valve opening is not less than 5%; the safe operating boundary includes the minimum stable frequency of water pumps and fans, the minimum start-up time and the minimum shut-off time of compressors, and the maximum slope limit for start-up and shutdown; at the same time, node flow continuity and differential pressure closure constraints are applied to the pipeline network.

[0014] As a preferred approach, the upper-level rolling optimization and the lower-level tracking control operate in tandem on two time scales: the upper layer solves for the reference setpoint set and distributes it over a prediction time domain of no less than 15 minutes with an optimization cycle of 30 to 120 seconds; the lower layer performs feedforward and proportional-integral-derivative composite control with a second sampling cycle of 1 to 10 seconds to track the reference trajectory and suppress short-term disturbances.

[0015] In addition, the method also includes coordinated tuning on the source side and the pipeline side: for a multi-heat-source or cold-source parallel system, the real-time performance coefficient and start-stop constraints of each heat source or cold source are calculated, the selection results and parallel configuration of heat sources or cold sources are generated and sent for execution; for a single heat source or cold source system, the operating point of the compressor or heat pump is kept in the high-efficiency operating range defined by the equipment efficiency relationship by adjusting the supply and return temperature difference and the flow of each branch.

[0016] As an alternative, time-of-use electricity price or carbon intensity signal is introduced as a time weight into the power part of the objective function to minimize the total power of the system in different time periods, and the upper limit constraint of the total power of the system is superimposed during the demand-limited period. The changes in time weight and power upper limit will affect the solution of the next upper-level rolling optimization, so that the reference setpoint is adaptively adjusted with the electricity price or carbon intensity and demand boundary.

[0017] Furthermore, the method also sets up a closed-loop adjustment for operational deviations: calculate the total power deviation and temperature deviation of the system; when any deviation exceeds the corresponding trigger threshold, adjust the power weight, temperature deviation weight and execution change weight by a preset step size, and tighten or relax the execution change rate limit; when the limit is exceeded multiple times in a row or the deviation shows a systematic trend, trigger the priority re-identification of unified model parameters; all of the above adaptive adjustments are subject to amplitude and minimum hold time constraints.

[0018] When a critical sensor failure, communication anomaly, severe model mismatch, or an urgent demand event is detected, the method switches to a degraded control mode: freezing the most recently stable reference setpoint, limiting the total system power to no more than the safe power limit, limiting the rate of change of the executed quantity to no more than the safe rate of change, and maintaining the area temperature within the safe temperature range. During degraded control, redundant sensor arbitration is enabled to suppress single-point drift using median or voting methods, conservative control commands are output, and event logs are recorded. After the sensor and communication self-tests pass, the method exits the degraded mode in the following order: tracking, unfreezing, relaxing restrictions, and resuming rolling optimization.

[0019] Compared with the prior art, the beneficial effects of the present invention are: The intelligent energy-saving control method for temperature regulation equipment proposed in this invention adopts a unified model that includes thermal, hydraulic and equipment efficiency relationships. With the minimum total system power as the main objective, it explicitly incorporates temperature, flow rate, pressure difference and equipment safe operation boundaries under the rolling optimization and tracking control architecture. This application can coordinate the allocation of flow rate and pressure difference in complex coupled pipe networks, so that source-side equipment such as pumps, fans, and heat pumps can operate in the high-efficiency zone for a long time. This helps to avoid the problems of increased total system energy consumption, multiple branch flow competition oscillation, frequent start-stop and temperature quality fluctuation caused by unilaterally reducing heat loss in existing solutions, thus achieving overall energy saving at the system level.

[0020] By updating unified model parameters through online identification and model residual feedback, performance degradation caused by model aging can be suppressed, which is beneficial for maintaining the accuracy of optimization solutions in the long term. Distributed coordinated solution and consistency constraints enable collaborative allocation between branches and trunk lines in multi-branch coupled networks, helping to avoid oscillations caused by flow preemption and pressure differential reconstruction. Time-of-use pricing or carbon intensity time-weighted pricing and demand ceiling constraints allow for time-based adjustments to energy-saving priorities, which is beneficial for peak shaving and valley shifting, reducing peak electricity costs and carbon intensity. Closed-loop adaptive weight adjustment and execution rate limits enable automatic adjustment of optimization priorities when operational deviations exceed limits, which helps to balance energy-saving effects and temperature quality. Degradation control modes enable safe operation and smooth recovery in the event of sensor or communication anomalies, which helps improve the reliability and maintainability of system operation. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the present invention clear and complete, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only some, not all, embodiments of the present invention, and are merely illustrative of the embodiments of the present invention. They are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figure 1 The present invention provides a technical solution: an intelligent energy-saving control method for temperature regulation equipment, comprising the following steps.

