Running state regulation and control system and method based on heat supply pipe network

Abstract

The invention discloses an operation state regulation and control system and method based on a heat supply pipe network, and relates to the technical field of intelligent control. Multi-dimensional data of the heat supply pipe network are integrated and collected in real time; a steady-state water conservancy model, a dynamic heat transfer model and a thermal characteristic model are constructed by utilizing multi-dimensional data of the heat supply pipe network, and the three models are combined to form a digital twin model of the heat supply pipe network; training the historical data by using machine learning to obtain a thermal load prediction model, and outputting the thermal load of the future heat supply pipe network; setting a control target according to the predicted total heat load, solving a decision variable according to the control target on the basis of a constraint condition in the digital twin model, outputting a control instruction sequence according to the decision variable, and taking the control instruction sequence as a feedforward instruction; acquiring an expected value and a measured value of each feedforward instruction output by the digital twin model to calculate an error, and calculating a feedback correction amount by using the error; and combining and superposing the feedforward instruction and the feedback correction, outputting a final control instruction, and executing the final control instruction.

Description

A system and method for controlling the operation status of a heating network Technical Field

[0001] This invention relates to the field of intelligent control technology, specifically to a system and method for regulating the operation status of a heating network. Background Technology

[0002] With the popularization of industrial automation technology, the regulation and control of heating pipe networks has entered the stage of "automatic monitoring + static control". The core progress includes: First, the initial construction of a monitoring system, deploying temperature, pressure and flow sensors along the pipeline and at heat exchange stations, realizing automatic data acquisition and local display through PLC, and introducing SCADA systems in some projects to achieve remote monitoring; Second, the realization of static hydraulic balance, using self-regulating flow control valves, differential pressure control valves and other equipment to statically distribute the flow of each branch of the pipeline, solving the basic problem of "overheating at the near end and undercooling at the far end"; Third, semi-automatic adjustment of heat source, automatically adjusting boiler output or primary network water supply temperature of heat exchange stations according to outdoor temperature and preset "temperature-load" curve, shortening the control cycle to the hour level. The technology at this stage has greatly improved heating stability and reduced manual operation and maintenance costs, but there are still limitations: traditional regulation mainly relies on "feedback control", but the heating network has the characteristics of large inertia and large delay. By the time problems are discovered and adjustments are made, users have been uncomfortable for a long time; thermal power plants or boilers often operate according to fixed curves and cannot cope with dynamic hydraulic and thermal changes in the pipeline network, or respond quickly to sudden weather changes or changes in user demand, resulting in energy waste or insufficient heating. Summary of the Invention

[0003] The purpose of this invention is to provide an operation status control system and method based on heating pipe networks to solve the problems raised in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for regulating the operation status of a heating network, the method comprising the following steps: S100, constructing a unified data platform to integrate and collect multi-dimensional data of the heating network in real time; further, the specific steps for integrating and collecting multi-dimensional data of the heating network in real time are as follows: S101, the multi-dimensional data includes meteorological data, user-side data, network operation data, and heat source data; the meteorological data is obtained from the meteorological department's database through an API interface, specifically including ambient temperature, wind speed, solar radiation intensity, and humidity; user-side and network data are collected in real time to ensure that the data accurately reflects the current real state of the heating system and the external environment, providing a fresh and reliable input basis for subsequent model construction and prediction.

[0005] The user-side data is obtained by deploying sensors in the user's room to acquire the indoor temperature and by acquiring historical and real-time heat load data from the heat meter of the user's residence; the pipeline operation data includes the pressure, temperature, flow rate, valve opening, and pump frequency of the heating pipeline network; the heat source data represents the adjustable output range, ramp rate, real-time heat supply, and water supply temperature of the heat source of the heating pipeline network.

[0006] The collection of multi-dimensional data covers four core categories: meteorology, user-side data, pipeline operation, and heat sources. This includes both external environmental influencing factors and internal system operating parameters and actual user demand data, avoiding the one-sided control caused by a single data dimension. The unified data platform breaks down data silos, integrating and managing data scattered across meteorological departments, user terminals, pipeline equipment, and heat sources, reducing redundant data transmission and processing steps, and improving data utilization efficiency.

