Intelligent control method of heating system

Abstract

The invention relates to the technical field of intelligent heating control, and discloses an intelligent control method for a heating system, which comprises the following steps of: constructing a data set by acquiring hardware configuration data, heating system operation data, house static information and outdoor meteorological data of a user side, and determining a service mode of the user side according to the acquired data and user preference; for a second user side selecting the balanced heat supply mode, processing the structured data set through an AI model and a clustering algorithm to generate a first control strategy, and receiving a second control strategy obtained by performing unit-level balance processing and cell-level optimization processing on the first control strategy; for a third user side selecting the personalized metering heat consumption mode, a third control strategy is generated through a prediction model processing data set obtained based on fusion of a physical model and an AI model; according to the service mode selected by the user side, controlling the user side to execute a second control strategy or a third control strategy; accurate regulation and control of the heating system are achieved, and energy waste is avoided.

Description

Technical Field

[0001] This invention relates to the field of intelligent heating control technology, and in particular to an intelligent control method for a heating system. Background Technology

[0002] Centralized heating has advantages such as high thermal efficiency, centralized operation, and ease of unified management, and is therefore widely used in both civil and industrial sectors. However, existing centralized heating systems still have certain technical limitations in actual operation.

[0003] On the one hand, existing centralized heating systems mostly use a billing method based on building area, lacking an adjustment and billing mechanism that matches users' actual heat demand. This makes it difficult to reflect fairness and energy-saving incentives, and also fails to meet users' personalized needs for indoor thermal environment. On the other hand, in terms of system operation and control, existing heating systems typically use a constant pressure differential pump control strategy, failing to fully consider the dynamic changes in pipe network resistance caused by load variations. Under low flow demand conditions, this can easily lead to insufficient flow in unfavorable loops, resulting in poor performance for some users. To prevent this, the worst-case scenario must always be used as the control target, leading to severe oversupply and significant energy waste.

[0004] In addition, during the heat regulation process on the user side, existing heating systems generally use proportional regulating valves to continuously regulate the flow rate. However, this regulation method has problems in practical applications: when the adjustment is small, the flow rate change is easily compensated by the change in the temperature difference between the supply and return water, and the heat dissipation regulation effect is not obvious; when the adjustment is large, it is easy to affect the hydraulic balance of the multi-loop heating system, which in turn leads to a deterioration in the heating effect in some areas.

[0005] Therefore, how to ensure user comfort, reduce energy waste, and improve the system's responsiveness to users' actual heat demands while maintaining the hydraulic stability of the heating system has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] This invention provides an intelligent control method for a heating system, which solves the problem of how to ensure user comfort, reduce energy waste, and improve the system's responsiveness to users' actual heat demands while maintaining the hydraulic stability of the heating system.

[0007] To address the aforementioned technical problems, the first aspect of this invention provides an intelligent control method for a heating system, comprising:

[0008] During the heating system configuration phase, hardware configuration data and static information of the house at the user end are acquired sequentially, and the operation data of the heating system and outdoor meteorological data are collected in real time to construct a dynamic structured dataset.

[0009] The virtual indoor temperature of the user terminal is obtained by processing the dynamic structured dataset through a pre-built first artificial intelligence model.

[0010] Based on the relationship between the virtual indoor temperature and the target temperature range, the temperature adjustment direction and deviation amplitude of the user terminal are determined, so as to generate a first control strategy through a clustering algorithm;

[0011] The system receives a second control strategy obtained by performing cell-level equalization and cell-level optimization on the first control strategy, and controls the user terminal to execute the second control strategy.

[0012] A second aspect of the present invention provides a heating system, including a user-end intelligent thermostatic valve, a unit-level intelligent gateway, and a cloud management platform; wherein,

[0013] The user-side intelligent temperature control valve is configured to implement the intelligent control method as described in the first aspect;

[0014] The unit-level intelligent gateway is configured to communicate with the user-end intelligent temperature control valve and the cloud management platform, and to perform unit-level equalization processing on the first control strategy sent by the user-end intelligent temperature control valve to obtain an equalization control strategy to be sent to the cloud management platform.

[0015] The cloud management platform is configured to communicate with the user-end smart temperature control valve and the unit-level smart gateway, and to perform community-level optimization processing on the balanced control strategy to obtain a second control strategy to be sent to the corresponding user-end smart temperature control valve.

[0016] Compared with the prior art, the beneficial effects of the embodiments of the present invention are as follows:

[0017] During the heating system configuration phase, hardware configuration data and static building information are sequentially acquired from the user end, and real-time operating data of the heating system and outdoor meteorological data are collected to construct a dynamic structured dataset. A pre-built first artificial intelligence model processes the dynamic structured dataset to obtain the virtual indoor temperature of the user end. Based on the relationship between the virtual indoor temperature and the target temperature range, the temperature adjustment direction and deviation amplitude of the user end are determined, and a first control strategy is generated through a clustering algorithm. A second control strategy obtained by performing unit-level equalization processing and community-level optimization processing on the first control strategy is received, and the user end is controlled to execute the second control strategy. By combining the artificial intelligence model with the clustering algorithm, heat waste is effectively reduced, and more precise on-demand heating is achieved. Through multi-level collaborative optimization at the unit and community levels, the system dynamically adapts to changes in pipeline resistance, avoids hydraulic imbalance, and reduces unnecessary heat consumption. While ensuring the hydraulic stability of the heating system, it guarantees user comfort, reduces energy waste, and improves the system's responsiveness to users' actual heat demands. Attached Figure Description

[0018] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of an intelligent control method for a heating system provided in a certain embodiment of the present invention;

[0020] Figure 2 This is a flowchart of an intelligent control method for a heating system provided in another embodiment of the present invention;

[0021] Figure 3 This is a structural diagram of a heating system provided in a certain embodiment of the present invention;

[0022] Figure label:

[0023] Among them, 10 is the user-end intelligent temperature control valve; 20 is the unit-level intelligent gateway; 30 is the cloud management platform; and 40 is the indoor temperature control panel. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings and examples. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0025] In this invention description, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In this invention description, unless otherwise stated, "a plurality of" means two or more. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art will understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0026] In the description of this invention, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is merely for describing specific embodiments and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0027] In one embodiment, such as Figure 1 As shown, the first aspect of the present invention provides an intelligent control method for a heating system, comprising:

[0028] S1. During the heating system configuration phase, the hardware configuration data and static information of the house at the user end are acquired sequentially, and the operation data of the heating system and outdoor meteorological data are collected in real time to construct a dynamic structured dataset.