[0024] S1. During the commissioning phase, engineers input the on-site pipeline network diagram into the control system, including the physical location, functional attributes, and connection relationships of each node with adjacent branches. For each branch, they fill in the length, diameter, local resistance, and insulation layer parameters, and simultaneously register the interface diameter and rated heat exchange capacity of the terminal heat exchangers. The system performs equipment inventory and address binding for flow meters, differential pressure transmitters, temperature sensors, and power metering points, establishing a one-to-one correspondence between measurements and topology and completing time synchronization. Based on the above information, a unified model is automatically generated: the hydraulic part uses the flow rate and differential pressure correspondence of each branch and node continuity constraints to form network equations; the thermal part uses the supply and return temperature difference and flow rate coupling relationship of the terminal heat exchangers to reflect the heat exchange capacity and the required supply and return temperature difference; the equipment efficiency relationship uses the efficiency, speed, and flow rate correspondence of water pumps and fans, as well as the performance coefficient, load, and supply and return temperature correspondence of compressors or heat pumps to form a performance map.

[0025] S2. After entering the operational phase, temperature, flow rate, differential pressure, and power data are synchronously collected on the real-time bus during the first sampling cycle. First, timestamp alignment, outlier limiting, and missing measurement compensation (last-most recent hold or linear extrapolation) are performed. Then, a unified model is used to predict key quantities at the same time and calculate model residuals, which are compared with a model residual threshold. If the residual does not exceed the threshold, the next cycle begins; if the residual exceeds the threshold, a parameter update process is triggered. Priority is given to fine-tuning local resistance and equivalent heat conduction parameters. Then, based on the deviation in efficiency relationships, small-step corrections are made to the key coefficients of water pumps, fans, compressors, or heat pumps. This prediction, comparison, and verification process is repeated until the residual falls back below the threshold or reaches the maximum number of updates allowed per cycle. The system sets a minimum effective step size and a minimum hold time for parameter updates to ensure timely adaptation to field changes within the first sampling cycle of 1 to 10 seconds without compromising the numerical stability of rolling optimization. Online identification results are written back to the unified model in real time for subsequent objective function generation and rolling optimization calls.

[0026] S3. Under the premise that the unified model is effective, generate an objective function for the current control time domain: Based on the instantaneous power estimation of the unified model, merge pumping power, fan power, and source-side power to obtain the total system power; read the set temperature and measured temperature of each control area to obtain a temperature difference deviation measure; use the difference between the historical reference setpoint and the current reference setpoint to represent the change in execution quantity; load three types of configurable parameters—power weight, temperature difference deviation weight, and execution change weight—to form a weighted objective, and establish a linkage with the trigger threshold in step S7 so that the emphasis is automatically adjusted when the deviation exceeds the limit; if time-of-use electricity price or carbon intensity signal is connected, introduce time weight in the power component to reflect time-period differences. All weights are parameterized and stored in the controller and can be modified online without interrupting execution.

[0027] S4. While generating the objective function, load the executable boundary of the project as optimization constraints: Set the temperature range and supply-return temperature difference range for each area according to process or comfort requirements, and limit the maximum rise and fall rate to avoid rapid temperature jumps; set the flow rate and pressure difference range for hydraulic channels, and define the minimum stable valve opening as not less than 5% to prevent control jitter and noise amplification caused by small openings; set the minimum stable frequency of water pumps and fans, the minimum start and stop time of compressors, and the maximum slope for start-up and shutdown for the equipment layer to ensure that the equipment does not enter the frequent start-up and shutdown or unstable operating range; apply node flow continuity and branch pressure difference closure constraints to the network layer to ensure the physical feasibility of the solution results; if it is during peak power consumption periods, the upper limit of the total system power is also added to the solution problem as a strong constraint. All constraints are uniformly converted into a combination of upper and lower limits and logical conditions before entering the solver, and unit consistency and physical feasibility verification are completed.