[0007] S200. Construct a steady-state hydraulic model, a dynamic heat transfer model, and a thermal characteristic model using multi-dimensional data from the heating network. Combine these three models to form a digital twin model of the heating network. Further, the specific steps for combining these three models to form the digital twin model of the heating network are as follows: S201. Set heat exchange stations, user heating areas, and heat sources in the heating network as heating nodes. Set the pipes connecting all heating nodes as pipe segments. Use the heating nodes and pipe segments to construct a topology diagram of the heating network. In the topology diagram, professionals set initial resistance coefficient values ​​in each pipe segment. Use network operation data as input and a genetic algorithm to correct the resistance coefficients in the topology diagram. The specific steps for constructing the steady-state hydraulic model are: △H = S × G 2 In the formula, ΔH represents the pressure drop in the pipe section, S represents the pipe section resistance coefficient, and G represents the fluid flow rate in the pipe section; S202, the dynamic heat transfer model is constructed for the heating pipeline at the same location at the same time using a one-dimensional unsteady heat transfer equation, specifically as follows: In the formula, T represents the fluid temperature inside the pipe, v represents the fluid velocity, U represents the overall heat transfer coefficient of the pipe, D represents the outer diameter of the pipe, ρ represents the fluid density, and C represents the fluid density. p T represents the specific heat capacity of the fluid, A represents the cross-sectional area of ​​the pipe, and T represents the specific heat capacity of the fluid. a The value represents the ambient temperature of the pipeline, and x represents a location within the axial distance the pipeline extends from the heat source. The value represents the infinitesimal length taken at position x in the pipe; the left side of the equation represents the total rate of change of fluid temperature in the pipe, obtained by adding the rate of change of temperature with time to the rate of change of temperature with flow; the right side of the equation represents the rate of change of temperature caused by the fluid dissipating heat to the environment through the pipe wall; the negative sign indicates heat dissipation from the fluid to the environment; S203, in the heating network, each user's heating area is simplified into a network composed of thermal resistance and heat capacity, where thermal resistance represents the ability of the building envelope of the user's heating area to impede heat loss, and heat capacity represents the ability of the building structure to store heat; the thermal characteristic model structure is defined using a first-order equivalent model, specifically: In the formula, T room (t) represents the indoor temperature, T supply (t) represents the fluid temperature in the heating pipe, I solar (t) represents the solar radiation intensity, α represents the effective solar radiation heat-receiving area, and T b (t) represents the outdoor temperature, R wall Let C represent the heat dissipation resistance of the building structure in the user's heating area to the outside, C represent the equivalent heat capacity of the indoor air, and R represent the equivalent thermal resistance from the heating network to the indoor air. Historical thermal characteristic data is collected for each user's heating area, including outdoor temperature, solar radiation intensity, heating network output, fluid temperature, and indoor temperature. Outdoor temperature, solar radiation intensity, heating network output, and average fluid temperature are used as inputs, and indoor temperature is used as the output. The equivalent heat capacity C of the indoor air, the equivalent thermal resistance R from the heating network to the indoor air, and the heat dissipation resistance Routside of the building structure in the user's heating area are calculated using a system identification algorithm. wall A digital twin model of the heating network is constructed by combining a steady-state hydraulic model, a dynamic heat transfer model, and a thermal characteristic model. A topology diagram is used to build the network structure framework, and a genetic algorithm is used to correct the resistance coefficient. This allows the steady-state hydraulic model to accurately simulate the relationship between pressure drop and flow rate in pipe sections. The dynamic heat transfer model captures the changes in fluid temperature over time and space through one-dimensional unsteady-state equations. The thermal characteristic model focuses on the heat exchange process at the user end. The combination of these three models achieves a precise digital replication of the entire heating network process from heat source to user end. The models correspond to the three core physical processes of hydraulic transmission, heat transfer, and user-end thermal response, respectively, comprehensively covering the key mechanisms of heating system operation and providing a simulation environment that closely matches actual operating conditions for subsequent control decisions.

[0008] Digital twin models, as "virtual mirrors" of physical pipeline networks, can intuitively present the network's operating status and support the simulation and deduction of different control strategies, providing a visual and verifiable basis for the generation of feedforward commands.

[0009] S300. A heat load prediction model is obtained by training historical data using machine learning, and the heat load of the future heating network is output. Further, the specific steps for outputting the heat load of the future heating network are as follows: S301. The historical data includes historical heat load data, historical meteorological data, and time data. The historical data is divided into a training set and a validation set according to the time series, and a heat load prediction model is obtained by training using a Long Short-Term Memory (LSTM) network. S302. Future meteorological data, historical heat load of the same period, date type, and time are used as input features of the model. The historical heat load of the same period represents the heat load in history that is the same as the current predicted future time. For example, if the predicted future date is March 10th, the historical heat load of the same period is the heat load of March 10th of last year. The date type includes weekdays, holidays, etc., and the time represents the peak periods of 8 am and 7 pm. The heat load prediction model outputs the predicted total heat load Q for the future time t. heat (t) For each heating node in the topology of the digital twin model, the proportion of the historical heat load of each heating node to the total heat load is used as an input feature. The heat load prediction model outputs the heat load proportion of each heating node, and the heat load of each heating node is obtained by multiplying the heat load proportion by the predicted total heat load. Based on time series training of the Long Short-Term Memory network, combined with historical heat load, meteorological data, and time features, especially the introduction of historical heat load data from the same period, the influence of seasonality and periodicity on heat load is fully considered, which greatly improves the prediction accuracy.

[0010] It not only outputs the total heat load but also achieves precise allocation of the total heat load to each heating node by using the historical load ratio of each node. This provides a basis for subsequent node-specific control and avoids localized overheating or underheating caused by "one-size-fits-all" control. By outputting the heat load for the next time t in advance, subsequent control commands can be planned ahead of time, avoiding passive responses and reserving adjustment time for optimized system operation.

[0011] S400. Set a control objective based on the predicted total heat load, solve for the decision variables in the digital twin model based on the control objective, and output a control command sequence based on the decision variables, using the control command sequence as feedforward commands; further, the specific steps for outputting the control command sequence based on the decision variables are as follows: S401. Set the control cycle of the heating network based on the time series, and set the control objective in each control cycle of the heating network to minimize the total operating cost while satisfying the predicted total heat load and the heat load of the heating nodes; set the decision variable U ff (t) = {T s，set (t), G total (t), K i (t)}, T s，set (t) represents the fluid temperature setpoint, G total (t) represents the total circulation flow, K i(t) represents the opening degree of the i-th valve; the decision variables include the fluid temperature setpoint, total circulation flow rate, and valve opening degree, covering the two core links of heat source output and pipeline transmission, and can realize multi-dimensional coordinated control.

[0012] Set the objective function as follows: In the objective function, J represents the total operating cost, and C... fuel C represents the unit price of fuel. elec Q represents the unit price of electricity. fuel (t) represents the fuel consumption at time t, P batt (t) represents the power consumption at time t; Δt represents the time step, and M represents the total number of time points within the control cycle; S402, set constraints, including equality constraints and inequality constraints; the equality constraints include the equations of three models in the digital twin model: steady-state hydraulic model, dynamic heat transfer model, and thermal characteristic model, requiring the decision variables to conform to the three models; the inequality constraints include equipment capacity limits, safety limits, and heat source ramp-up rate limits; where the equipment capacity limit is specifically: T s，min ≤T s，set (t)≤T s，max T s，min This indicates the minimum fluid temperature setpoint that the heating network equipment can provide, in T. s，max Indicates the maximum fluid temperature setpoint that the heating network equipment can provide; G min ≤G total (t)≤G max G min G represents the minimum circulating flow rate of the heating network. max This indicates the maximum circulating flow rate of the heating network; the specific safety limit is: △P min ≤△H≤△P max , △P min Indicates the minimum pressure drop in the pipe section, ΔP max This indicates the maximum pressure drop in the pipe section; the specific limit for the heat source ramp-up rate is: |T s，set (t+1)-T s，set (t)|≤△T max ×△t;T s，set (t+1) represents the fluid temperature setpoint at the next time step, ΔT max This represents the maximum change in fluid temperature within the heating network per unit time; equality constraints ensure that decisions conform to the physical laws of the model, while inequality constraints ensure that the system operates within the equipment's carrying capacity and safety threshold, avoiding equipment damage or safety hazards caused by excessive regulation.