[0029] Specifically, each user terminal is equipped with a household constant-flow intelligent thermostatic valve. This valve sets and maintains a fixed flow rate in the user terminal branch and integrates supply and return water temperature sensors to collect heating system operation data. This data includes at least supply water temperature, return water temperature, valve on / off status, and cumulative valve opening time. User terminal hardware configuration data is registered and acquired once during system deployment. This data includes whether the user terminal is equipped with an indoor thermostatic panel, the communication capability between the indoor thermostatic panel and the household constant-flow intelligent thermostatic valve, and pairing information. The indoor thermostatic panel is optional and is used to collect real-time room temperature data and support user-set target temperatures. House static information is collected or entered once during system deployment. This information includes at least static parameters characterizing building thermal inertia and heat dissipation conditions, such as house area, insulation type, floor level, and orientation. Outdoor meteorological data is collected in real-time and continuously updated by outdoor meteorological monitoring equipment. This data includes at least outdoor temperature and outdoor humidity. Subsequently, the collected and acquired data were preprocessed, including data cleaning, outlier removal and time alignment. The processed heating system operation data, building static information and outdoor meteorological data were then combined according to unified fields and timestamps to form a dynamic structured dataset for subsequent thermal modeling and control strategy generation.

[0030] In one embodiment, before step S2, the method further includes:

[0031] Based on the hardware configuration data, the user terminal is first-level determined, and the user terminal that passes the first-level determination is designated as the first user terminal.

[0032] The first user terminal is evaluated based on user data. The user terminal that fails the second-level evaluation is combined with the user terminal that passes the first-level evaluation to form the second user terminal, and the user terminal that passes the second-level evaluation is used as the third user terminal.

[0033] Specifically, this invention classifies users based on their heating participation status and determines their corresponding control strategy participation mode. That is, based on the user's hardware configuration data and the user's current control behavior, it determines whether the user has entered an autonomous temperature adjustment state and determines whether the user participates in the system's balanced heating coordinated control.

[0034] During operation, the heating system by default incorporates all user terminals into the balanced heating control system and, based on the structured dataset constructed in step S1, enters the first artificial intelligence model processing link to generate a first control strategy for system balanced regulation. When the system detects that a user terminal is equipped with an indoor temperature control panel, and the user actively sets a target temperature or enables the autonomous temperature adjustment function through the indoor temperature control panel, the system determines that the user terminal has entered a personalized metering heating state with autonomous temperature adjustment.

[0035] For user terminals that enter autonomous temperature regulation mode, the system removes them from the balanced heating collaborative control and no longer participates in the generation process of the first control strategy and the subsequent third control strategy based on virtual room temperature ranking and hydraulic balance constraints. The second artificial intelligence model deployed in the household constant flow intelligent temperature control valve independently calculates the valve on / off time ratio for the next control cycle based on the target temperature set by the user and the historical operating data of the user terminal, and executes it locally on the user terminal to keep the indoor temperature stably maintained near the user-set comfort target range.

[0036] For user terminals that have not entered the autonomous temperature regulation state, the system maintains its participation in the balanced heating control system, continues to use the first control strategy generated by the first artificial intelligence model, and accepts the constraints of the third control strategy formed by unit-level balancing processing and community-level collaborative optimization in subsequent steps, so as to achieve overall system thermal balance and energy efficiency optimization.

[0037] In other words, this invention uses hardware configuration data—whether an indoor temperature control panel is installed—as the primary criterion for judgment, and designates users who pass the primary criterion as the first user and personalized heating users. Users who do not have an indoor temperature control panel installed are balanced heating users. Personalized heating users can set their target desired temperature. This invention uses user data—whether the target temperature setting function is used as the secondary criterion for judgment. The first user who uses the target temperature setting function becomes the third user (for the third user, subsequent heating costs are calculated based on the amount of heat required to reach the customized temperature, i.e., a personalized metering heating service mode user). The first user who does not choose to use the target temperature setting function and all balanced heating users are combined to form the second user (for the second user, heating costs are charged according to the traditional area, i.e., a traditional area service mode user).

[0038] Through the above methods, the present invention does not require explicit service mode division in the initial stage of the system. It can simultaneously support two control modes, namely balanced heating and personalized autonomous temperature adjustment, within the same heating system. This achieves dynamic coordination between user autonomous comfort and system-level thermal balance, directly solving the problems of lack of fairness and energy-saving incentives, and improving user satisfaction and fairness.

[0039] S2. The dynamic structured dataset is processed by a pre-built first artificial intelligence model to obtain the virtual indoor temperature of the user terminal;

[0040] In one embodiment, step S2 includes:

[0041] Dynamic features are constructed based on the dynamic structured dataset of the second user terminal;

[0042] The dynamic features are input into the first large-scale artificial intelligence model for processing to obtain the virtual indoor temperature of the second user terminal.

[0043] Specifically, this invention unifies all data in the dynamic structured dataset of the second user terminal to a 1-minute time granularity to generate continuous dynamic features that include heating system operation data, outdoor environmental data, etc.

[0044] Since the second user terminal lacks a temperature control panel and its indoor temperature cannot be directly collected, this invention estimates the actual indoor temperature of the second user terminal using an AI model: The constructed dynamic features are input into a pre-trained first large-scale artificial intelligence model (e.g., a Transformer-based time series prediction model, with 3 Transformer layers (balancing accuracy and computational cost), 4 attention heads, 64 hidden layer dimensions, and outputting the virtual indoor temperature at time t). This model has been trained on massive amounts of historical data, specifically based on the actual indoor temperature data of all second user terminals in the same community. By learning the mapping relationship between supply and return water temperature changes → valve opening and closing rhythm → outdoor temperature influence → actual indoor temperature, a multi-dimensional time series prediction model is established. This model takes dynamic features as input and the predicted virtual indoor temperature as output. An encoder-decoder architecture (such as Seq2Seq with Attention or Transformer) is employed to adapt to the complex mapping from system operation data to indoor temperature. Taking Transformer as an example: Encoder layers: 3 layers; Decoder layers: 1 layer; Attention heads: 4 heads; Hidden layer dimension: 64 dimensions. Input: The model receives a multi-dimensional time series of length k=24 (e.g., data from the past 2 hours, with a sampling period of 5 minutes). This series consists of the following features: supply water temperature time series, return water temperature time series, valve opening and closing status time series (0 for closed, 1 for open), and outdoor temperature time series. All input features are standardized (Z-score standardization) before being input into the model. The output is a scalar value: the virtual indoor temperature, i.e., the estimated room temperature at the current moment. Training Data: Supervised training is conducted using historical data from all second-user terminals within the same heating system. Training data samples must include time series data containing the aforementioned input features, and the actual indoor temperature of the corresponding second-user terminal is used as the training label. Training Objective: To optimize model parameters by minimizing the difference between the model's predicted virtual temperature and the actual temperature. The loss function used is Smooth L1 Loss, which is insensitive to outliers and improves model robustness. Training Process: The AdamW optimizer (with weight decay) is used, with an initial learning rate set to 5 × 10⁻⁶. −5A cosine annealing scheduler with a batch size of 64 was used. The trained first AI model was deployed on the user-side intelligent control valve to serve its corresponding second user. In each control cycle, it collected data from each user, batch-inferred the virtual indoor temperature of all second user terminals, and output the virtual indoor temperature of the second user terminals as well as the predicted indoor temperature for the next 30 minutes. The temperature prediction process of the first AI model is expressed by the following formula:

[0045]

[0046] In the formula, This represents the virtual indoor temperature. This is the first large-scale artificial intelligence model; These are the parameters of the first large-scale artificial intelligence model; , , , These are the time series of water supply temperature, return water temperature, valve on / off status, and outdoor temperature from time tk to time t, respectively, with a sampling granularity of 1 minute for each time. is time; k is the length of the input sequence (unit: min), preferably 1440 min, to ensure coverage of the daily cycle of thermal change.

