Method and system for establishing online heating demand network

Abstract

The present invention provides a method and system for establishing an online heating demand network, applicable to the field of data processing technology. The method comprises: obtaining user operation data for indoor radiators, indoor temperature, and indoor radiator performance parameters; constructing a user dynamic demand response function based on the user operation data, indoor temperature, and indoor radiator performance parameters; obtaining a preset optimization model; and constructing an indoor heating demand network based on the user dynamic demand response function and the preset optimization model, thereby implementing indoor heating based on instructions output by the indoor heating demand network. Through multi-source data fusion, dynamic model construction, and intelligent network optimization, the present invention achieves precise heating demand response, intelligent control strategy (real-time balancing of multiple objectives), and flexible system architecture (supporting full-scenario adaptation), thereby significantly improving user experience and energy efficiency, and enhancing the efficiency and accuracy of indoor heating demand.

Description

Technical Field

[0001] The present invention relates to data processing technology, and in particular to a method and system for establishing an online heating demand network. Background Art

[0002] As a major energy consumer (accounting for 30%-40% of global building energy consumption), the intelligent and energy-saving transformation of building heating systems has become a key demand.

[0003] Currently, there is no online heating demand network that can meet users’ heating needs, resulting in poor user experience and low efficiency and accuracy of indoor heating demand. Summary of the Invention

[0004] Based on the above problems, the present invention is proposed to provide a method and system for establishing an online heating demand network to overcome the above problems or at least partially solve the above problems.

[0005] According to one aspect of the present invention, a method for establishing an online heating demand network is provided, comprising the following steps:

[0006] Obtain user operation data on indoor radiators, indoor temperature, and performance parameters of indoor radiators;

[0007] A user dynamic demand response function is constructed based on the user's operation data on the indoor radiator, the indoor temperature and the performance parameters of the indoor radiator;

[0008] Get the preset optimization model;

[0009] An indoor heating demand network is constructed based on the user's dynamic demand response function and a preset optimization model, so as to realize indoor heating based on the instructions output by the indoor heating demand network.

[0010] Optionally, a user dynamic demand response function is constructed based on the user's operation data on the indoor radiator, the indoor temperature, and the performance parameters of the indoor radiator, including:

[0011] Based on the user's operation data on the indoor radiator, the user's operation frequency on the indoor radiator in each time period is calculated to obtain a user operation probability matrix;

[0012] Based on the indoor temperature and the performance parameters of the indoor radiator, a temperature-performance probability matrix is ​​constructed;

[0013] Obtain the target switching probability matrix of indoor radiators based on indoor temperature and reinforcement learning algorithm;

[0014] Based on the user operation probability matrix, temperature-performance probability matrix and indoor radiator target switching probability matrix, a user dynamic demand response function is constructed.

[0015] Optionally, the user's operation frequency for the indoor radiator in each time period is calculated based on the user's operation data for the indoor radiator to obtain a user operation probability matrix, including:

[0016] Processing the user's operation data on the indoor radiator to obtain the number of operations of the user on the indoor radiator in each time period and the total number of operations of the user on the indoor radiator in all time periods;

[0017] The user's operation frequency for the indoor radiator in each time period is obtained by dividing the number of times the user operates the indoor radiator in each time period by the total number of times the user operates the indoor radiator in all time periods.

[0018] The user's operation frequency of the indoor radiator in each time period is normalized to obtain the user operation probability matrix.

[0019] Optionally, based on the indoor temperature and the performance parameters of the indoor radiator, a temperature-performance probability matrix is ​​constructed, including:

[0020] Obtaining performance attenuation data of the indoor radiator from the performance parameters of the indoor radiator;

[0021] Get the indoor temperature at each time period;

[0022] Based on the indoor temperature and indoor radiator performance attenuation data of each time period, a temperature-performance probability matrix is ​​constructed.

[0023] Optionally, obtaining a target switching probability matrix of the indoor radiator based on the indoor temperature and a reinforcement learning algorithm includes:

[0024] The initial switching probability matrix of indoor radiators is generated through interactive iterative training of reinforcement learning algorithm and indoor temperature.

[0025] Based on the temporal difference error and experience replay strategy, the initial switching probability matrix of the indoor radiator is dynamically adjusted to obtain the intermediate switching probability matrix of the indoor radiator.

[0026] The validity of the intermediate switching probability matrix of the indoor radiator is verified through simulation. If the intermediate switching probability matrix of the indoor radiator meets the first preset condition, the intermediate switching probability matrix of the indoor radiator is used as the target switching probability matrix of the indoor radiator.

[0027] Optionally, based on the user operation probability matrix, the temperature-performance probability matrix, and the target switching probability matrix of the indoor radiator, a user dynamic demand response function is constructed, including:

[0028] Feature extraction is performed on the user operation probability matrix, the temperature-performance probability matrix, and the indoor radiator target switching probability matrix to obtain the eigenvectors of the user operation probability matrix, the temperature-performance probability matrix, and the indoor radiator target switching probability matrix;

[0029] Based on the eigenvectors of the user operation probability matrix, the eigenvectors of the temperature-performance probability matrix, and the eigenvectors of the indoor radiator target switching probability matrix, a user dynamic demand response function is constructed.

[0030] Optionally, an indoor heating demand network is constructed based on a user dynamic demand response function and a preset optimization model, so as to realize indoor heating based on instructions output by the indoor heating demand network, including:

[0031] Obtain indoor temperature comfort, indoor radiator energy consumption cost, and indoor radiator operating life;

[0032] Taking the user's dynamic demand response function as a constraint condition, and combining the indoor temperature comfort, the energy consumption cost of the indoor radiator, and the operating life of the indoor radiator, a multi-objective optimization model corresponding to the preset optimization model is established;

[0033] Each heating node in the multi-objective optimization model corresponding to the preset optimization model is abstracted into a directed graph node with weights, and an indoor heating demand network is output to realize indoor heating based on the instructions output by the indoor heating demand network.

[0034] Optionally, a multi-objective optimization model corresponding to the preset optimization model is established with the user dynamic demand response function as a constraint condition and combined with indoor temperature comfort, energy consumption cost of the indoor radiator, and operating life of the indoor radiator, including:

[0035] Input the indoor temperature comfort, the energy consumption cost of the indoor radiator, and the operating life of the indoor radiator into the user dynamic demand response function, and output the user dynamic demand response result;

[0036] If the user's dynamic demand response result meets the second preset condition, a multi-objective optimization model corresponding to the preset optimization model is established.