[0028] S5. Based on a unified model and external disturbance prediction, the rolling optimization controller solves the future prediction time domain in each upper-level optimization cycle to minimize the total system power and satisfy the constraints of step S4. It then obtains the reference setpoints for the current cycle and immediately issues them down. These reference setpoints include pump speed, fan speed, valve opening, source-side frequency, and supply-return temperature difference. When the pipeline network is a multi-branch coupled network, distributed coordinated solving is automatically enabled. The entire network is divided into local subproblems of each branch that can be solved in parallel. Consistency constraints are applied to shared variables such as total flow and mains pressure difference. The process stops and outputs the coordinated results when the change in the reference setpoint set falls below the consistency threshold or reaches the maximum number of iterations. Simultaneously, validity periods and change rate requirements are added to each reference setpoint for smooth tracking by the execution layer.

[0029] S6. The tracking controller reads the reference setpoint in the second sampling period and generates a control signal by combining it with real-time measurements. It uses feedforward and proportional-integral-derivative composite control to achieve smooth adjustment of pumps, fans, valves, and compressors or heat pumps. The second sampling period is the same as or shorter than the first sampling period. Throughout the process, it applies execution quantity change rate and soft start-stop limits to valve opening, speed and source-side frequency, and activates anti-saturation logic when the boundary is reached to ensure that the measured state closely follows the reference setpoint. At the same time, it reports the executed instructions and tracking error for deviation evaluation in step S7.

[0030] S7. The system calculates the total power deviation and temperature deviation according to the upper-level optimization cycle and the second sampling cycle, respectively, and compares them with the corresponding trigger thresholds: if no limit is exceeded, the existing weights and restrictions are maintained; if any limit is exceeded, the power weight, temperature deviation weight, and execution change weight are adjusted according to the preset step size, and the execution change rate limit is tightened or relaxed accordingly; if the limit is exceeded multiple times consecutively or the deviation shows a systematic trend, the priority re-identification in step S2 is triggered. All adaptive adjustments are constrained by amplitude and minimum hold time to avoid oscillation.

[0031] S8. When a critical sensor failure, communication anomaly, severe model mismatch, or an urgent demand event is detected, the system immediately enters degraded control: freezing the most recent stable reference setpoint, limiting the total system power to no more than the safe power limit, limiting the rate of change of the executed quantity to no more than the safe rate of change, and maintaining the area temperature within the safe temperature range; simultaneously, enabling redundant sensor arbitration to suppress single-point drift using median or voting methods, outputting conservative control commands, and recording the event log. After the sensor and communication self-tests pass and the measurement is effectively restored, the system exits degraded control in the order of tracking, unfreezing, relaxing restrictions, and resuming rolling optimization, returning to normal collaborative control.

[0032] As an example, two types of tasks—upper-level rolling optimization and lower-level tracking control—are run simultaneously on the edge controller. The upper layer is triggered periodically with an optimization cycle of 30 to 120 seconds. It reads the unified model, engineering constraints, external disturbance predictions, and the most recent running data. For a prediction time domain of no less than 15 minutes, it generates a continuous sequence of reference setpoints and sends the first reference setpoint of the current time period to the execution layer. The reference setpoints include at least pump speed, fan speed, valve opening, source-side frequency, and supply-return temperature difference, and carry validity period and change rate requirements. The lower layer works continuously during a second sampling cycle of 1 to 10 seconds: the feedforward calculates the expected execution quantity based on the static mapping of the unified model, the proportional-integral-derivative loop corrects the instantaneous error, and applies change rate and soft-start / stop limits to the valve opening and speed. When the boundary is reached, anti-saturation is activated to prevent overshoot. If the upper layer does not provide a new setpoint within the specified time, the lower layer maintains the most recent stable reference and executes slowly according to the change rate limit until a new reference is received or a degraded mode is entered. At the end of each second sampling period, the lower layer sends back the executed instructions and tracking errors, which the upper layer uses together with power and temperature difference statistics in the next round of optimization.