[0047] In addition, the parameters of the first artificial intelligence model are corrected by comparing the actual indoor temperature of other second or third user terminals in the same unit with the corresponding virtual temperature every hour. To ensure that the estimation error is ≤0.5℃, this invention constructs a virtual indoor temperature using a large AI model, thus solving the problem of missing data on the second user end.

[0048] S3. Based on the relationship between the virtual indoor temperature and the target temperature range, determine the temperature adjustment direction and deviation amplitude of the user terminal, so as to generate a first control strategy through a clustering algorithm;

[0049] In one embodiment, step S3 includes:

[0050] Based on the relationship between the virtual indoor temperature of the second user terminal and the target temperature range, the temperature adjustment direction and deviation amplitude of the second user terminal are determined.

[0051] Static features are constructed based on the dynamic structured dataset of all user terminals, and the static features, along with the temperature adjustment direction and trend amplitude of the second user terminal, are input into the clustering algorithm for processing, so as to group user terminals with similar building physical characteristics and heating needs into the same cluster, and obtain the user terminal-cluster ID mapping relationship.

[0052] Based on the first historical control strategy corresponding to the third user terminal, determine the average control strategy of the third user terminal within each cluster;

[0053] Based on the user terminal-cluster ID mapping relationship, the average control strategy within the same cluster is used as the first control strategy for the second user terminal in that cluster.

[0054] Based on the relationship between the predicted virtual indoor temperature of the second user terminal and its target temperature range, this invention can determine the temperature adjustment direction and deviation amplitude of the second user terminal. The goal of clustering users using a clustering algorithm is to group users with similar physical characteristics and heating needs into the same cluster, ensuring that the heating needs of users within the cluster are consistent with the heat dissipation characteristics of the building. This lays the foundation for subsequent heat allocation by cluster and shared control strategies. Therefore, features that reflect the differences between user heating characteristics and building physics are selected as the core of clustering to avoid invalid features interfering with the clustering results.

[0055] All static information about the houses from all user terminals (such as area, floor, orientation, insulation level, building location, etc.) is used as static features and standardized. This involves converting numerical features (such as area and virtual room temperature) into the [0,1] interval (to avoid large numerical features dominating clustering, such as the area's weight suppressing the orientation). Encoding is then performed, converting categorical features (such as orientation and insulation level) into one-thermal codes (e.g., south-facing = [1,0,0,0]). Dimensionality reduction is then applied to the virtual indoor temperature data, compressing the virtual room temperature sequence into 3-5 principal components using PCA (Principal Component Analysis), retaining over 90% of the information and improving clustering efficiency, resulting in statically processed features. Subsequently, for the third user terminal, its actual first historical control strategy and actual indoor temperature are used as labels; for the second user terminal, the predicted virtual indoor temperature is used as a label, combined with the static features.

[0056] The processed static features, along with the temperature adjustment direction and trend amplitude, are input into a clustering algorithm to group users with similar building physical characteristics and heating needs into the same cluster, resulting in a user-cluster ID mapping relationship. For example, Cluster 1: {Top floor, west-facing, large window-to-wall ratio, current virtual / actual temperature is low} – identified as "high heat dissipation, heating needs to be strengthened"; Cluster 2: {Middle floor, south-facing, high insulation, current virtual / actual temperature is moderate} – identified as "low heat dissipation, heating needs to be maintained"; Cluster 3: {Bottom floor, north-facing, current virtual / actual temperature is low but predicted to rise} – identified as "moderate heat dissipation, heating needs to be reduced in advance", etc. The clustering algorithm is selected based on the number of users in the community and the characteristics of data distribution to ensure that the clustering results conform to the principle of "physical similarity". For example, the K-Means algorithm is suitable for scenarios with a large number of users (>50 households) and relatively uniform data distribution. By presetting the number of clusters K (determined by the elbow method, preferably 5, which can be adjusted according to actual conditions), it minimizes the sum of squared feature distances of users within a cluster, achieving "compact within clusters and separation between clusters". Another example is the DBSCAN algorithm (the neighborhood radius is preferably 0.3, and the minimum number of samples is preferably 5): it is suitable for scenarios with a small number of users (<50 households) and uneven data density (such as houses with extreme heat insulation / heat dissipation). It automatically identifies the number of clusters through "density accessibility", avoiding the problem of K-Means being sensitive to outliers, etc., and can be selected according to actual conditions. In addition, it should be noted that the LSTM, Transformer, K-Means, DBSCAN and other algorithms used in the above schemes are all existing algorithms. This invention has not improved them, and their specific usage process can refer to existing technologies, only the application scenarios are transferred.

[0057] Within each cluster, users belonging to the personalized metering heating service mode (third user terminals) are selected, and the average value of the first historical control strategies adopted by these users to maintain a comfortable temperature (e.g., 20-22°C) under similar weather conditions over the past week is analyzed (e.g., the average valve opening time of each third user terminal and the daily total heating curve and its average value). The average valve opening time of all third user terminals within each cluster is calculated, as well as the target temperature range when these users reach a comfortable state. Then, based on the user terminal-cluster ID mapping relationship, the average control strategy and target temperature range of each cluster are used as the control strategy and corresponding target temperature range for all area service mode users (second user terminals) within that cluster. Referring to the instruction generation process in the first control strategy, the valve opening time of the second user terminals is converted into executable valve action instructions, thus obtaining the first control strategy for each second user terminal.

[0058] This invention standardizes, encodes, and reduces the dimensionality of static features, eliminating the interference of data of different dimensions and types on the clustering results and improving the accuracy and stability of subsequent clustering algorithms. Through clustering, users with similar physical characteristics of their houses (static features) and real-time heat demand (virtual indoor temperature) are grouped together. The control strategies that have been successfully verified by users of personalized metering heating service mode are safely and reliably transferred to users of area service mode in the same cluster through the "bridge" established by clustering. This allows the second user to also enjoy the precise and comfortable control brought by data-driven approaches, realizing true on-demand heating and fair distribution.

[0059] S4. Receive the second control strategy obtained by performing unit-level balancing and community-level optimization on the first control strategy, and control the user terminal to execute the second control strategy. Specifically, the user terminal smart thermostatic valve of the second user terminal sends its corresponding first control strategy to its unit-level smart gateway for unit-level balancing. Subsequently, the gateway sends the processing result to the cloud management platform for community-level optimization, obtaining the second control strategy, which is then sent to the corresponding user terminal smart thermostatic valve via the unit-level smart gateway or directly. This allows the second user terminal to receive and execute the second control strategy obtained by performing unit-level balancing and community-level optimization on its first control strategy, thereby optimizing the energy efficiency and improving temperature comfort of the heating system. Simultaneously, the final strategy can dynamically adapt to changes in pipeline resistance, avoiding disruption of system pressure balance due to large-scale valve opening adjustments, fundamentally preventing and solving the hydraulic imbalance problem.