[0013] S403. Using a model predictive control framework, the system solves the problem based on the control objective and constraints, and outputs the optimal control command U for the future time period. ff (t) = {Ts，set (t), G total (t), K i (t)}; Based on the time series, the control command at time t=0 is sent to the heat source control system of the heating network to form a feedforward command. Based on the model predictive control framework, the optimal control command for the future time period is output and sent out in advance as a feedforward command, so that the system can adjust its operating state in advance to adapt to future heat load demand.

[0014] S500. Collect the expected and measured values ​​of each feedforward instruction output by the digital twin model, calculate the error, and use the error to calculate the feedback correction amount; further, the specific steps for calculating the feedback correction amount using the error are as follows: S501. Collect the expected and measured values ​​of each feedforward instruction output by the digital twin model, where the expected value of the feedforward instruction represents the decision variable within the predicted feedforward instruction output by the digital twin model; subtract the expected value from the measured value to obtain the variable error, and calculate the feedback correction amount for each type of variable error using the proportional-integral-differential algorithm, the formula being: In the formula, U p U c U d The coefficients for the proportional, integral, and differential terms are represented by e(t), which are obtained through expert experience; e(t) represents the variable error, t represents the upper limit of integration, and ΔU represents the coefficients for the integral term. fb (t) represents the feedback correction amount, which is sent to the heat source control system as a feedback command. The variable error is obtained by the difference between the measured value and the expected value. Combined with the proportional-integral-derivative (PID) algorithm, the error can be responded to quickly, the steady-state deviation can be eliminated, and the fluctuation can be suppressed, so as to achieve accurate correction of the feedforward command.

[0015] Dynamic corrections are made to address deviations between model predictions and actual operation, compensating for minor differences between the digital twin model and the physical system, as well as prediction errors caused by unforeseen external factors, thereby improving the system's anti-interference capability. As a supplement to feedforward control, feedback correction transforms the control process from "one-way prediction-execution" to a closed loop of "prediction-execution-feedback-correction," ensuring that the control effect continuously aligns with actual needs.

[0016] S600 combines the feedforward command and the feedback correction amount, superimposes them to output the final control command, and executes it.

[0017] Furthermore, the specific steps for combining and superimposing the feedforward command and feedback correction to output the final control command are as follows: S601, the same decision variable in the feedforward command and feedback command output by the digital twin model are combined to generate the final control command, the formula being: U final (t) = U ff (t) + △U fb (t), U final(t) represents the final control command; the heat source control system of the heating network executes the final control command. The final command is directly sent to the heat source control system for execution, reducing intermediate transmission links, ensuring that the command is implemented quickly, and adjusting the system operating status in a timely manner. The synergistic effect of feedforward and feedback enables the final command to accurately match the actual heat load demand, avoiding the prediction deviation of single feedforward control or the lag of single feedback control, and ensuring stable heating quality.

[0018] An operational status control system based on a heating network is disclosed. The system includes a data acquisition module, a digital twin module, a heat load prediction module, a feedforward instruction module, a feedback instruction module, and a final instruction execution module. The data acquisition module constructs a unified data platform, integrating and collecting multi-dimensional data from the heating network in real time. The digital twin module utilizes the multi-dimensional data to construct a steady-state hydraulic model, a dynamic heat transfer model, and a thermal characteristic model, combining these three models to form a digital twin model of the heating network. The heat load prediction module trains a heat load prediction model using machine learning on historical data, outputting the future heat load of the heating network. The feedforward instruction module sets a control target based on the predicted total heat load, solves for decision variables in the digital twin model based on the control target, and outputs a sequence of control instructions, which serves as the feedforward instruction. The feedback instruction module collects the expected and measured values ​​of each feedforward instruction output by the digital twin model, calculates the error, and uses the error to calculate a feedback correction. The final instruction execution module combines the feedforward instructions and the feedback correction, superimposes them, outputs the final control instruction, and executes it.

[0019] The digital twin module includes a steady-state hydraulic model unit, a dynamic heat transfer model unit, and a thermal characteristic model unit. The steady-state hydraulic model unit is used to set heat exchange stations, user heating areas, and heat sources in the heating network as heating nodes, and pipes connecting all heating nodes as pipe segments. The heating nodes and pipe segments are used to construct a topology diagram of the heating network, and resistance coefficients are set in the topology diagram to construct a steady-state hydraulic model. The dynamic heat transfer model unit is used to construct a dynamic heat transfer model for heating pipes at the same location at the same time using a one-dimensional unsteady heat transfer equation. The thermal characteristic model unit is used to simplify each user heating area in the heating network into a network composed of thermal resistance and heat capacity, and defines the thermal characteristic model structure using a first-order equivalent model.

[0020] The feedforward instruction module includes a control target unit and a constraint unit. The control target unit is used to set the control cycle of the heating network based on the time series. In each control cycle of the heating network, the control target is set to minimize the total operating cost while satisfying the predicted total heat load and the heat load of the heating nodes. The constraint unit is used to set constraints, which include equality constraints and inequality constraints.

[0021] Compared with the prior art, the beneficial effects of the present invention are: 1. The present invention achieves precise control of each heating node through accurate heat load prediction, refined load distribution, and closed-loop command correction, effectively avoiding local overheating and underheating problems, ensuring stable indoor temperature for users, and improving heating satisfaction.

[0022] 2. This invention incorporates safety constraints and equipment capacity constraints throughout the entire process, and combines feedback correction to promptly correct abnormal deviations, ensuring that the system always operates within safe, compliant, and equipment-capacity limits, thereby reducing operational risks.