[0037] Optionally, each heating node in the multi-objective optimization model corresponding to the preset optimization model is abstracted into a directed graph node with a weight, and an indoor heating demand network is output to realize indoor heating based on the instructions output by the indoor heating demand network, including:

[0038] Abstracting each heating node in the multi-objective optimization model corresponding to the preset optimization model into a directed graph node with a weight, and obtaining a node set of the directed graph nodes;

[0039] Obtaining a transmission path of each heating node in a node set of directed graph nodes, and calculating the strength of the transmission path of each heating node to establish a strength set;

[0040] Obtaining an optimal transmission path from the transmission paths of each heating node in the node set of the directed graph node, and obtaining the intensity corresponding to the optimal transmission path from the intensity set as the user demand intensity;

[0041] An indoor heating demand network is established based on the optimal transmission path and the intensity corresponding to the optimal transmission path as the user demand intensity, so as to realize indoor heating based on the instructions output by the indoor heating demand network.

[0042] According to another aspect of the present invention, there is provided a system for establishing an online heating demand network, comprising:

[0043] The first acquisition module is used to obtain user operation data on the indoor radiator, indoor temperature and performance parameters of the indoor radiator;

[0044] A function building module, for building a user dynamic demand response function based on user operation data for the indoor radiator, indoor temperature and performance parameters of the indoor radiator;

[0045] A second acquisition module is used to acquire a preset optimization model;

[0046] The network establishment module is used to build an indoor heating demand network based on the user's dynamic demand response function and a preset optimization model, so as to realize indoor heating based on the instructions output by the indoor heating demand network.

[0047] According to the present invention's solution, the user's operating data for the indoor radiator, indoor temperature, and indoor radiator performance parameters are first obtained. A dynamic user demand response function is then constructed based on the user's operating data, indoor temperature, and indoor radiator performance parameters. A preset optimization model is then obtained, and an indoor heating demand network is constructed based on the user's dynamic demand response function and the preset optimization model. Indoor heating is then implemented based on instructions output by the indoor heating demand network. Through multi-source data fusion, dynamic model construction, and intelligent network optimization, the present invention achieves precise heating demand response, intelligent control strategies (real-time balancing of multiple objectives), and flexible system architecture (supporting full-scenario adaptation), significantly improving user experience and energy efficiency, and enhancing the efficiency and accuracy of indoor heating demand. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A flow chart of a method for establishing an online heating demand network according to an embodiment of the present invention is shown;

[0049] Figure 2A structural block diagram of a system for establishing an online heating demand network according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0050] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0051] To solve the problems in the prior art, the inventors propose the solution of the present invention. One embodiment of the present invention provides a patent document management method and device based on artificial intelligence.

[0052] like Figure 1 As shown, the present embodiment proposes a method for establishing an online heating demand network, comprising the following steps:

[0053] Step S101: obtaining user operation data on an indoor radiator, indoor temperature, and performance parameters of the indoor radiator.

[0054] In the initial stage of building an indoor heating demand network, it is necessary to systematically collect three types of core data to provide basic input for the subsequent construction of user dynamic demand response functions and optimization models. The specific implementation process is as follows:

[0055] 1. Targeted Collection of User Operation Data

[0056] First, we focus on the interaction behavior data between users and radiators. By deploying sensors on the radiator control terminal (such as button sensors, touch screen interaction modules) or the background of the smart home system, we can record in real time the user's operation actions on the radiator at different time periods, including power on and off time, temperature adjustment range, and operation mode switching (such as energy-saving mode / comfort mode). On this basis, we divide the original operation records into granularity of time dimensions, for example, taking 15 minutes as a time period, and count the specific types and frequencies of user operations in each time period. To ensure data integrity, we need to clean abnormal data (such as duplicate records and data that has not been saved due to timeout), and form a standardized user operation data set through timestamp calibration and operation logic verification.

[0057] 2. Real-time monitoring and calibration of indoor temperature

[0058] Simultaneously conduct dynamic monitoring of the indoor ambient temperature. Deploy high-precision temperature sensors in key indoor areas (such as living rooms and bedrooms) and collect temperature data at a fixed frequency (e.g., once per minute) throughout the day. To account for the uneven distribution of the temperature field, perform a spatially weighted average of the multi-sensor data. For example, weight coefficients are set based on the room area and sensor location to calculate a representative composite indoor temperature value. To ensure data reliability, regularly calibrate the sensors (e.g., monthly) and compare them with a standard thermometer to correct for device drift errors. A time series archive of temperature data is also established to clearly record temperature fluctuations over each period.

[0059] 3. Full-dimensional acquisition of radiator performance parameters

[0060] Finally, complete the collection of the radiator's inherent performance parameters and attenuation characteristic data. First, extract basic performance parameters from the equipment's factory technical documentation, including rated heat dissipation, initial value of heat conduction efficiency, fluid circulation resistance coefficient, material temperature resistance threshold, etc. For performance attenuation data, obtain the attenuation coefficient that changes with usage time through equipment operation and maintenance records or built-in monitoring modules, including:

[0061] The thermal conductivity attenuation coefficient reflects the decline in the thermal conductivity of heat dissipation materials after long-term use and is calculated by comparing actual heat dissipation with the rated heat dissipation. The fluid circulation attenuation coefficient reflects the reduction in circulation efficiency caused by scaling on the inner wall of the pipe or aging of the pump body and can be indirectly assessed by monitoring the flow difference between the supply and return water. The material structure attenuation coefficient records changes in structural stability caused by factors such as corrosion and deformation and is quantified in combination with the degree of material damage reported in regular maintenance reports. These attenuation coefficients are serialized along the timeline to form a performance attenuation curve corresponding to the length of use, providing a basis for subsequent analysis of the coupled relationship between temperature and performance.