[0060] In one embodiment, the method further includes:

[0061] A room thermal balance physical model is constructed based on the dynamic structured dataset of the third user terminal, and the room thermal balance physical model is adaptively corrected by a pre-trained second artificial intelligence model to obtain a fusion prediction base model.

[0062] The dynamic structured dataset of the third user terminal is input into the fusion prediction base model for processing to obtain a third control strategy to control the execution of the third user terminal.

[0063] Specifically, for third-party users (i.e., users in the personalized service mode), this invention uses machine learning technology combined with physics knowledge to calculate the heat required to reach the desired temperature, and then controls the heating system based on this heat. This integrates physical models with AI, combining the deterministic laws of the physical world with the uncertainties and nonlinear relationships learned by AI from data. This not only improves the accuracy of room temperature prediction, but also achieves precise control of heat dissipation, effectively overcoming the nonlinear control problem of traditional valves when adjusting at small openings, and realizing a leap from coarse adjustment to precise control.

[0064] In one embodiment, the step of constructing a room thermal balance physical model based on the dynamic structured dataset of the third user terminal, and adaptively correcting the room thermal balance physical model through a pre-trained second artificial intelligence model to obtain a fusion prediction base model, includes:

[0065] The room thermal balance physical model is constructed based on the dynamic structured dataset of the third user terminal and the room thermal balance equation.

[0066] The dynamic structured dataset from the third user terminal is input into the second artificial intelligence big model for parameter correction, resulting in corrected room heat capacity and corrected heat transfer coefficient to optimize the room heat balance physical model, thus obtaining the fusion prediction basic model.

[0067] Specifically, this invention performs time alignment on the dynamic structured dataset from the third-party user terminal and historical baseline data including indoor temperature change curves over the past 7 days, valve opening and closing time series, and supply and return water temperature fluctuation data. This involves unifying data from different sampling frequencies to a 1-minute time granularity and calculating the indoor-outdoor temperature difference and supply-return water temperature difference as derived features. These derived features are then combined with the time-aligned heating system operation data, outdoor environmental data, and historical baseline data to construct a structured data feature set. Based on this structured data feature set and the room heat balance equation, a fundamental physical model—the room heat balance physical model—is constructed, expressed by the following formula:

[0068]

[0069] In the formula, , Let be the indoor temperatures of the i-th third user terminal at times t and t+1, respectively; To control the cycle, 30 minutes is preferred; The room's heat capacity (unit: kJ / ℃) is initially determined by the room area S. i and thermal insulation level I i The estimated result is: =k1 S i I i Thermal insulation level I i (Dimensionless, mapped to levels 1-5: Level 1 (worst, such as no insulation layer), Level 2 (simple insulation), Level 3 (ordinary insulation), Level 4 (high-quality insulation), Level 5 (ultra-low energy consumption insulation)) k1 is the empirical coefficient of heat capacity (unit: kJ / (m²·℃)), which is adjusted according to the insulation level. Level 1 corresponds to 120kJ / (m²·℃), and Level 5 corresponds to 180kJ / (m²·℃). The instantaneous heat supply (unit: kJ / min) to the i-th third user terminal at time t. The heat transfer coefficient (unit: kJ / (min·℃)) is initially determined by the orientation correction factor O. i (Dimensionless, mapping rules: North = 4, West = 3, East = 2, South = 1, reflecting stronger heat dissipation in the North / West direction) and floor correction factor F i (Dimensionless, mapping rule: bottom layer = 3, middle layer = 1, top layer = 2, reflecting stronger heat dissipation at the bottom / top layer) Estimation: =k2 F i +k4 O i k2 and k4 are empirical coefficients for floor level and orientation, respectively (unit: kJ / (m²·℃)), and their preferred values ​​are 0.8 and 1.2, respectively. This refers to the outdoor temperature.

[0070] The structured data feature set from the third user terminal is input into a pre-trained second large-scale artificial intelligence model (e.g., a neural network model based on LSTM or Transformer; taking a Transformer encoder as an example, its number of layers: 2 (balancing complexity and real-time requirements), number of attention heads: 2, hidden layer dimension: 32, feedforward network dimension: 64). The input is a multi-dimensional time series of length L (e.g., L=12, representing data from the past hour, with a sampling period of 5 minutes), which consists of the following features: indoor temperature, supply water temperature, return water temperature, valve opening status, outdoor temperature, indoor-outdoor temperature difference, and supply and return water... Temperature difference; Output: Two scalar values: corrected room heat capacity and corrected heat transfer coefficient; Training data used by the model: Supervised training using massive historical heating data of similar buildings. The training data samples must contain time series of the above input features, as well as the corresponding real room heat capacity and real heat transfer coefficient as labels. These real labels can be obtained by calibration under specific experimental conditions using high-precision instruments, or by inversion calculation from long-term series data; Training objective: To optimize the model parameters by minimizing the difference between the corrected values ​​output by the model and the real labels. The loss function used is the mean squared error loss; Training process: The Adam optimizer is used, with the initial learning rate set to 1×10⁻⁶. -4The batch size is 32, and multiple rounds of iterative training are performed until the model converges. The trained second AI model is integrated into the edge computing chip of the user-end smart thermostat valve. At the beginning of each control cycle, the thermostat valve inputs the latest time-series data into the model, and infers the corrected room heat capacity and corrected heat transfer coefficient in real time. This is used to update the fusion prediction base model, thereby generating a precise second control strategy. This model analyzes data patterns and dynamically adjusts the room heat capacity and heat transfer coefficient by minimizing the average absolute error between the predicted room temperature and the actual room temperature through gradient descent. It outputs the corrected room heat capacity and corrected heat transfer coefficient to optimize the basic physical model (i.e., replace the original room heat capacity and heat transfer coefficient with the corrected data) to obtain the fusion prediction base model.

[0071] This invention, by fusing physical and AI models, can more accurately predict room thermal dynamics, taking into account changes in key parameters such as room heat capacity and heat transfer coefficient, thereby quickly calculating the amount of heat required to stabilize the room temperature to the target desired temperature, avoiding overshoot or lag problems in traditional control methods.

[0072] In one embodiment, the step of inputting the dynamic structured dataset from the third user terminal into the fusion prediction base model for processing to obtain a third control strategy for controlling the execution of the third user terminal includes:

[0073] Based on the target demand temperature of the third user terminal and the fusion prediction basic model, the predicted heat value required for the room temperature of the third user terminal to stabilize to the target demand temperature is determined.

[0074] Real-time supply water temperature and real-time return water temperature are extracted from the dynamic structured dataset of the third user terminal to quantify the unit heat supply, and the valve opening time of the third user terminal is determined based on the heat prediction value and the unit heat supply.

[0075] The valve opening duration of the third user terminal is converted into the third control strategy to control the execution of the third user terminal.

[0076] Specifically, this invention calculates the heat required to compensate for the difference between the current indoor temperature and the target indoor temperature based on the current indoor temperature data in the dynamic structured data feature set of the third user terminal, the target temperature (user-defined target temperature) of the third user terminal, the predicted outdoor temperature (such as the weather trend for the next 30 minutes predicted using an LSTM model), and the fusion prediction base model. It is expressed by the following formula:

[0077]

[0078] In the formula, The target required temperature; To correct the room's heat capacity.