[0023] 3. This invention enables the system to minimize fuel and electricity consumption and reduce total operating costs while meeting heating demand through accurate simulation of digital twin models, advance prediction of heat load, and feedforward-feedback coordinated control. At the same time, it avoids ineffective equipment operation and extends equipment service life. Attached Figure Description

[0024] Figure 1 is a schematic diagram of the steps of a method for regulating the operation status of a heating network according to the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on 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.

[0026] Example: As shown in Figure 1, the present invention provides a technical solution, a method for regulating the operation status of a heating network, the method comprising the following steps: S100, constructing a unified data platform to integrate and collect multi-dimensional data of the heating network in real time; the specific steps for integrating and collecting multi-dimensional data of the heating network in real time are as follows: S101, the multi-dimensional data includes meteorological data, user-side data, network operation data, and heat source data; the meteorological data is obtained from the meteorological department database through an API interface, specifically including ambient temperature, wind speed, solar radiation intensity, and humidity; user-side and network data are collected in real time to ensure that the data can accurately reflect the real state of the current heating system and external environment, providing a fresh and reliable input basis for subsequent model construction and prediction.

[0027] The user-side data is obtained by deploying sensors in the user's room to acquire the indoor temperature and by acquiring historical and real-time heat load data from the heat meter of the user's residence; the pipeline operation data includes the pressure, temperature, flow rate, valve opening, and pump frequency of the heating pipeline network; the heat source data represents the adjustable output range, ramp rate, real-time heat supply, and water supply temperature of the heat source of the heating pipeline network.

[0028] The collection of multi-dimensional data covers four core categories: meteorology, user-side data, pipeline operation, and heat sources. This includes both external environmental influencing factors and internal system operating parameters and actual user demand data, avoiding the one-sided control caused by a single data dimension. The unified data platform breaks down data silos, integrating and managing data scattered across meteorological departments, user terminals, pipeline equipment, and heat sources, reducing redundant data transmission and processing steps, and improving data utilization efficiency.

[0029] S200. Construct a steady-state hydraulic model, a dynamic heat transfer model, and a thermal characteristic model using multi-dimensional data from the heating network. Combine these three models to form a digital twin model of the heating network. The specific steps for combining these three models to form a digital twin model of the heating network are as follows: S201. Set heat exchange stations, user heating areas, and heat sources in the heating network as heating nodes. Set the pipes connecting all heating nodes as pipe segments. Use the heating nodes and pipe segments to construct a topology diagram of the heating network. In the topology diagram, professionals set initial resistance coefficient values ​​in each pipe segment. Use the network operation data as input and use a genetic algorithm to correct the resistance coefficients in the topology diagram. The specific steps for constructing the steady-state hydraulic model are: △H=S×G 2 In the formula, ΔH represents the pressure drop in the pipe section, S represents the pipe section resistance coefficient, and G represents the fluid flow rate in the pipe section; S202, the dynamic heat transfer model is constructed for the heating pipeline at the same location at the same time using a one-dimensional unsteady heat transfer equation, specifically as follows: In the formula, T represents the fluid temperature inside the pipe, v represents the fluid velocity, U represents the overall heat transfer coefficient of the pipe, D represents the outer diameter of the pipe, ρ represents the fluid density, and C represents the fluid density. p T represents the specific heat capacity of the fluid, A represents the cross-sectional area of ​​the pipe, and T represents the specific heat capacity of the fluid. a The value represents the ambient temperature of the pipeline, and x represents a location within the axial distance the pipeline extends from the heat source. The value represents the infinitesimal length taken at position x in the pipe; the left side of the equation represents the total rate of change of fluid temperature in the pipe, obtained by adding the rate of change of temperature with time to the rate of change of temperature with flow; the right side of the equation represents the rate of change of temperature caused by the fluid dissipating heat to the environment through the pipe wall; the negative sign indicates heat dissipation from the fluid to the environment; S203, in the heating network, each user's heating area is simplified into a network composed of thermal resistance and heat capacity, where thermal resistance represents the ability of the building envelope of the user's heating area to impede heat loss, and heat capacity represents the ability of the building structure to store heat; the thermal characteristic model structure is defined using a first-order equivalent model, specifically: In the formula, T room (t) represents the indoor temperature, T supply (t) represents the fluid temperature in the heating pipe, I solar (t) represents the solar radiation intensity, α represents the effective solar radiation heat-receiving area, and T b(t) represents the outdoor temperature, R wall Let C represent the heat dissipation resistance of the building structure in the user's heating area to the outside, C represent the equivalent heat capacity of the indoor air, and R represent the equivalent thermal resistance from the heating network to the indoor air. Historical thermal characteristic data is collected for each user's heating area, including outdoor temperature, solar radiation intensity, heating network output, fluid temperature, and indoor temperature. Outdoor temperature, solar radiation intensity, heating network output, and average fluid temperature are used as inputs, and indoor temperature is used as the output. The equivalent heat capacity C of the indoor air, the equivalent thermal resistance R from the heating network to the indoor air, and the heat dissipation resistance Routside of the building structure in the user's heating area are calculated using a system identification algorithm. wall A digital twin model of the heating network is constructed by combining a steady-state hydraulic model, a dynamic heat transfer model, and a thermal characteristic model. A topology diagram is used to build the network structure framework, and a genetic algorithm is used to correct the resistance coefficient. This allows the steady-state hydraulic model to accurately simulate the relationship between pressure drop and flow rate in pipe sections. The dynamic heat transfer model captures the changes in fluid temperature over time and space through one-dimensional unsteady-state equations. The thermal characteristic model focuses on the heat exchange process at the user end. The combination of these three models achieves a precise digital replication of the entire heating network process from heat source to user end. The models correspond to the three core physical processes of hydraulic transmission, heat transfer, and user-end thermal response, respectively, comprehensively covering the key mechanisms of heating system operation and providing a simulation environment that closely matches actual operating conditions for subsequent control decisions.

[0030] Digital twin models, as "virtual mirrors" of physical pipeline networks, can intuitively present the network's operating status and support the simulation and deduction of different control strategies, providing a visual and verifiable basis for the generation of feedforward commands.