[0062] After acquiring the above data, data collaboration and preprocessing are required. Specifically, after the three types of data are collected, they must be time-aligned and formatted uniformly. Using the natural day as the primary cycle, user operation data, temperature data, and performance parameters are grouped into the same time period (e.g., one hour as a data unit) and associated to ensure data dimension consistency during subsequent analysis. Data standardization (e.g., unit and dimension normalization) creates a structured dataset suitable for model input, laying the foundation for constructing the user dynamic demand response function in step S102.

[0063] Step S102: constructing a user dynamic demand response function based on the user's operation data on the indoor radiator, the indoor temperature, and the performance parameters of the indoor radiator.

[0064] Optionally, a user dynamic demand response function is constructed based on the user's operation data on the indoor radiator, the indoor temperature and the performance parameters of the indoor radiator, including: calculating the user's operation frequency on the indoor radiator in each time period based on the user's operation data on the indoor radiator to obtain a user operation probability matrix; constructing a temperature-performance probability matrix based on the indoor temperature and the performance parameters of the indoor radiator; obtaining a target switching probability matrix of the indoor radiator based on the indoor temperature and the reinforcement learning algorithm; and constructing a user dynamic demand response function based on the user operation probability matrix, the temperature-performance probability matrix and the target switching probability matrix of the indoor radiator.

[0065] Optionally, the user's operation frequency on the indoor radiator in each time period is calculated based on the user's operation data on the indoor radiator to obtain a user operation probability matrix, including: processing the user's operation data on the indoor radiator to obtain the number of operations of the user on the indoor radiator in each time period and the total number of operations of the user on the indoor radiator in all time periods; dividing the number of operations of the user on the indoor radiator in each time period by the total number of operations of the user on the indoor radiator in all time periods to obtain the user's operation frequency on the indoor radiator in each time period; normalizing the user's operation frequency on the indoor radiator in each time period to obtain the user operation probability matrix.

[0066] For example, we first conduct an in-depth analysis of user operation data for indoor radiators. By analyzing the time series of operation data, we can clearly identify the user's operation behavior on the radiator at different time periods. Specifically, we can count the number of times a user operated the indoor radiator in each time period, and also calculate the total number of times a user operated the indoor radiator across all time periods.

[0067] Next, the number of times the user operated the indoor radiator in each time period is divided by the total number of operations to obtain the user's operation frequency for the indoor radiator in each time period. This operation frequency reflects the relative frequency of the user's operation of the radiator in each specific time period.

[0068] To make the operation frequencies across different time periods comparable and useful for further analysis, these frequencies need to be normalized. Normalization converts the operation frequencies into probability values, ensuring their sum equals 1. This results in a user operation probability matrix. This matrix, with time periods as rows and operation types as columns, clearly demonstrates the probability of users selecting various operations during different time periods.

[0069] This embodiment of the application quantifies user behavior and establishes a data-driven decision-making foundation. First, by analyzing the time series of user operation data, the number of operations per time period and the total number of operations are counted, converting user operation behavior on indoor radiators into quantifiable numerical indicators. This quantification process avoids subjective judgment and enables subsequent analysis to be based on objective data. For example, by identifying specific patterns such as "users adjust radiators most frequently between 8:00 PM and 10:00 PM," this provides precise user behavior characteristics for system optimization, rather than relying on vague empirical judgments.

[0070] Furthermore, it reflects operational tendencies and captures time-of-day differences in user needs: By calculating the frequency of operations in each time period (number of operations / total number of operations), it can intuitively reflect user preferences for radiator operation at different times. For example, the frequency of operations on weekday mornings may be lower than in the evenings, while the frequency of operations on weekends may be higher than on weekdays. This difference reflects the temporal changes in user needs (such as the impact of changes in work and rest schedules and ambient temperature on operations), providing a key basis for dynamic system adjustments. Compared to simply counting the total number of operations, calculating the frequency of operations can better reflect the relative importance of time periods, helping the system prioritize responses to needs during high-frequency operation periods.

[0071] Furthermore, normalization is performed to form a standardized probability model: the operation frequencies are normalized into a probability matrix (summing to 1). This gives the data the mathematical properties of a probability distribution, facilitating integration and analysis with other probabilistic models (such as the temperature-performance probability matrix and the radiator switching probability matrix). For example, the user operation probability matrix can be used as the probabilistic input for "user-initiated adjustment behavior" and combined with probabilistic models such as "radiator operating state switching" and "the impact of temperature changes on performance" to construct a user dynamic demand response function. This normalization eliminates the influence of data dimensionality, improving the compatibility of data of different dimensions and the universality of the model.

[0072] Furthermore, it supports intelligent control and improves the accuracy of system optimization: the user operation probability matrix is ​​one of the core foundational data for the subsequent construction of intelligent heating systems. Using this matrix, the system can predict the likelihood that users will actively adjust radiators during specific periods (for example, users are more likely to reduce heating power when the room temperature is too high). It then adjusts heating strategies in advance based on factors such as temperature fluctuations and radiator operating status. For example, during periods of high user activity (such as evenings after work), the system can optimize radiator switching logic in advance, ensuring user comfort while reducing energy consumption. This "data-driven + model-predictive" approach avoids the rigidity of traditional fixed strategies and makes the heating system more tailored to actual user needs.

[0073] Thirdly, the system enhances adaptability, balancing energy conservation and user experience. By capturing user operational patterns, the system strikes a balance between proactive user intervention and automatic control. During periods of high user activity, the system prioritizes responding to user adjustments, avoiding user experience conflicts caused by over-automation. During periods of low user activity, the system autonomously optimizes based on historical probability models and environmental data to reduce inefficient energy consumption. For example, if a user tends to lower heating power before bedtime, the system can identify this pattern through a probability matrix and automatically adjust the system in advance, reducing manual user intervention while achieving energy conservation goals. This two-way adaptive mechanism enhances the system's user-friendliness and operational efficiency.

[0074] In summary, constructing a user operation probability matrix transforms user behavior into a computable and integrable core data asset. Through the logic of "data quantification → pattern extraction → model standardization," it provides key input for subsequent reinforcement learning training, dynamic response function construction, and heating network optimization. Ultimately, it achieves a closed-loop optimization process: "accurately capturing user needs → intelligently adjusting system operation → balancing comfort and energy savings." This provides a crucial technical foundation for the transition of intelligent heating systems from an "experience-driven" to a "data-driven" approach.