[0079] Next, heat loss prediction is performed, which involves calculating the heat loss through walls, doors, and windows over 30 minutes based on the outdoor temperature. It is expressed by the following formula:

[0080]

[0081] In the formula, To correct the heat transfer coefficient; This is the predicted outdoor temperature.

[0082] The calculated heat and heat loss By adding them together, we can obtain the total heat demand for the next control cycle, which is the predicted heat value required for the current stable room temperature and the target demand temperature at the third user terminal.

[0083] This invention converts heat demand into specific valve opening times based on real-time heating power: First, it obtains the real-time supply water temperature and return water temperature at the third user end to calculate the heat supply per unit time, i.e., the unit heat supply. (Unit: kJ / min), which is expressed by the following formula:

[0084]

[0085] In the formula, The user's water flow rate (in L / h) is initialized at 4-6 liters / square meter, i.e. =k3 S i k3 is the flow coefficient (preferably 5 L / (m²·h)), which is divided by 60 to convert to L / min; For specific heat capacity, ; The density of water is 1 kg / L; , These are the supply water temperature and the return water temperature, respectively.

[0086] The ratio of the predicted heat value to the unit heat supply is used as the valve opening time at the third user end. (Accurate to 0.1 minutes), wherein the valve opening time is no more than 30 minutes. If the calculated value exceeds 30 minutes, then 30 minutes shall be used.

[0087] Finally, the valve opening duration from the third user terminal is converted into its corresponding second control strategy, which is the executable valve action command: when the valve opening duration is 0, the command is fully closed, meaning the valve remains closed for 30 minutes; when the valve opening duration is 30, the command is fully open, meaning the valve remains open for 30 minutes; when the valve opening duration is between 0 and 30, the command is open. Minutes + Off (30 minus) ( ) minutes. Finally, the obtained second control strategy is transmitted to the local chip of its temperature control valve without waiting for the cloud response; at the same time, the instruction execution timestamp is recorded (such as "10:00-10:15 open, 10:15-10:30 close").

[0088] This invention, based on accurate heat prediction and real-time data quantification of unit heat supply, can optimize valve control strategies, avoid unnecessary heating, reduce energy waste, lower operating costs, and meet the requirements of green building and smart energy management. It utilizes AI models to correct physical parameters in real time, enabling the model to adapt to different room structural characteristics, external environmental changes, and user behavior, thus improving system robustness and personalized service levels. By combining structured data with target temperature requirements, it automatically generates control strategies through model prediction, reducing manual intervention, improving the automation level of the heating system, and enhancing the user experience.

[0089] The process of another intelligent control method for heating systems is as follows: Figure 2 As shown, this invention addresses the challenge of ensuring user comfort, reducing energy waste, and improving the system's responsiveness to actual user heat demands while maintaining stable hydraulic operation of the heating system. An intelligent control method for the heating system is designed, which determines the service mode based on hardware configuration data and user preferences. This solves the problems of a single service mode, lack of fairness, and insufficient energy-saving incentives, achieving diversified and personalized service modes and improving user satisfaction and fairness. For users with personalized metered heating service modes, a physical model is integrated with AI, combining the deterministic laws of the physical world with the uncertainties and nonlinear relationships learned by AI from data. This combination not only improves the accuracy of room temperature prediction but also enables precise control of heat dissipation, effectively overcoming the nonlinearity problem of traditional valves when adjusting small openings, and achieving a leap from coarse adjustment to precise control. For users in the area service mode, the generated control strategy will undergo unit-level equalization processing and community-level optimization processing, enabling the final strategy to dynamically adapt to changes in pipeline resistance, avoiding disruption of system pressure balance due to large opening adjustments of individual valves, and fundamentally preventing and solving hydraulic imbalance problems. Through multi-level optimization processing, the waste of operating with high flow and high pressure differential to accommodate the most unfavorable loop can be avoided, thus preventing unnecessary energy consumption.

[0090] It should be noted that although the steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order requirement for the execution of these steps, and they can be executed in other orders.

[0091] In another embodiment, such as Figure 3 As shown, a second aspect of the present invention provides a heating system, including a user-end intelligent thermostatic valve 10, a unit-level intelligent gateway 20, and a cloud management platform 30; wherein,

[0092] The user-end intelligent temperature control valve 10 is configured to implement the intelligent control method as described in the first aspect;

[0093] The unit-level intelligent gateway 20 is configured to communicate with the user-end intelligent temperature control valve 10 and the cloud management platform 30, and to perform unit-level equalization processing on the first control strategy sent by the user-end intelligent temperature control valve 10 to obtain an equalization control strategy to be sent to the cloud management platform 30.

[0094] The cloud management platform 30 is configured to communicate with the user-end smart temperature control valve 10 and the unit-level smart gateway 20, and to perform cell-level optimization processing on the balanced control strategy to obtain a second control strategy to be sent to the corresponding user-end smart temperature control valve 10.

[0095] Specifically, if heating is provided for a residential community, then each household within that community will have a user-end smart thermostatic valve 10 installed at the inlet of the user's heating pipe. This valve incorporates an AI chip, temperature sensor, and on / off controller, possessing edge computing capabilities. It is responsible for collecting key thermodynamic data and implementing the intelligent control method described in the first aspect. The heating system also includes an indoor thermostatic panel 40, which is connected to the user-end smart thermostatic valve 10 at the first user end. This invention categorizes users based on whether they have an optional indoor thermostatic panel 40 and generates corresponding control strategies to execute switching actions according to each category. The number of unit-level smart gateways 20 corresponds to the number of units within the community. Deployed in the building unit's computer room, it completes communication adaptation with the user-end temperature control valves within the unit. It is responsible for collecting data from all user-end smart temperature control valves 10 within its managed units, and has a built-in AI algorithm chip to execute the balanced heat control logic within the unit. It also communicates with the cloud management platform 30 via 4G network or other wireless means. The cloud management platform 30 is deployed at the community heating station and communicates with all unit-level smart gateways 20 and user-end smart temperature control valves 10 wirelessly to perform community-level optimization of the balanced control strategy for all units. After obtaining the third control strategy, it sends it to the corresponding user-end smart temperature control valve 10, or it can send it to the corresponding user-end smart temperature control valve 10 through the corresponding unit-level smart gateway 20.

[0096] In one embodiment, the unit-level smart gateway 20 is configured as follows:

[0097] Extract the pipeline operation data within the unit from the dynamic structured dataset corresponding to the second user terminal contained within the unit; wherein, the pipeline operation data within the unit includes the branch flow reference value, valve opening status and cumulative valve opening time of the second user terminal within the unit;

[0098] Calculate the deviation between the virtual indoor temperature of each second user terminal in the unit and the median of its target temperature range, and determine the temperature adjustment direction and deviation amplitude of each second user terminal in the unit based on the relationship between the deviation and the deviation threshold.

[0099] Based on the hydraulic balance constraints within the unit and the pipeline operation data within the unit, the direction and magnitude of each temperature adjustment are verified in multiple dimensions to generate a preliminary correction strategy to correct the first control strategy corresponding to each second user terminal, thereby obtaining a balanced control strategy.