[0031] S300. A heat load prediction model is obtained by training historical data using machine learning, and the heat load of the future heating network is output. The specific steps for outputting the heat load of the future heating network are as follows: S301. The historical data includes historical heat load data, historical meteorological data, and time data. The historical data is divided into a training set and a validation set according to the time series, and a heat load prediction model is obtained by training using a Long Short-Term Memory (LSTM) network. S302. The future meteorological data, historical heat load of the same period, date type, and time are used as input features of the model. The historical heat load of the same period refers to the heat load in the past that is the same as the current predicted future time. For example, the predicted heat load for the future date is March 10, and the historical heat load of the same period is the heat load of March 10 last year. The date type includes weekdays, holidays, etc., and the time refers to the peak periods of 8 am and 7 pm in a day. The heat load prediction model outputs the predicted total heat load Q for the future time t. heat(t) For each heating node in the topology of the digital twin model, the proportion of the historical heat load of each heating node to the total heat load is used as an input feature. The heat load prediction model outputs the heat load proportion of each heating node, and the heat load of each heating node is obtained by multiplying the heat load proportion by the predicted total heat load. Based on time series training of the Long Short-Term Memory network, combined with historical heat load, meteorological data, and time features, especially the introduction of historical heat load data from the same period, the influence of seasonality and periodicity on heat load is fully considered, which greatly improves the prediction accuracy.

[0032] It not only outputs the total heat load but also achieves precise allocation of the total heat load to each heating node by using the historical load ratio of each node. This provides a basis for subsequent node-specific control and avoids localized overheating or underheating caused by "one-size-fits-all" control. By outputting the heat load for the next time t in advance, subsequent control commands can be planned ahead of time, avoiding passive responses and reserving adjustment time for optimized system operation.

[0033] S400. Set the control objective based on the predicted total heat load, solve the decision variables in the digital twin model based on the control objective, and output the control command sequence based on the decision variables, using the control command sequence as feedforward commands. The specific steps for outputting the control command sequence based on the decision variables are as follows: S401. Set the control cycle of the heating network based on the time series. In each control cycle of the heating network, set the control objective to minimize the total operating cost while satisfying the predicted total heat load and the heat load of the heating nodes; set the decision variable U. ff (t) = {T s，set (t), G total (t), K i (t)}, T s，set (t) represents the fluid temperature setpoint, G total (t) represents the total circulation flow, K i (t) represents the opening degree of the i-th valve; the decision variables include the fluid temperature setpoint, total circulation flow rate, and valve opening degree, covering the two core links of heat source output and pipeline transmission, and can realize multi-dimensional coordinated control.

[0034] Set the objective function as follows: In the objective function, J represents the total operating cost, and C... fuel C represents the unit price of fuel. elec Q represents the unit price of electricity. fuel (t) represents the fuel consumption at time t, P batt(t) represents the power consumption at time t; Δt represents the time step, and M represents the total number of time points within the control cycle; S402, set constraints, including equality constraints and inequality constraints; the equality constraints include the equations of three models in the digital twin model: steady-state hydraulic model, dynamic heat transfer model, and thermal characteristic model, requiring the decision variables to conform to the three models; the inequality constraints include equipment capacity limits, safety limits, and heat source ramp-up rate limits; where the equipment capacity limit is specifically: T s，min ≤T s，set (t)≤T s，max T s，min This indicates the minimum fluid temperature setpoint that the heating network equipment can provide, in T. s，max Indicates the maximum fluid temperature setpoint that the heating network equipment can provide; G min ≤G total (t)≤G max G min G represents the minimum circulating flow rate of the heating network. max This indicates the maximum circulating flow rate of the heating network; the specific safety limit is: △P min ≤△H≤△P max , △P min Indicates the minimum pressure drop in the pipe section, ΔP max This indicates the maximum pressure drop in the pipe section; the specific limit for the heat source ramp-up rate is: |T s，set (t+1)-T s，set (t)|≤△T max ×△t;T s，set (t+1) represents the fluid temperature setpoint at the next time step, ΔT max This represents the maximum change in fluid temperature within the heating network per unit time; equality constraints ensure that decisions conform to the physical laws of the model, while inequality constraints ensure that the system operates within the equipment's carrying capacity and safety threshold, avoiding equipment damage or safety hazards caused by excessive regulation.

[0035] S403. Using a model predictive control framework, the system solves the problem based on the control objective and constraints, and outputs the optimal control command U for the future time period. ff (t) = {T s，set (t), G total (t), K i (t)}; Based on the time series, the control command at time t=0 is sent to the heat source control system of the heating network to form a feedforward command. Based on the model predictive control framework, the optimal control command for the future time period is output and sent out in advance as a feedforward command, so that the system can adjust its operating state in advance to adapt to future heat load demand.

[0036] S500. Collect the expected and measured values ​​of each feedforward instruction output by the digital twin model, calculate the error, and use the error to calculate the feedback correction amount. The specific steps for calculating the feedback correction amount using the error are as follows: S501. Collect the expected and measured values ​​of each feedforward instruction output by the digital twin model, where the expected value of the feedforward instruction represents the decision variable within the predicted feedforward instruction in the digital twin model; subtract the expected value from the measured value to obtain the variable error, and calculate the feedback correction amount for each type of variable error using the proportional-integral-differential algorithm, as shown in the formula: In the formula, U p U c U d The coefficients for the proportional, integral, and differential terms are represented by e(t), which are obtained through expert experience; e(t) represents the variable error, t represents the upper limit of integration, and ΔU represents the coefficients for the integral term. fb (t) represents the feedback correction amount, which is sent to the heat source control system as a feedback command. The variable error is obtained by the difference between the measured value and the expected value. Combined with the proportional-integral-derivative (PID) algorithm, the error can be responded to quickly, the steady-state deviation can be eliminated, and the fluctuation can be suppressed, so as to achieve accurate correction of the feedforward command.

[0037] Dynamic corrections are made to address deviations between model predictions and actual operation, compensating for minor differences between the digital twin model and the physical system, as well as prediction errors caused by unforeseen external factors, thereby improving the system's anti-interference capability. As a supplement to feedforward control, feedback correction transforms the control process from "one-way prediction-execution" to a closed loop of "prediction-execution-feedback-correction," ensuring that the control effect continuously aligns with actual needs.