[0075] Optionally, a temperature-performance probability matrix is ​​constructed based on the indoor temperature and the performance parameters of the indoor radiator, including: obtaining the performance attenuation data of the indoor radiator from the performance parameters of the indoor radiator; obtaining the indoor temperature in each time period; and constructing the temperature-performance probability matrix based on the indoor temperature in each time period and the performance attenuation data of the indoor radiator.

[0076] The performance attenuation data of the indoor radiator include but are not limited to the heat conduction attenuation coefficient, the fluid circulation attenuation coefficient and the material structure attenuation coefficient. The above attenuation coefficients all show corresponding performance attenuation as time goes by.

[0077] In one embodiment, based on the indoor temperature of each time period and the performance attenuation data of the indoor radiator, a temperature-performance probability matrix is ​​constructed, including: obtaining the indoor temperature, heat conduction attenuation coefficient, fluid circulation attenuation coefficient and material structure attenuation coefficient of each time period; searching for the probability matching the indoor temperature from the first probability table corresponding to the heat conduction attenuation coefficient based on the first comparison table to obtain the thermal conductivity attenuation rate of the heat conduction attenuation coefficient; searching for the probability corresponding to the heat conduction attenuation coefficient and the indoor temperature from the second probability table corresponding to the fluid circulation attenuation coefficient based on the second comparison table to obtain the fluid friction loss rate of the fluid circulation attenuation coefficient; searching for the probability corresponding to the heat conduction attenuation coefficient, indoor temperature and fluid circulation attenuation coefficient from the third probability table corresponding to the material structure attenuation coefficient based on the third comparison table to obtain the material loss rate of the material structure attenuation coefficient; dividing the indoor temperature of each time period into the first time period, the heat conduction attenuation coefficient and the indoor temperature; and obtaining the material loss rate of the material structure attenuation coefficient. The first time period, the second time period, and the third time period are selected; based on the indoor temperature of the first time period, a preset number of thermal conductivity attenuation rates are selected from the thermal conductivity attenuation rates of the heat conduction attenuation coefficient as the second row data of the temperature-performance probability matrix; based on the indoor temperature of the second time period, a preset number of fluid friction loss rates are selected from the fluid friction loss rate of the fluid circulation attenuation coefficient as the third row data of the temperature-performance probability matrix; based on the indoor temperature of the third time period, a preset number of material loss rates are selected from the material loss rate of the material structure attenuation coefficient as the fourth row data of the temperature-performance probability matrix; the average value of each column of the matrix formed by the second row data, the third row data, and the fourth row data is used as the first data of each column, and the first row data is formed by the first data of each column; the temperature-performance probability matrix is ​​composed of the first row data, the second row data, the third row data, and the fourth row data.

[0078] For example, in a smart indoor heating system, constructing a temperature-performance probability matrix is ​​a key step in connecting indoor environmental parameters with radiator performance degradation. This matrix quantifies the degradation probability of a radiator's core performance indicators under different temperature conditions, providing physical characteristics for subsequent user demand response analysis. The following describes the construction process in detail, following a logical sequence:

[0079] 1. Basic Data Preparation: Performance Degradation and Temperature Data Acquisition

[0080] First, extract performance degradation data from the indoor radiator's historical operation archive or equipment monitoring system, including:

[0081] Thermal conductivity attenuation coefficient: reflects the decrease in thermal conductivity of heat dissipation materials (such as metal fins) over time, quantified by the ratio of actual heat dissipation to rated heat dissipation;

[0082] Fluid circulation attenuation coefficient: reflects the increase in circulation resistance of the fluid (such as water or thermal oil) in the pipeline due to impurity deposition and pump aging, and is evaluated by the ratio of the inlet and outlet flow rate difference to the rated flow rate;

[0083] Material structure attenuation coefficient: records the structural loss of the radiator body (such as solder joints and pipe walls) due to corrosion and deformation, and converts the damage degree score through regular maintenance into a numerical indicator.

[0084] The indoor temperature of each time period monitored in step S101 is synchronously acquired to form a temperature sequence aligned with the timestamp of the performance degradation data (eg, with 1 hour as a time period, covering 24 hours a day).

[0085] 2. Probability Mapping: Calculation of Decay Rate Based on Lookup Table

[0086] To establish a quantitative relationship between temperature and performance degradation, a pre-established three-tier comparison table system is required:

[0087] The first layer: Thermal Conductivity Attenuation Rate Mapping uses the first comparison table (thermal conductivity attenuation coefficient-temperature probability table) to input the current time period's indoor temperature and thermal conductivity attenuation coefficient to find the corresponding thermal conductivity attenuation rate. This table is based on experimental data from material thermal properties. For example, when the indoor temperature is 20°C and the thermal conductivity attenuation coefficient is 0.85, the table shows a thermal conductivity attenuation rate of 12%, indicating a 12% decrease in actual thermal efficiency compared to the initial value.

[0088] The second layer: Fluid circulation loss rate mapping uses a second comparison table (fluid circulation attenuation coefficient - heat conduction attenuation coefficient - temperature probability table) to calculate the fluid friction loss rate based on the first layer's results, combining the fluid circulation attenuation coefficient with the indoor temperature. This table integrates fluid dynamics models with engineering experience. For example, when the heat conduction attenuation rate is 12%, the indoor temperature is 20°C, and the fluid circulation attenuation coefficient is 0.90, the table determines that the fluid friction loss rate is 8%, indicating that the circulation pump needs to consume 8% more energy to overcome resistance.

[0089] The third layer: Material structure loss rate mapping. The third comparison table (material structure attenuation coefficient - thermal conduction attenuation coefficient - fluid circulation attenuation coefficient - temperature probability table) combines the first three parameters to determine the material loss rate. This table is based on material fatigue theory and life prediction models. For example, when the thermal conductivity attenuation rate is 12%, the fluid friction loss rate is 8%, the indoor temperature is 20°C, and the material structure attenuation coefficient is 0.95, the table shows a material loss rate of 5%, indicating a 5% decrease in structural stability compared to the initial state.

[0090] 3. Time period division and feature data extraction

[0091] To simplify the matrix dimensions and highlight the temperature variation patterns, the entire day is divided into three representative time periods based on the temperature fluctuation characteristics:

[0092] The first time period is the low-temperature period (e.g., 10:00 PM to 6:00 AM the next day), which corresponds to the high-load operation of the radiator.