[0100] Specifically, at the beginning of each control cycle (30 minutes), the unit-level intelligent gateway 20 synchronizes user temperature data from the user-end intelligent temperature control valve 10: the real-time temperature of all users in the unit and the target temperature range of the second user end, as well as the pipeline operation data in the unit: the branch flow reference value of the second user end contained in the unit (the preset value of the household constant flow temperature control valve, set according to the standard of 4-6 liters / square meter), the valve opening status (open / closed) and the cumulative valve opening time.

[0101] Each unit-level smart gateway 20 calculates the deviation between the virtual indoor temperature of each second user terminal within its unit and the median of its target temperature range (e.g., the median of the target temperature range [20,22]℃ is 21℃), and determines the temperature adjustment direction and deviation amplitude of each second user terminal within the unit based on its relationship with the deviation threshold: When the deviation is <-0.5℃ (temperature is too low): the initial plan is to extend the valve opening time, with the amplitude calculated according to the deviation ratio, and an adjustment of 10% corresponding to each 1℃ deviation, such as extending by 10% corresponding to a deviation of -1℃; when the deviation is >0.5℃ (temperature is too high): the initial plan is to shorten the valve opening time, with the same amplitude; no adjustment is made when the deviation is within ±0.5℃.

[0102] In one embodiment, the step of performing multi-dimensional verification of the temperature regulation direction and deviation amplitude based on the hydraulic balance constraints within the unit and the pipeline network operation data within the unit, and generating a preliminary correction strategy, includes:

[0103] Based on the valve opening duration in each of the temperature adjustment directions and deviation amplitudes and the cumulative valve opening duration at the corresponding second user terminal, a single adjustment amplitude constraint verification is performed on each of the first control strategies to obtain the first adjustment strategy.

[0104] Based on the branch flow baseline value and the valve opening duration in the first adjustment strategy, the first adjustment strategy is verified by the unit total flow fluctuation constraint to obtain the second adjustment strategy.

[0105] Based on the location of each second user terminal, the valve action staggered constraint verification of the second adjustment strategy is performed to obtain a preliminary correction strategy.

[0106] This invention verifies the direction and magnitude of temperature adjustment at the second user terminal within each unit based on intra-unit hydraulic balance constraints to ensure that adjustments do not cause hydraulic imbalance.

[0107] The single adjustment range constraint stipulates that the valve opening duration adjustment ratio for a single user must not exceed 10%. The ratio of the valve opening duration in the temperature adjustment direction and trend amplitude of the second user to its own cumulative valve opening duration is used as the valve opening duration adjustment ratio. If this ratio exceeds 10%, it indicates that the single adjustment range constraint check has not been passed, and the valve opening duration adjustment ratio for that user is forcibly limited to 10%, resulting in the first adjustment strategy. For example, if a user's valve opening duration was originally 15 minutes and is planned to be adjusted to 18 minutes, but the 20% increase does not meet the single adjustment range constraint, it is forcibly limited to 10%, i.e., adjusted to 16.5 minutes. If the single adjustment range constraint check is passed, the adjusted strategy in the temperature adjustment direction and trend amplitude is used as the first adjustment strategy. This invention avoids sudden changes in branch flow caused by large valve movements of a single user through single adjustment range constraints.

[0108] The constraint on total flow fluctuation within a unit is that the deviation between the total flow within the unit after adjustment and the total flow before adjustment must be ≤15%. This invention calculates the equivalent proportion of valve opening status in the first adjustment strategy for each second user terminal within the unit, multiplied by its branch flow baseline value, to obtain the adjusted user flow. The adjusted user flow of all second user terminals within the unit is then summed to obtain the adjusted total flow within the unit. Based on this calculation principle, the current total flow within the unit is calculated as the total flow within the unit before adjustment. The absolute value of the difference between the adjusted and the previous total flow is divided by the previous total flow. If the quotient is not greater than 15%, the first adjustment strategy passes the unit total flow fluctuation constraint verification and is used as the second adjustment strategy. Otherwise, it is sorted according to "temperature deviation priority," prioritizing adjustments for users with larger deviations (e.g., users with a deviation of -1℃ are prioritized over users with a deviation of -0.6℃), reducing the adjustment range of lower-priority users (10% adjustment for every 1℃ deviation) until the total flow fluctuation meets the standard, thus obtaining the second adjustment strategy. This invention prevents pressure imbalance in the main pipeline caused by sudden increases or decreases in the total flow of the unit through a constraint verification of the total flow fluctuation of the unit.

[0109] The valve action staggering constraint requires that the valve opening / closing actions of the second user terminal within this unit be performed at staggered times to avoid the instantaneous impact of synchronous actions on the pipeline network. The gateway groups the users to be adjusted by floor, dividing the users in each unit into three groups proportionally: low, medium, and high (low: one-third of the building from the ground floor upwards; high: one-third of the building from the top floor downwards; the rest are in the medium group). Based on the grouping results, the low-floor users are adjusted first, followed by the medium-floor users after five minutes, and then the high-floor users after ten minutes, so that adjacent groups are spaced five minutes apart, generating a staggering schedule, and thus obtaining a third adjustment strategy as a preliminary correction strategy. This invention reduces pressure fluctuations caused by the superposition of instantaneous flow rates through valve action staggering constraint verification.

[0110] This invention combines a unified control strategy generated by clustering with real-time operational data within the unit, overcoming the limitations of macro-level strategies in adapting to the dynamic changes in the micro-network. By constraining total flow fluctuations, it avoids system pressure surges caused by simultaneous valve adjustments, improving system reliability. A valve actuation staggering mechanism prevents simultaneous equipment start-up and shutdown, reducing mechanical wear and extending valve lifespan. Verification based on branch flow benchmarks ensures the system operates within its design range. A three-level constraint verification (amplitude constraint, flow constraint, and staggering constraint) effectively prevents hydraulic imbalances in the network caused by valve actuations. Through a multi-level verification mechanism, it ensures that control commands meet both user-specific needs and network operational constraints, guaranteeing the hydraulic stability of the heating system.

[0111] In one embodiment, the cloud management platform 30 is configured as follows:

[0112] The system acquires real-time thermal data of each unit and global pipeline network operation data, and determines the adjustment strategy for each unit based on the relationship between the thermal data and the global target temperature; wherein, the global pipeline network operation data includes secondary side pipeline network pressure data, secondary side water pump flow data, and the current total flow of each unit;

[0113] Based on global hydraulic balance constraints and global pipeline network operation data, the adjustment strategies of each unit are verified in multiple dimensions, and a secondary correction strategy is generated to correct the equilibrium control strategy, resulting in an optimized control strategy as the second control strategy.