[0038] S600 combines the feedforward command and the feedback correction amount, superimposes them to output the final control command, and executes it.

[0039] The specific steps for combining and superimposing the feedforward command and feedback correction to output the final control command are as follows: S601, combine the same decision variable in the feedforward command and feedback command output by the digital twin model to generate the final control command, the formula is: U final (t) = U ff (t) + △U fb (t), U final (t) represents the final control command; the heat source control system of the heating network executes the final control command. The final command is directly sent to the heat source control system for execution, reducing intermediate transmission links, ensuring that the command is implemented quickly, and adjusting the system operating status in a timely manner. The synergistic effect of feedforward and feedback enables the final command to accurately match the actual heat load demand, avoiding the prediction deviation of single feedforward control or the lag of single feedback control, and ensuring stable heating quality.

[0040] An operational status control system based on a heating network is disclosed. The system includes a data acquisition module, a digital twin module, a heat load prediction module, a feedforward instruction module, a feedback instruction module, and a final instruction execution module. The data acquisition module constructs a unified data platform, integrating and collecting multi-dimensional data from the heating network in real time. The digital twin module utilizes the multi-dimensional data to construct a steady-state hydraulic model, a dynamic heat transfer model, and a thermal characteristic model, combining these three models to form a digital twin model of the heating network. The heat load prediction module trains a heat load prediction model using machine learning on historical data, outputting the future heat load of the heating network. The feedforward instruction module sets a control target based on the predicted total heat load, solves for decision variables in the digital twin model based on the control target, and outputs a sequence of control instructions, which serves as the feedforward instruction. The feedback instruction module collects the expected and measured values ​​of each feedforward instruction output by the digital twin model, calculates the error, and uses the error to calculate a feedback correction. The final instruction execution module combines the feedforward instructions and the feedback correction, superimposes them, outputs the final control instruction, and executes it.

[0041] The digital twin module includes a steady-state hydraulic model unit, a dynamic heat transfer model unit, and a thermal characteristic model unit. The steady-state hydraulic model unit is used to set heat exchange stations, user heating areas, and heat sources in the heating network as heating nodes, and pipes connecting all heating nodes as pipe segments. The heating nodes and pipe segments are used to construct a topology diagram of the heating network, and resistance coefficients are set in the topology diagram to construct a steady-state hydraulic model. The dynamic heat transfer model unit is used to construct a dynamic heat transfer model for heating pipes at the same location at the same time using a one-dimensional unsteady heat transfer equation. The thermal characteristic model unit is used to simplify each user heating area in the heating network into a network composed of thermal resistance and heat capacity, and defines the thermal characteristic model structure using a first-order equivalent model.

[0042] The feedforward instruction module includes a control target unit and a constraint unit. The control target unit is used to set the control cycle of the heating network based on the time series. In each control cycle of the heating network, the control target is set to minimize the total operating cost while satisfying the predicted total heat load and the heat load of the heating nodes. The constraint unit is used to set constraints, which include equality constraints and inequality constraints.

[0043] Example: This example focuses on a centralized heating network system in a city. The system covers 3 heat source plants, 8 heat exchange stations, and 20 residential communities (including 5000 households). The total length of the heating network is approximately 80km, with pipe diameters ranging from DN150 to DN600. The system operates from November 15th to March 15th of the following year (heating season). A unified data acquisition and storage system is built based on an industrial internet platform. The MQTT protocol is used to achieve real-time communication between sensors, meters, and the server, with a data transmission cycle of 15 minutes per transmission. Real-time data acquisition categories and examples: Meteorological data: December 20, 2024, 10:00 AM, ambient temperature -3℃, wind speed 2.5m / s, solar radiation intensity 200W / m². 2 Humidity 60%; User-side data: In Community A (1000 households), the average indoor temperature at 10:00 AM was 19.2℃, and the real-time heat load was 8MW; Pipeline operation data: In Pipeline segment 1 (DN300, 3km long), the pressure at 10:00 AM was 0.8MPa, the temperature was 75℃, and the flow rate was 500m³ / h. 3 / h, valve opening 60%, pump frequency 45Hz; heat source data: heat source plant 1 has an adjustable output range of 60-100MW, real-time heating capacity of 75MW, water supply temperature of 85℃, and ramp rate of 3MW / h.

[0044] Steady-state hydraulic model construction: A pipeline topology was constructed using 3 heat source plants, 8 heat exchange stations, and 20 residential communities as heating nodes, with 80km of pipelines connecting these nodes as pipe segments. Nearly one month's worth of pipeline operation data (pressure, flow rate) was input into the topology. A genetic algorithm was used to correct the resistance coefficient of each pipe segment, ultimately obtaining the resistance coefficient S = 0.03mH2O / (m²). 3 / h) 2 Based on the formula △H=S×G 2 When the flow rate of pipe section 1 is G=500m³ 3 When the pressure drop in the pipe section is calculated to be ΔH = 0.03 × 500 at a rate of / h, the pressure drop is calculated to be ΔH = 0.03 × 500. 2 =7500mH2O (actual measured value 7480mH2O, error ≤0.3%).

[0045] Dynamic heat transfer model construction: For pipe segment 1 (diameter DN300, outer diameter 325mm, cross-sectional area A=0.083m²), 2 Given that the fluid density ρ = 995 kg / m³ 3 Specific heat capacity Cp = 4.18 kJ / (kg・℃), overall heat transfer coefficient of the pipe U = 3.5 W / (m²) 2•℃), ambient temperature Ta=-3℃; based on the one-dimensional unsteady heat transfer equation, when the fluid velocity v=0.8m / s, the rate of change of fluid temperature in the pipe with axial distance x is calculated: when x=1km, the rate of change of temperature is -0.02℃ / m (that is, the fluid temperature drops by 0.02℃ for every 1m extension), and the error with the measured value (-0.021℃ / m) is ≤5%.