[0033] In another embodiment, during each upper-level optimization cycle, the real-time performance coefficients, availability status, minimum start-up and minimum shutdown times, start-up / shutdown ramp limits, and maintenance priorities of each heat or cold source are simultaneously evaluated. These, along with the temperature range, flow rate, and differential pressure range on the pipeline side, serve as decision constraints to generate source-side selection results and parallel configuration ratios. Specifically, several candidate combinations are first given. For each candidate combination, feasible supply-return temperature difference and flow rate allocation are calculated in the prediction time domain. Combinations that do not meet engineering boundaries or demand limits are eliminated. Then, the scheme with the lowest total system power and the most stable temperature quality is selected from the remaining combinations. This forms an integrated reference for equipment start-up / shutdown selection, parallel configuration ratio, supply-return temperature difference targets, and branch flow rate targets, which is then sent to the execution layer. The source-side controller smoothly switches according to the minimum start-up and minimum shutdown times and ramp limits. For a single heat or cold source system, the operating point of the compressor or heat pump is kept within the high-efficiency operating range defined by the equipment efficiency relationship by adjusting the supply-return temperature difference and the flow rate of each branch.

[0034] In another embodiment, time-of-use electricity prices or carbon intensity signals are introduced as time weights into the power component of the objective function. The total system power is minimized using a weighted approach across different time periods, and a maximum system power constraint is superimposed during demand-constrained periods. Changes in the time weights and power upper limit influence the solution of the next upper-level rolling optimization, causing the reference setpoint to adaptively adjust with the electricity price or carbon intensity and demand boundaries. If a signal is temporarily unavailable, the default weights are reverted while the hard constraint of the demand upper limit is retained, ensuring strategy stability and energy compliance.

[0035] As an example, this application also provides an intelligent energy-saving control device for temperature regulation equipment, comprising: a model building module for acquiring the topology and parameters of the temperature regulation pipeline network and generating a unified model including thermal relationships, hydraulic relationships, and equipment efficiency relationships; an objective function generation module for generating an objective function with the minimum total system power as the main objective, while simultaneously balancing temperature difference deviation and execution quantity changes; a boundary setting module for setting the temperature range, flow rate and pressure difference range, and the safe operating boundaries of the actuators and equipment; a rolling optimization module for running a rolling optimization controller based on the unified model and external disturbance predictions, solving for the reference setpoint and sending it to the execution layer; and a tracking control module for running a tracking controller with a second sampling period, outputting control signals to water pumps, fans, valves, and compressors or heat pumps, so that the measured state tracks the reference setpoint. The device may also include an online identification module, a distributed coordinated solution module, an adaptive adjustment module, a degradation control module, a source-side collaborative tuning module, and a time weighting module, respectively used to implement the corresponding functions in the above methods. Each of the above modules can be implemented by a processor in an edge controller, industrial computer, or server.

[0036] In one embodiment, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the intelligent energy-saving control method for any of the above-described temperature regulation devices.

[0037] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent energy-saving control method for a temperature regulation device, characterized in that, Includes the following steps: The topology and parameters of the temperature regulation pipeline network are collected to generate a unified model that includes thermal, hydraulic, and equipment efficiency relationships. The thermal relationships include the pipeline network diagram structure and the coupling relationship between the supply and return temperature difference and the flow rate of the terminal heat exchanger. The pipeline network diagram structure includes nodes, branches and their lengths, diameters, local resistances, and insulation layer parameters. The hydraulic relationships include the correspondence between the flow rate and pressure difference of the branches and the continuity constraints of the nodes. The equipment efficiency relationships include the efficiency, speed, and flow rate of the water pumps and fans, as well as the coefficient of performance of the compressor or heat pump and the correspondence between the load and the supply and return temperatures. Generate an objective function that takes the minimum total system power as the main objective, while also weighing the temperature difference deviation and the change in the execution quantity. The total system power consists of pumping power, fan power and source-side power. Set the temperature range, flow rate and differential pressure range, as well as the safe operating boundaries of the actuator and equipment; Based on the unified model and external disturbance prediction, the rolling optimization controller is run to solve for the reference setpoints and send them down to the execution layer. The reference setpoints include pump speed, fan speed, valve opening, source-side frequency and supply-return temperature difference. The tracking controller operates in the second sampling period, outputting control signals to the water pump, fan, valve, and compressor or heat pump to make the measured state track the reference set value.

2. The intelligent energy-saving control method for temperature regulation equipment according to claim 1, characterized in that, Temperature, flow rate, pressure difference, and power data are collected during the first sampling period. It is determined whether the model residual of the unified model exceeds the model residual threshold. If it does, the parameters of the unified model are updated until the model residual does not exceed the model residual threshold. The first sampling period is 1 to 10 seconds, and the model residual threshold is calibrated during the debugging phase.