[0093] The second time period: normal temperature period (such as 6:00-10:00 in the morning and 16:00-22:00 in the afternoon), corresponding to daily comfortable operation scenarios;

[0094] The third time period: high temperature period (such as 10:00-16:00 noon), corresponding to the radiator low load or standby scenario (if natural heating exists).

[0095] For each time period, select a preset number of feature values ​​from the corresponding decay rate sequence (for example, take the decay rates of 4 key time points in each time period):

[0096] From the thermal conductivity decay rate corresponding to the temperature series in the first time period, select four typical values ​​as the second row of the matrix data (reflecting the change in thermal conductivity performance under low temperature environment);

[0097] From the fluid friction loss rate corresponding to the temperature series of the second time period, select four typical values ​​as the third row of matrix data (reflecting the changes in fluid circulation performance under normal temperature environment);

[0098] From the material loss rate corresponding to the temperature series of the third time period, four typical values ​​are selected as the fourth row data of the matrix (reflecting the change in structural stability under high temperature environment).

[0099] 4. Matrix Synthesis: Building a Multi-Level Probabilistic Model

[0100] To create a comprehensive matrix structure, we first calculated the average of each column in the second, third, and fourth rows of data. This average is used as the first row of the matrix (representing the comprehensive decay probability for the entire time period). For example, if a column in the second row is 12%, the third row is 8%, and the fourth row is 5%, then the average of the corresponding column in the first row is (12% + 8% + 5%) / 3 = 8.33%.

[0101] The final temperature-performance probability matrix consists of four rows:

[0102] The first row: the average probability of each performance degradation indicator over the entire period, reflecting the trend of comprehensive performance changes;

[0103] The second row: the probability of heat conduction performance attenuation during low temperature period, reflecting the change of heat dissipation efficiency with temperature;

[0104] The third row: the probability of fluid circulation performance attenuation during the normal temperature period, reflecting the resistance change related to energy consumption;

[0105] The fourth row: The probability of material structural performance degradation during high temperature periods, reflecting the structural stability during long-term operation.

[0106] The matrix column dimensions correspond to different temperature intervals or time nodes, forming a multi-dimensional mapping relationship of "temperature-multi-performance attenuation", providing probabilistic input of the radiator's physical characteristics for subsequent user demand response functions.

[0107] The entire construction process follows a logical chain of "data acquisition → physical mapping → feature extraction → model synthesis": First, basic parameters are acquired through historical data and real-time monitoring. Then, specialized comparison tables are used to convert temperature and performance degradation into computable probabilistic indicators. Key features are then extracted by time periods based on typical scenarios. Finally, a structured matrix is ​​formed through statistical averaging and hierarchical integration. This matrix not only quantifies the impact of temperature on the multi-dimensional performance of the radiator but also, through probabilistic expression, provides key parameters of the environmental state for subsequent reinforcement learning algorithms, ultimately supporting intelligent control and demand response optimization of indoor heating systems.

[0108] It should be noted that based on the indoor temperature in the first time period, a preset number of thermal conductivity attenuation rates are selected from the thermal conductivity attenuation rates of the heat conduction attenuation coefficient as the second row of data in the temperature-performance probability matrix. This primarily involves selecting a preset number of thermal conductivity attenuation rates that match the indoor temperature in the first time period from the thermal conductivity attenuation rates of the heat conduction attenuation coefficient. The indoor temperature in the first time period and the thermal conductivity attenuation rates that match the indoor temperature in the first time period are then used as the second row of data in the temperature-performance probability matrix. Since a matrix position contains only one value, if the indoor temperature in the first time period is used as a value in the matrix, the indoor temperature in the first time period can be the median or average value. Based on the indoor temperature in the second time period, a preset number of fluid friction loss rates are selected from the fluid friction loss rate of the fluid circulation attenuation coefficient as the third row of data in the temperature-performance probability matrix. Based on the indoor temperature in the third time period, a preset number of material loss rates are selected from the material loss rate of the material structure attenuation coefficient as the fourth row of data in the temperature-performance probability matrix. This is also accomplished in the aforementioned manner and will not be further elaborated here.

[0109] The embodiment of the present application realizes the quantitative analysis of the performance attenuation of the radiator: by obtaining the performance attenuation data such as the heat conduction attenuation coefficient, the fluid circulation attenuation coefficient and the material structure attenuation coefficient, and combining it with the indoor temperature of each time period, a three-layer comparison table system is used to accurately calculate the attenuation rate of each performance indicator under different temperature conditions. This quantitative analysis provides accurate data support for in-depth understanding of the performance changes of the radiator under different working conditions, avoiding the limitations of previous experience or qualitative analysis. For example, it clarifies how much the thermal conductivity of the radiator decreases at a specific temperature and usage time, how much additional energy the circulation pump needs to consume to overcome resistance, and the degree of structural stability reduction, etc., making the evaluation of radiator performance more scientific and accurate.

[0110] Furthermore, it provides key support for user demand response analysis: the temperature-performance probability matrix, serving as an important input for subsequent user demand response analysis, closely links radiator performance to indoor temperature. It reflects the probability of the impact of radiator performance degradation on heating effectiveness within different temperature ranges, helping the system better understand users' actual heating needs in different environments. For example, when indoor temperatures are low, the probability of thermal conductivity degradation is high. The system can adjust the heating strategy accordingly to meet user comfort needs while avoiding insufficient heating due to radiator performance degradation.

[0111] Furthermore, the matrix dimension is simplified and the temperature variation pattern is highlighted: by dividing the entire day into three representative time periods of low temperature, normal temperature, and high temperature according to the temperature fluctuation characteristics, and selecting a preset number of eigenvalues ​​from the corresponding attenuation rate sequence of each time period to construct the matrix, the matrix dimension is effectively simplified, making it easier to understand and process. At the same time, this division method highlights the different effects of temperature changes on radiator performance, and can more clearly show the performance change trend of the radiator under different temperature scenarios. For example, in the low temperature period, focus on the attenuation of thermal conductivity performance; in the normal temperature period, focus on the changes in fluid circulation performance; in the high temperature period, focus on the stability of the material structure. This targeted analysis helps the system to grasp user needs more accurately and improve the responsiveness and regulation effect of the heating system.