[0114] Specifically, this invention coordinates the adjustment strategies of each unit through a cloud platform, avoiding overall pressure imbalance and flow distribution disorder caused by cross-unit control. Every hour, the cloud management platform summarizes the average unit temperature, temperature compliance rate (i.e., the percentage of users meeting the standard within the cluster), and target temperature range within the cluster uploaded by the smart gateways of each unit as thermal data; as well as secondary side pipeline pressure data: total pressure of the secondary side pipeline (collected by the main pipeline pressure sensor, accuracy ±0.01MPa) and inlet pressure of each unit (pressure of the unit's branch pipe); and secondary side water pump flow data: current frequency, rated flow, real-time total flow (collected by the flow meter), and current total flow of each unit (i.e., the sum of the current flow of all users within the unit; for a certain user, when their valve is open, the corresponding flow rate for that user is...). When the valve is closed, the flow rate for that user is 0. Simultaneously, the cloud management platform 30 establishes a baseline model linking "flow rate-pressure-pump frequency" based on historical data to record the normal pressure range under different operating conditions (e.g., the stable pressure range of the secondary network: 0.25~0.35MPa). Based on this data, the cloud management platform 30 calculates the deviation between each unit and the global target temperature of the community, and determines the adjustment strategy for each unit according to the relationship between each deviation and the deviation threshold. Specifically, if the deviation > 1℃ (unit is too hot): the initial plan is to lower the target temperature range within the unit cluster by 0.5℃, and simultaneously shorten the valve opening time for all users within the unit by 5%; if the deviation < -1℃ (unit is too cold): the initial plan is to raise the target temperature range within the cluster by 0.5℃, and extend the opening time by 5%; if the deviation is within ±1℃, no adjustment is made for the time being.

[0115] In one embodiment, the step of performing multi-dimensional verification of the adjustment strategies of each unit based on global hydraulic balance constraints and global pipeline network operation data to generate a secondary correction strategy includes:

[0116] The estimated total flow corresponding to the adjustment strategy of each unit is input into the baseline model for processing to obtain the predicted pressure of the secondary side pipeline network. This is then combined with the pressure data of the secondary side pipeline network to perform global differential pressure fluctuation constraint verification on the adjustment strategy of each unit, thus obtaining the initial adjustment strategy of each unit.

[0117] Based on the estimated total flow corresponding to the initial adjustment strategy of each unit and the current total flow of each unit, the initial adjustment strategy of each unit is checked for total flow change constraints to obtain the secondary adjustment strategy of each unit as the secondary correction strategy.

[0118] The global differential pressure fluctuation constraint requires that, after simulation adjustment, the deviation between the total pressure of the secondary pipeline network and the current pressure must be ≤10%, and the deviation between the inlet pressure of each unit and the unit's design pressure must be ≤10%. The estimated total flow corresponding to the adjustment strategy of each unit is input into the baseline model for processing to obtain the total pressure of the secondary pipeline network and the predicted inlet pressure of each unit. This is then compared with the corresponding pressure in the secondary pipeline network pressure data. If the deviations are both less than 10%, it indicates that the adjustment strategy of each unit has passed the global differential pressure fluctuation constraint verification, and the adjustment strategy of each unit is used as the initial adjustment strategy for each unit. Otherwise, the adjustment range of that unit is reduced; for every 10% deviation, the opening duration in the adjustment strategy is reduced by 5%, or it can be adjusted to be completed in two cycles. This yields the initial adjustment strategy for each unit.

[0119] The total flow change constraint verification requires that the deviation between the adjusted total flow and the current total flow must be ≤20%. The absolute value of the difference between the estimated total flow corresponding to the initial adjustment strategy of each unit and the current total flow of each unit is divided by the current total flow of each unit. The quotient is the deviation. If the deviation meets the total flow change constraint verification, the initial adjustment strategy of each unit is used as the secondary adjustment strategy of each unit; otherwise, the units are sorted according to the absolute value of the temperature deviation, and the units with larger deviations are given priority for adjustment (e.g., units with a deviation ΔTk of -2℃ are given priority over units with a deviation of -1.2℃). The adjustment range of low-priority units is reduced (the opening time in the adjustment strategy is reduced by 5% for every 10% deviation) until the total flow fluctuation meets the standard. The secondary adjustment strategy of each unit is obtained as a secondary correction strategy, thereby avoiding pump overload or inefficient operation.

[0120] Finally, the secondary adjustment strategies of each unit are sent to the corresponding unit-level intelligent gateway 20 or directly to the corresponding user-end intelligent temperature control valve 10 as its second control strategy. This invention prevents overpressure operation of the pipeline network by constraining differential pressure fluctuations, ensuring equipment safety, and avoids the impact of sudden flow changes on the heat source and pipeline network by constraining total flow changes.

[0121] In one embodiment, the cloud management platform 30 is further configured to:

[0122] When the average temperature of a unit deviates from the global target temperature by more than a preset threshold, the cloud management platform 30 generates a corresponding adjustment trend instruction; for units with high temperatures, it generates an adjustment suggestion to weaken the heating trend; for units with low temperatures, it generates an adjustment suggestion to strengthen the heating trend; for units with temperatures within the target range, it generates an adjustment suggestion to maintain the current heating state; the adjustment suggestion can be expressed by adjusting the target temperature range within the unit or by constraining the valve on / off time ratio, and the specific adjustment range can be dynamically determined according to the system operating status.

[0123] In one embodiment, the cloud management platform 30 is further configured to:

[0124] Based on global hydraulic balance constraints, the unit-level gateway adjustment suggestions are verified at multiple levels to avoid adverse effects of coordinated adjustment on the stability of the heating system.

[0125] The multi-level verification includes:

[0126] The pressure change of the secondary pipeline network is simulated and analyzed based on the estimated adjustment results, and the simulation results are compared with real-time pressure data to determine whether the pipeline network pressure fluctuation constraint is met after adjustment.

[0127] The magnitude of the change in total flow in the cell is evaluated based on the estimated adjustment results, and the evaluation results are compared with the range of total flow change allowed by the system to determine whether the flow stability constraint is met after adjustment.

[0128] When the adjustment result fails to meet any constraint, the cloud management platform 30 reduces the adjustment range of the corresponding unit or performs corrections in stages until the global hydraulic constraint requirements are met, thereby generating a corrected unit-level adjustment strategy.

[0129] This invention overcomes the limitations of local perspectives imposed by unit-level gateways, coordinating and optimizing the entire heating network. It constructs a complete system operational status awareness by acquiring real-time thermal data from each unit and global network operation data. Employing a three-level global constraint verification (pressure difference fluctuation constraint and total flow change constraint), it effectively prevents network pressure surges and hydraulic imbalances caused by simultaneous adjustments from multiple units. By predicting pressure changes through a baseline model, it identifies potential system stability risks in advance. For the control strategy of balanced heating mode, this invention ensures accurate temperature for individual users through user-end autonomous generation, resolves local temperature differences through unit-level fine-tuning, and avoids cross-unit imbalances and network problems through community-level coordination. These three elements form a closed loop of "local generation → local correction → global optimization," guaranteeing both real-time control (user-end local response) and overall balance (hierarchical collaboration). Furthermore, through on / off time control and hydraulic constraints, it fundamentally solves the nonlinearity and imbalance problems of traditional proportional regulation.