3. The intelligent energy-saving control method for temperature regulation equipment according to claim 1, characterized in that, When the pipeline network is a multi-branch coupled network, the rolling optimization controller adopts a distributed coordinated solution method, which divides the entire network into local subproblems of each branch that can be solved in parallel, and applies consistency constraints to shared variables such as total flow and main pipeline pressure difference. When the change of the reference set obtained in two adjacent iterations is lower than the consistency threshold, a coordinated reference set covering the entire network is generated and output.

4. The intelligent energy-saving control method for temperature regulation equipment according to claim 1, characterized in that, The temperature difference deviation is measured by the weighted sum of the squares of the deviations between the set temperature and the measured temperature in each control zone, and the change in the execution quantity is measured by the weighted sum of the changes in valve opening and rotation speed at adjacent times; the power weight, temperature difference deviation weight, and execution change weight are configurable parameters.

5. The intelligent energy-saving control method for temperature regulation equipment according to claim 1, characterized in that, The temperature range includes the upper and lower limits of the temperature of each area and the supply-return temperature difference, as well as the maximum rate of rise and fall; the flow rate and differential pressure range includes the upper and lower limits of flow rate, differential pressure, and valve opening, wherein the minimum stable valve opening is not less than 5%; the safe operating boundaries include the minimum stable frequency of the water pump and fan, the minimum start-up time and the minimum shut-off time of the compressor, and the maximum slope limit for start-up and shutdown; the method also includes applying node flow continuity and branch differential pressure closure constraints to the pipeline network.

6. The intelligent energy-saving control method for temperature regulating equipment according to claim 1, characterized in that, The rolling optimization controller solves for the reference setpoint set for a prediction time domain of not less than 15 minutes with an upper-level optimization cycle of 30 to 120 seconds and then issues the result; the tracking controller performs feedforward and proportional-integral-derivative composite control with a second sampling cycle of 1 to 10 seconds to track the reference trajectory and suppress short-term disturbances.

7. The intelligent energy-saving control method for temperature regulation equipment according to claim 1, characterized in that, The method also includes coordinated tuning of the source side and the pipeline side: In a multi-heat-source or parallel-connected cold-source system, the real-time performance coefficients and start-stop constraints of each heat source or cold source are calculated, and the selection results and parallel configuration of the heat source or cold source are generated and sent out for execution. In a single heat source or cold source system, the operating point of the compressor or heat pump is kept in the high-efficiency operating range defined by the equipment efficiency relationship by adjusting the supply and return temperature difference and the flow rate of each branch.

8. The intelligent energy-saving control method for temperature regulation equipment according to claim 1, characterized in that, Time-of-use electricity price or carbon intensity signal is introduced as a time weight into the power part of the objective function. The total power of the system is minimized in different time periods, and the upper limit constraint of the total power of the system is superimposed during the demand-limited period. The changes in time weight and power upper limit are used to affect the solution of the next upper-level rolling optimization, so that the reference setpoint is adaptively adjusted with the electricity price or carbon intensity and demand boundary.

9. The intelligent energy-saving control method for temperature regulation equipment according to claim 4, characterized in that, The method further includes: calculating the total power deviation and temperature deviation of the system; when any deviation exceeds the corresponding trigger threshold, adjusting the power weight, temperature deviation weight and execution change weight by a preset step size, and tightening or relaxing the execution change rate limit; when the limit is exceeded multiple times consecutively or the deviation shows a systematic trend, triggering the parameter re-identification of the unified model; and adaptively adjusting the amplitude limit and the shortest hold time constraint.

10. The intelligent energy-saving control method for temperature regulating equipment according to claim 1, characterized in that, When a critical sensor failure or communication anomaly is detected, switch to degraded control mode, freeze the most recently stable reference setpoint, limit the total system power to the safe power limit, limit the rate of change of the execution quantity to the safe rate of change, and maintain the area temperature within the safe temperature difference range until the fault is cleared and rolling optimization is resumed. The degraded control mode includes: enabling redundant sensor arbitration, suppressing single-point drift by means of median or voting, outputting conservative control commands, and recording event logs; after the sensor and communication self-tests pass, exiting the degraded control mode in the order of tracking, unfreezing, relaxing restrictions, and restoring rolling optimization.