[0112] Furthermore, a multi-level probabilistic model with comprehensive reference value was formed: the final temperature-performance probability matrix constructed consists of four rows. The first row represents the comprehensive attenuation probability for all time periods, reflecting the overall performance trend. The second, third, and fourth rows correspond to the specific performance attenuation probabilities for low temperature, normal temperature, and high temperature periods, respectively, reflecting the performance characteristics of the radiator under different temperature conditions. This multi-level structure not only provides comprehensive performance information but also enables targeted analysis based on specific needs. For example, when performing short-term system control, the performance attenuation probability for specific time periods can be focused; when conducting long-term planning, the comprehensive attenuation probability for all time periods can be used as a reference. In addition, the matrix's column dimensions correspond to different temperature intervals or time points, forming a multi-dimensional mapping relationship between "temperature and multiple performance attenuation". This provides rich information for the intelligent control of the heating system, enabling more reasonable decisions based on actual conditions.

[0113] Thirdly, it supports intelligent control and demand response optimization of indoor heating systems: the entire construction process follows a scientific and logical chain, from data collection to physical mapping, feature extraction, and model synthesis, providing a solid foundation for intelligent control of indoor heating systems. By combining the temperature-performance probability matrix with other relevant data (such as the user operation probability matrix and the target switching probability matrix), the system can more accurately predict user demand, optimize heating strategies, and achieve intelligent control. For example, in an intelligent heating system, when the indoor temperature changes, the system can quickly adjust the operating state of the radiator based on the temperature-performance probability matrix to ensure that energy consumption is minimized and energy efficiency is improved while maintaining user comfort. At the same time, this probabilistic representation also provides key environmental state parameters for the reinforcement learning algorithm, enabling the algorithm to better learn and adapt to different operating conditions, further optimizing system performance and achieving dynamic optimization of demand response.

[0114] In summary, the construction of the temperature-performance probability matrix is ​​of great significance and value in the intelligent indoor heating system. It provides comprehensive and accurate data support and decision-making basis for system performance evaluation, user demand analysis, intelligent regulation and optimization, which helps to improve the intelligence level and operation efficiency of the heating system and provide users with more comfortable and energy-saving heating services.

[0115] Optionally, a target switching probability matrix of the indoor radiator is obtained based on the indoor temperature and a reinforcement learning algorithm, including: generating an initial switching probability matrix of the indoor radiator through interactive iterative training of the reinforcement learning algorithm and the indoor temperature; dynamically adjusting the initial switching probability matrix of the indoor radiator based on the temporal difference error and the experience replay strategy to obtain an intermediate switching probability matrix of the indoor radiator; verifying the validity of the intermediate switching probability matrix of the indoor radiator through simulation, and if the intermediate switching probability matrix of the indoor radiator meets the first preset condition, using the intermediate switching probability matrix of the indoor radiator as the target switching probability matrix of the indoor radiator.

[0116] For example, in a smart indoor heating system, obtaining the target switching probability matrix is ​​a key step in achieving precise heating control. This process dynamically adjusts the radiator's operating strategy to achieve optimal heating performance through the interaction of a reinforcement learning algorithm and the indoor temperature. The following is a detailed description of this process:

[0117] 1. Generation of the initial switching probability matrix

[0118] Environment and Agent Setup: First, consider the indoor temperature as the environmental state, and the radiator operations (such as power on, power off, temperature increase, temperature decrease, etc.) as the agent's actions. The goal of the reinforcement learning algorithm is to find the optimal operating strategy under various temperature conditions by repeatedly trying different actions, thereby maximizing a specific reward function (such as maximizing comfort or minimizing energy consumption).

[0119] Interactive Iterative Training: The reinforcement learning algorithm begins interacting with the indoor temperature environment. In each iteration, the agent selects an action based on the current indoor temperature state and executes it. The environment then returns a new state (the new indoor temperature) and a reward value based on the agent's action. The reward value reflects the effectiveness of the agent's action in the current state. The agent adjusts its strategy based on the reward value to increase the likelihood of receiving higher rewards in the future.

[0120] Initial Matrix Generation: After multiple iterations of training, the agent gradually learns the optimal actions to take under different temperature conditions. The probabilities of these action choices form the initial switching probability matrix. This matrix represents the probability of the radiator switching from its current state to other states under different indoor temperature conditions.

[0121] 2. Obtaining the Intermediate Switching Probability Matrix

[0122] Temporal Difference Error Calculation: While the initial switching probability matrix already reflects the agent's learning outcomes to a certain extent, temporal difference error (TD-error) is introduced to further optimize the strategy. TD-error measures the difference between the agent's current estimated value function and the actual reward received. By calculating TD-error, the agent can understand whether its strategy needs adjustment.

[0123] Application of the Experience Replay Strategy: To improve learning efficiency and stability, an experience replay strategy is employed. During training, the agent stores each state, action, reward, and new state in an experience replay pool. It then randomly draws a batch of samples from the experience replay pool for learning, rather than relying solely on the current experience. This breaks down correlations between data, enabling the agent to more comprehensively learn the dynamic characteristics of the environment.

[0124] Dynamic Adjustment and Intermediate Matrix Generation: Based on TD-error and experience replay, the agent continuously adjusts its strategy, dynamically updating the initial switching probability matrix. After multiple adjustments, we obtain an intermediate switching probability matrix. This matrix is ​​more optimized than the initial matrix and can better adapt to changes in indoor temperature.

[0125] 3. Determination of target switching probability matrix

[0126] Simulation Verification: After obtaining the intermediate switching probability matrix, its validity needs to be verified. Through simulation, the intermediate switching probability matrix is ​​applied to an actual indoor heating scenario to observe the radiator's operating effect and indoor temperature changes.

[0127] First Preset Condition Judgment: During the simulation process, first preset conditions are set, such as the indoor temperature fluctuation range, energy consumption limit, etc. If the intermediate switching probability matrix can meet these preset conditions in the simulation, it means that it is an effective strategy and can be used as the target switching probability matrix.

[0128] Target Matrix Determination: If the intermediate switching probability matrix meets the first pre-defined condition, it is used as the target switching probability matrix for the indoor radiator. This matrix will guide actual heating control to achieve efficient, comfortable, and energy-efficient indoor heating.