[0046] Thermal characteristic model construction: Community A is simplified into a "thermal resistance-heat capacity" network. Historical thermal characteristic data (outdoor temperature, solar radiation intensity, heating supply, fluid temperature, and indoor temperature) for the past year are collected. Using outdoor temperature, solar radiation intensity, heating supply, and average fluid temperature as inputs, and indoor temperature as output, the equivalent heat capacity C = 1.2 × 10⁻⁶ is obtained through a system identification algorithm. 7 kJ / ℃, equivalent thermal resistance R=0.005℃・h / kJ, heat dissipation resistance to the outside R wall =0.01℃・h / kJ; Substituting into the thermal characteristic model formula, when T supply (t) = 75℃, I solar (t) = 200W / m 2 α=8000m 2 T b When (t) = -3℃, T is calculated. room (t) = 19.3℃, with an error of ≤0.5% compared to the measured value of 19.2℃.

[0047] Load Forecast: Input Characteristics: Weather forecast data for December 21, 2024 (Saturday, a public holiday) (daily average temperature -5℃, wind speed 3m / s, solar radiation intensity 150W / m²) 2 The total heat load on December 21st last year was 45MW, with peak times at 8 AM and 7 PM. Model output: Total heat load Q on December 21st, 2024. heat (t) = 48MW; Node load allocation: Based on the historical load ratio of each community, community A accounts for 16.7% of the total load, and the daily heat load of community A is calculated to be 48 × 16.7% ≈ 8MW.

[0048] Control objectives and parameter settings: The control cycle is set to 1 hour; the control objective is to "meet the total heat load of 48MW and the load at each node, while minimizing fuel and electricity costs"; Decision variable: T s，set (t) (fluid temperature setpoint), G total (t) (total circulating flow), K i (t) (the opening degree of the i-th valve); Cost parameters: fuel price Cfuel = 2.8 yuan / kg, electricity price Celec = 0.6 yuan / kWh; Constraints: T s，min =60℃, T s，max =95℃; Gmin =3000m 3 / h, G max =8000m 3 / h;△P min =500mH2O, ΔP max =10000mH2O; △T max =2℃ / h, Δt=1h, therefore |T s，set (t+1)-T s，set (t)|≤2℃.

[0049] Solution and Command Output: Using a model predictive control framework combined with the constraints of a digital twin model, the solution outputs the feedforward command for 8:00 AM on December 21, 2024 (t=0): T s，set (0) = 88℃; G total (0) = 5500m 3 / h; The valve opening degree K1(0) corresponding to community A is 65%.

[0050] During one control cycle, data is collected at time t=0: the expected value output by the digital twin model (88℃, 5500m). 3 / h, 65%), actual measured value (87.5℃, 5480m 3 / h, 64.8%.

[0051] The specific time within the control period is as follows: Time axis: t0, t1, t2, t3, t4. Operation: At time t0: Predict [t0, t4] → Collect the measured value at t0 ← → Compare with the expected value at time t0 → Execute the t0 command; At time t1: Collect the measured value at t1 ← → Compare with the predicted expected value at t0 → Correct the t1 command; At time t2: Collect the measured value at t2 ← → Compare with the predicted expected value at t0 → Correct the t2 command; Calculate the parameter error: e_Ts(0) = 87.5 - 88 = -0.5℃; e_G(0) = 5480 - 5500 = -20m 3 / h;e_K1(0)=64.8%-65%=-0.2%.

[0052] PID Algorithm Correction: Setting PID Parameters (U p =0.8, U c =0.2, U d =0.1), substituting into the PID formula to calculate the feedback correction: -0.4℃; -16m 3 / h; -0.16%.

[0053] Calculate the final instruction: T s，final (0) = 88 + (-0.4) = 87.6℃; G final (0) = 5500 + (-16) = 5484m3 / h;K 1，final (0) = 65% + (-0.16%) = 64.84%.

[0054] The heat source control system receives the final control command, which drives the heat source plant temperature controller to adjust the supply water temperature to 87.6℃, and the flow regulating valve to adjust the total circulating flow to 5484m³. 3 / h, the electric regulating valve corresponding to community A will adjust the opening to 64.84%.

[0055] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for regulating the operation status of a heating network, characterized in that: The method includes the following steps: S100, constructing a unified data platform to integrate and collect multi-dimensional data of the heating network in real time; S200, using the multi-dimensional data of the heating network to construct a steady-state hydraulic model, a dynamic heat transfer model, and a thermal characteristic model, and combining the three models to form a digital twin model of the heating network; S300, using machine learning to train historical data to obtain a heat load prediction model, and outputting the future heat load of the heating network; S400, setting a control target based on the predicted total heat load, constructing an objective function based on the control target, and setting constraints, including equality constraints and inequality constraints; solving for decision variables based on the control target in the digital twin model based on the constraints, outputting a control command sequence based on the decision variables, and using the control command sequence as feedforward commands; S500, collecting the expected value and measured value of each feedforward command output by the digital twin model to calculate the error, and using the error to calculate the feedback correction amount; S600, combining and superimposing the feedforward commands and feedback correction amounts to output the final control command and execute it.

2. The method for regulating the operation status of a heating network according to claim 1, characterized in that: The steady-state hydraulic model in S200 is specifically as follows: Heat exchange stations, user heating areas, and heat sources in the heating network are designated as heating nodes; pipes connecting all heating nodes are designated as pipe segments; a topology diagram of the heating network is constructed using the heating nodes and pipe segments; initial resistance coefficient values ​​are set for each pipe segment in the topology diagram by professionals; network operation data is used as input; and a genetic algorithm is used to correct the resistance coefficients in the topology diagram. The steady-state hydraulic model is specifically constructed as: ΔH = S × G 2 In the formula, ΔH represents the pressure drop in the pipe section, S represents the pipe section resistance coefficient, and G represents the fluid flow rate in the pipe section. The dynamic heat transfer model is specifically constructed using a one-dimensional unsteady-state heat transfer equation for the heating pipeline at the same location at the same time. The left side of the equation represents the rate of change of temperature with time plus the rate of change of temperature with flow, which gives the total rate of change of fluid temperature in the pipeline. The right side of the equation represents the rate of change of fluid temperature due to heat dissipation to the environment through the pipe wall. The thermal characteristic model is specifically simplified in the heating network as a network composed of thermal resistance and heat capacity. Thermal resistance represents the ability of the building envelope of the user's heating area to impede heat loss, and heat capacity represents the ability of the building structure to store heat. The thermal characteristic model structure is defined using a first-order equivalent model; historical thermal characteristic data is collected for each user's heating area, including outdoor temperature, solar radiation intensity, heating network heat supply, fluid temperature and indoor temperature, with outdoor temperature, solar radiation intensity, heating network heat supply and average fluid temperature as inputs and indoor temperature as output. The equivalent heat capacity C of indoor air, the equivalent thermal resistance R from the heating network to the indoor air, and the heat dissipation resistance R of the building structure in the user's heating area to the outside are solved by a system identification algorithm. wall .