[0130] In summary, this invention relates to the field of intelligent heating control technology and discloses an intelligent control method for a heating system. The method involves acquiring hardware configuration data, heating system operation data, building static information, and outdoor meteorological data from the user end to construct a dataset. Based on the acquired data and user preferences, a service mode for the user end is determined. For a second user end selecting a balanced heating mode, an AI model and clustering algorithm are used to process the structured dataset to generate a first control strategy. A second control strategy is obtained by performing unit-level balancing and community-level optimization on the first control strategy. For a third user end selecting a personalized heating mode, a prediction model based on the fusion of a physical model and an AI model is used to process the dataset to generate a third control strategy. Depending on the service mode selected by the user end, the system controls the execution of either the second or third control strategy. This achieves precise regulation of the heating system and avoids energy waste.

[0131] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on its differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0132] The above-described embodiments are merely preferred embodiments of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention. Therefore, the scope of protection of this invention should be determined by the scope of the claims.

Claims

1. An intelligent control method for a heating system, characterized in that, include: During the heating system configuration phase, hardware configuration data and static information of the house at the user end are acquired sequentially, and the operation data of the heating system and outdoor meteorological data are collected in real time to construct a dynamic structured dataset. The virtual indoor temperature of the user terminal is obtained by processing the dynamic structured dataset through a pre-built first artificial intelligence model. Based on the relationship between the virtual indoor temperature and the target temperature range, the temperature adjustment direction and deviation amplitude of the user terminal are determined, so as to generate a first control strategy through a clustering algorithm; The system receives a second control strategy obtained by performing cell-level equalization and cell-level optimization on the first control strategy, and controls the user terminal to execute the second control strategy.

2. The intelligent control method for a heating system according to claim 1, characterized in that, Before constructing the room thermal balance physical model using the dynamic structured dataset, the following steps are included: Based on the hardware configuration data, the user terminal is first-level determined, and the user terminal that passes the first-level determination is designated as the first user terminal. The first user terminal is evaluated based on user data. The user terminal that fails the second-level evaluation is combined with the user terminal that passes the first-level evaluation to form the second user terminal, and the user terminal that passes the second-level evaluation is used as the third user terminal.

3. The intelligent control method for a heating system according to claim 2, characterized in that, The process of processing the dynamic structured dataset using a pre-built first artificial intelligence model to obtain the virtual indoor temperature on the user end includes: Dynamic features are constructed based on the dynamic structured dataset of the second user terminal; The dynamic features are input into the first large-scale artificial intelligence model for processing to obtain the virtual indoor temperature of the second user terminal.

4. The intelligent control method for a heating system according to claim 3, characterized in that, The step of determining the temperature adjustment direction and deviation amplitude of the user terminal based on the relationship between the virtual indoor temperature and the target temperature range, and generating a first control strategy through a clustering algorithm, includes: Based on the relationship between the virtual indoor temperature of the second user terminal and the target temperature range, the temperature adjustment direction and deviation amplitude of the second user terminal are determined. Static features are constructed based on the dynamic structured dataset of all user terminals, and the static features, along with the temperature adjustment direction and trend amplitude of the second user terminal, are input into the clustering algorithm for processing, so as to group user terminals with similar building physical characteristics and heating needs into the same cluster, and obtain the user terminal-cluster ID mapping relationship. Based on the first historical control strategy corresponding to the third user terminal, determine the average control strategy of the third user terminal within each cluster; Based on the user terminal-cluster ID mapping relationship, the average control strategy within the same cluster is used as the first control strategy for the second user terminal in that cluster.

5. The intelligent control method for a heating system according to claim 2, characterized in that, The method further includes: A room thermal balance physical model is constructed based on the dynamic structured dataset of the third user terminal, and the room thermal balance physical model is adaptively corrected by a pre-trained second artificial intelligence model to obtain a fusion prediction base model. The dynamic structured dataset of the third user terminal is input into the fusion prediction base model for processing to obtain a third control strategy to control the execution of the third user terminal.

6. The intelligent control method for a heating system according to claim 5, characterized in that, The room thermal balance physical model is constructed based on the dynamic structured dataset of the third user terminal, and the room thermal balance physical model is adaptively corrected by a pre-trained second artificial intelligence model to obtain a fusion prediction base model, including: The room thermal balance physical model is constructed based on the dynamic structured dataset of the third user terminal and the room thermal balance equation. The dynamic structured dataset from the third user terminal is input into the second artificial intelligence big model for parameter correction, resulting in corrected room heat capacity and corrected heat transfer coefficient to optimize the room heat balance physical model, thus obtaining the fusion prediction basic model.

7. The intelligent control method for a heating system according to claim 5, characterized in that, The step of inputting the dynamic structured dataset from the third user terminal into the fusion prediction base model for processing to obtain a third control strategy to control the execution of the third user terminal includes: Based on the target demand temperature of the third user terminal and the fusion prediction basic model, the predicted heat value required for the room temperature of the third user terminal to stabilize to the target demand temperature is determined. Real-time supply water temperature and real-time return water temperature are extracted from the dynamic structured dataset of the third user terminal to quantify the unit heat supply, and the valve opening time of the third user terminal is determined based on the heat prediction value and the unit heat supply. The valve opening duration of the third user terminal is converted into the third control strategy to control the execution of the third user terminal.

8. A heating system, characterized in that, This includes user-end intelligent temperature control valves, unit-level intelligent gateways, and cloud management platforms; among which, The user-end intelligent temperature control valve is configured to implement the intelligent control method as described in any one of claims 1-7; The unit-level intelligent gateway is configured to communicate with the user-end intelligent temperature control valve and the cloud management platform, and to perform unit-level equalization processing on the first control strategy sent by the user-end intelligent temperature control valve to obtain an equalization control strategy to be sent to the cloud management platform. The cloud management platform is configured to communicate with the user-end smart temperature control valve and the unit-level smart gateway, and to perform community-level optimization processing on the balanced control strategy to obtain a second control strategy to be sent to the corresponding user-end smart temperature control valve.

9. A heating system according to claim 8, characterized in that, The unit-level intelligent gateway is configured as follows: Extract the pipeline operation data within the unit from the dynamic structured dataset corresponding to the second user terminal contained within the unit; Calculate the deviation between the virtual indoor temperature of each second user terminal in the unit and the median of its target temperature range, and determine the temperature adjustment direction and deviation amplitude of each second user terminal in the unit based on the relationship between the deviation and the deviation threshold. Based on the hydraulic balance constraints within the unit and the pipeline network operation data within the unit, the temperature adjustment direction and deviation amplitude are verified in multiple dimensions to generate a preliminary correction strategy to correct the second control strategy corresponding to each second user terminal, thereby obtaining a balanced control strategy.

10. A heating system according to claim 8, characterized in that, The cloud management platform is configured as follows: Real-time acquisition of thermal data from each unit and global pipeline operation data, and determination of adjustment strategies for each unit based on the relationship between the thermal data and the global target temperature; Based on global hydraulic balance constraints and global pipeline network operation data, the adjustment strategies of each unit are verified in multiple dimensions, and a secondary correction strategy is generated to correct the equilibrium control strategy, resulting in an optimized control strategy as the second control strategy. The heating system also includes an indoor temperature control panel, which is connected to the user-end intelligent temperature control valve of the first user terminal.

Images