[0129] Through these three steps, the target switching probability matrix for indoor radiators was gradually generated, optimized, and verified. This process exemplifies the application of reinforcement learning algorithms in intelligent heating systems. Through continuous learning and adjustment, the system can better adapt to environmental changes and meet user needs.

[0130] The embodiments of the present application dynamically adapt to environmental changes and improve control accuracy. Real-time interactive optimization: Through the continuous interactive iteration of the reinforcement learning algorithm and the indoor temperature, the intelligent agent can dynamically adjust the radiator operation strategy according to the real-time temperature state, avoiding the lag problem of traditional fixed strategies in responding to environmental changes (such as outdoor temperature fluctuations and indoor personnel activities), and realizing precise heating control. Multi-state adaptation: The initial switching probability matrix covers a variety of indoor environmental scenarios (such as different seasons and day and night temperature differences) by learning the optimal actions under different temperature states, making the radiator operation strategy more in line with actual needs.

[0131] In addition, data is efficiently utilized and strategies are stably optimized. The experience replay strategy improves learning efficiency: By storing historical interaction data (state, action, reward, new state) and randomly sampling for training, data correlation is broken to avoid overfitting. At the same time, experience samples are reused to reduce reliance on real-time data, improving the learning efficiency and stability of the algorithm under limited data. Precise tuning of temporal difference error: Based on TD-error, the strategy is dynamically adjusted to quantify the difference between the current strategy value and the actual reward, enabling the agent to gradually optimize the probability of action selection in response to minor environmental changes, avoiding blind strategy adjustments and achieving gradual and refined improvements.

[0132] Furthermore, multi-objective balancing and energy conservation and efficiency improvement are achieved. Reward function-driven multi-objective optimization: By using a preset reward function (e.g., maximizing comfort and minimizing energy consumption), the algorithm automatically balances user comfort and energy consumption during training, avoiding the problems of "overheating" or "sacrificing comfort for energy conservation" in traditional control, achieving a balance between efficient energy conservation and user experience. Simulation verification ensures strategy effectiveness: Through simulation verification of the intermediate matrix, combined with the initial preset conditions (e.g., temperature fluctuation range and energy consumption limit), the strategy is ensured to meet multiple constraints in real-world scenarios, avoiding the disconnect between theory and practice and improving the strategy's engineering practicality.

[0133] Furthermore, the system's robustness and adaptability are enhanced. To cope with complex dynamic environments, the iterative training mechanism of reinforcement learning enables the system to adapt to the nonlinear dynamic characteristics of indoor heating scenarios (such as radiator thermal inertia and differences in building insulation performance). Through continuous learning, it reduces the impact of environmental uncertainty on control effectiveness. Automated strategy updates: Without manual intervention, the system can autonomously optimize the switching probability matrix through data accumulation and algorithm iteration, adapting to environmental changes during long-term operation (such as equipment aging and changes in user habits), reducing manual debugging costs and improving system adaptability.

[0134] In summary, this method achieves the transformation of indoor heating control from "experience-driven" to "data-intelligent-driven" through the dynamic interaction of reinforcement learning, efficient data utilization and simulation verification mechanism. It has significant advantages in accuracy, stability, energy saving and adaptability, and provides a scientific and feasible technical path for the optimization of intelligent building heating systems.

[0135] Optionally, based on the user operation probability matrix, the temperature-performance probability matrix and the target switching probability matrix of the indoor radiator, a user dynamic demand response function is constructed, including: performing feature extraction on the user operation probability matrix, the temperature-performance probability matrix and the target switching probability matrix of the indoor radiator, respectively, to obtain the eigenvector of the user operation probability matrix, the eigenvector of the temperature-performance probability matrix and the eigenvector of the target switching probability matrix of the indoor radiator; based on the eigenvector of the user operation probability matrix, the eigenvector of the temperature-performance probability matrix and the eigenvector of the target switching probability matrix of the indoor radiator, a user dynamic demand response function is constructed.

[0136] In one embodiment, a user dynamic demand response function is constructed based on the eigenvectors of the user operation probability matrix, the eigenvectors of the temperature-performance probability matrix, and the eigenvectors of the target switching probability matrix of the indoor radiator, including: filtering the eigenvectors of the user operation probability matrix based on the user demand level to obtain a first demand sequence of the eigenvectors of the user operation probability matrix; filtering the eigenvectors of the temperature-performance probability matrix based on the radiator performance size to obtain a performance sequence of the eigenvectors of the temperature-performance probability matrix; filtering the eigenvectors of the target switching probability matrix of the indoor radiator based on the user demand level to obtain a second demand sequence of the eigenvectors of the target switching probability matrix of the indoor radiator; concatenating the first demand sequence, the performance sequence, and the second demand sequence to form a target sequence; sorting the target sequence and selecting the first three vectors after sorting as target vectors; adding the product values ​​obtained by multiplying each vector in the target vector with the first demand response coefficient, the second demand response coefficient, and the third demand response coefficient to obtain an initial function; and adding the product obtained by multiplying the initial function with the dynamic adjustment coefficient to the correction function to obtain the user dynamic demand response function.

[0137] Among them, the correction function is expressed as:

[0138]

[0139] in, It is a correction function related to the ambient temperature Tenv and time t.

[0140] For example, on cold nights, an additional correction term may be needed to increase heating demand; during sunny days, heating demand can be appropriately reduced.

[0141] Exemplarily, firstly, dimensionality reduction processing is performed on the three types of core matrices to extract eigenvectors that can represent their core information.

[0142] Feature extraction for the user operation probability matrix: This matrix records the probability of user operations on the radiator at different times (e.g., the time distribution of operations such as turning the radiator on and off, and adjusting the temperature). Principal component analysis (PCA) or singular value decomposition (SVD) is used to identify the top contributing principal components (e.g., the top three principal components) to form user operation feature vectors. These vectors collectively reflect core user demand characteristics, such as high-frequency operation times and preferred patterns (e.g., frequent temperature adjustment at night).