3. The method for regulating the operation status of a heating network according to claim 1, characterized in that: The control objective set according to the predicted total heat load in S400 is specifically as follows: the control cycle of the heating network is set based on the time series, and the control objective in each control cycle of the heating network is to minimize the total operating cost while satisfying the predicted total heat load and the heat load of the heating nodes. Set decision variable U ff (t) = {T s，set (t), G total (t), K i (t)}, T s，set (t) represents the fluid temperature setpoint, G total (t) represents the total circulation flow, K i (t) represents the opening degree of the i-th valve.

4. The method for regulating the operation status of a heating network according to claim 1, characterized in that: The S400 equation constraints include three model equations in the digital twin model: steady-state hydraulic model, dynamic heat transfer model, and thermal characteristic model. The decision variables must conform to the three models.

5. The method for regulating the operation status of a heating network according to claim 3, characterized in that: The inequality constraints in S400 include equipment capacity limitations, safety limitations, and heat source ramp-up rate limitations; specifically, the equipment capacity limitation is: T s，min ≤T s，set (t)≤T s，max T s，min This indicates the minimum fluid temperature setpoint that the heating network equipment can provide, in T. s，max Indicates the maximum fluid temperature setpoint that the heating network equipment can provide; G min ≤G total (t)≤G max G min G represents the minimum circulating flow rate of the heating network. max This indicates the maximum circulating flow rate of the heating network; the specific safety limit is: △P min ≤△H≤△P max , △P min Indicates the minimum pressure drop in the pipe section, ΔP max This indicates the maximum pressure drop in the pipe section; the specific limit for the heat source ramp-up rate is: |T s，set (t+1)-T s，set (t)|≤△T max ×△t;T s，set (t+1) represents the fluid temperature setpoint at the next time step, ΔT max This indicates the maximum change in fluid temperature within the heating network per unit time.

6. The method for regulating the operation status of a heating network according to claim 1, characterized in that: The specific steps for calculating the feedback correction amount using the error in S500 are as follows: the expected value and the measured value of each feedforward command output by the digital twin model are collected respectively, the variable error is obtained by subtracting the expected value from the measured value, the feedback correction amount is calculated using the proportional-integral-differential algorithm for each type of variable error, and the feedback correction amount is sent to the heat source control system as a feedback command.

7. The method for regulating the operation status of a heating network according to claim 1, characterized in that: In S600, the combination and superposition of feedforward commands and feedback corrections to output and execute the final control command specifically involves: The feedforward and feedback commands output from the digital twin model are combined collaboratively to generate the final control command, as shown in the formula: U final (t) = U ff (t) + △U fb (t), U final (t) represents the final control command, U ff (t) represents the feedforward instruction, △U fb (t) indicates a feedback command; the heat source control system of the heating network executes the final control command.

8. A system for controlling the operation status of a heating network, characterized in that: The operation status control system includes a data acquisition module, a digital twin module, a heat load prediction module, a feedforward instruction module, a feedback instruction module, and a final instruction execution module. The data acquisition module is used to build a unified data platform, integrating and collecting multi-dimensional data from the heating network in real time. The digital twin module uses the multi-dimensional data from the heating network to construct a steady-state hydraulic model, a dynamic heat transfer model, and a thermal characteristic model, combining these three models to form a digital twin model of the heating network. The heat load prediction module uses machine learning to train a heat load prediction model on historical data, outputting the future heat load of the heating network. The feedforward instruction module sets control targets based on the predicted total heat load, solves for decision variables in the digital twin model based on the control targets, outputs a sequence of control instructions based on the decision variables, and uses this sequence as feedforward instructions. The feedback instruction module is used to collect the expected and measured values ​​of each feedforward instruction output by the digital twin model to calculate the error, and use the error to calculate the feedback correction amount; the final instruction execution module is used to combine the feedforward instruction and the feedback correction amount, superimpose them to output the final control instruction and execute it.

9. The operation status control system based on a heating network according to claim 8, characterized in that: The digital twin module includes a steady-state hydraulic model unit, a dynamic heat transfer model unit, and a thermal characteristic model unit. The steady-state hydraulic model unit is used to set heat exchange stations, user heating areas, and heat sources in the heating network as heating nodes, and pipes connecting all heating nodes as pipe segments. The heating nodes and pipe segments are used to construct a topology diagram of the heating network, and resistance coefficients are set in the topology diagram to construct a steady-state hydraulic model. The dynamic heat transfer model unit is used to construct a dynamic heat transfer model for heating pipes at the same location at the same time using a one-dimensional unsteady heat transfer equation. The thermal characteristic model unit is used to simplify each user heating area in the heating network into a network composed of thermal resistance and heat capacity, and defines the thermal characteristic model structure using a first-order equivalent model.

10. The operation status control system based on a heating network according to claim 8, characterized in that: The feedforward instruction module includes a control target unit and a constraint condition unit. The control target unit is used to set the control cycle of the heating network based on the time series. In each control cycle of the heating network, the control target is set to minimize the total operating cost while satisfying the predicted total heat load and the heat load of the heating nodes. The constraint condition unit is used to set constraint conditions, which include equality constraints and inequality constraints.

Images