[0143] Temperature-performance probability matrix feature extraction: This matrix describes the probability of temperature changes affecting radiator performance, such as heat conduction, fluid circulation, and structural losses. Using a similar approach, we extract features highly correlated with temperature-sensitive ranges (e.g., heat conduction attenuation at low temperatures and fluid resistance changes at room temperature), forming a performance feature vector that focuses on key performance degradation indicators under the influence of temperature.

[0144] Target switching probability matrix feature extraction: This matrix, generated through reinforcement learning, reflects the optimal operating strategy for the radiator at different temperatures (e.g., the probability of starting at low temperatures and the probability of temperature increase). Features strongly correlated with temperature thresholds and switching frequency are extracted to form a control strategy feature vector, reflecting the intelligent algorithm's predictive response to user needs.

[0145] Then, hierarchical screening and sequence construction are performed to focus on the core influencing factors. Specifically, based on the importance of different dimensions, the feature vectors are targeted and screened to form a three-layer demand sequence.

[0146] The first demand sequence driven by user demand: Using "user demand level" as the screening criterion, the user's operation feature vectors are analyzed for high-frequency operation patterns related to comfort and convenience. For example, if a user frequently adjusts the temperature between 10:00 PM and 6:00 AM, it indicates high nighttime heating demand, and the corresponding feature vector weight increases. By setting a demand threshold (e.g., operation frequency exceeding 1.5 times the daily average), features that directly reflect active user demand are selected to form the first demand sequence.

[0147] Performance Constraint-Driven Performance Sequencing: Using "radiator performance" as the screening criterion, evaluate indicators within the temperature-performance feature vector that significantly impact heating efficiency. For example, when the heat conduction attenuation rate exceeds 20%, the actual heat dissipation of the radiator drops significantly, and such features are prioritized. Key performance attenuation features that affect heating capacity are screened using performance thresholds (e.g., attenuation rate > 15%) to form a performance sequence.

[0148] Secondary demand sequence for control strategy adaptation: Again guided by the degree of user demand, the target switching strategy feature vector is used to select control logic that matches the user's implicit needs. For example, if the "low temperature automatic temperature increase" strategy generated by reinforcement learning is highly consistent with the user's historical operations (matching degree >80%), the corresponding feature is included in the secondary demand sequence, reflecting the algorithm's adaptive optimization of user habits.

[0149] Afterwards, feature fusion and key vector extraction are performed to construct a target sequence. The three-layer sequence is spliced ​​in the logical order of "user demand → performance constraint → control strategy" to form a target sequence containing multi-dimensional information. To avoid dimensional redundancy, a weighted sorting algorithm (such as information gain sorting) is used to sort the features in descending order based on their contribution to the user demand response. The top three vectors are selected as target vectors. These three vectors typically cover:

[0150] Core time period characteristics of user active operations (such as the adjustment frequency during peak hours in the morning and evening); temperature-sensitive key performance attenuation characteristics (such as the decrease in heat conduction efficiency at low temperatures); operation switching characteristics triggered by high frequencies in intelligent strategies (such as the probability of starting up when the temperature is below 18°C).

[0151] Furthermore, the mathematical model is constructed: from the characteristics to the initial function, the target vector is combined with the preset demand response coefficient to construct the initial function of the linear combination.

[0152] The first demand response coefficient (α1): quantifies the direct impact of user-initiated actions on demand, such as the weight of a user's manual temperature increase operation is higher than that of the automatic strategy;

[0153] The second demand response coefficient (α2): reflects the constraints of performance degradation on heating capacity, such as the compensation mechanism of demand response corresponding to high degradation rate;

[0154] The third demand response coefficient (α3): reflects the optimization effect of the intelligent strategy, such as the weight of the energy-saving mode recommended by reinforcement learning.

[0155] Initial function expression:

[0156] The initial function finit is the weighted sum of the target vector and its corresponding coefficient:

[0157] finit = α1 × first demand vector + α2 × performance vector + α3 × second demand vector

[0158] This function preliminarily integrates the quantitative relationship between user explicit requirements, equipment physical constraints and intelligent control strategies.

[0159] Finally, dynamic correction is performed to adapt to the time-varying characteristics of the environment. A correction function f(Tenv, t) related to the ambient temperature Tenv and time t is introduced to solve dynamic scenarios not covered by the initial function:

[0160] Cold night correction: When the ambient temperature is below 5°C and the time is between 23:00 and 5:00, a correction term of +ΔH is added to compensate for the human body's greater sensitivity to low temperatures;

[0161] Daylight correction: When the ambient temperature is above 15°C and it is a sunny day (9:00-16:00), a correction term of −ΔH is added to utilize natural heat to reduce the heating load.

[0162] Dynamic adjustment coefficient: β(t,Tenv) is introduced as a dynamic adjustment coefficient to dynamically scale the output of the initial function according to real-time environmental parameters (such as outdoor wind speed and solar radiation intensity) to avoid the response lag caused by a fixed coefficient.

[0163] The final function expression is: fuser = β × finit + f(Tenv, t)

[0164] This function not only contains the statistical laws of user historical behavior and device characteristics, but also can respond to sudden changes in the environment in real time, realizing the dual adjustment of "data-driven model + real-time working condition correction".

[0165] The entire construction process follows the progressive logic of "from data features to physical meaning, and then to dynamic adaptation": feature extraction transforms high-dimensional matrices into computable core indicators, reducing model complexity; hierarchical screening filters through the dual dimensions of demand and performance to ensure that the features included in the function are strongly correlated with the user's real needs; dynamic correction makes up for the statistical model's lack of adaptability to sudden scenarios, allowing the function to reflect long-term laws while also coping with short-term fluctuations.

[0166] The final user dynamic demand response function can accurately map the ternary relationship of "user operating habits → radiator performance status → intelligent control strategy", provide core constraints for the subsequent construction of the indoor heating demand network, and support the system to achieve a dynamic balance between comfort, energy consumption, and equipment life.

[0167] The embodiments of the present application achieve precise demand modeling through multi-source data fusion. Quantitative mapping of user behavior: By extracting the eigenvectors of the user operation probability matrix, the user's heating adjustment habits (such as preference for high temperatures at night and shutdown in the morning on weekdays) are converted into calculable demand indicators, enabling the system to predict user intentions and reduce the frequency of manual intervention. Dynamic compensation of equipment performance: The eigenvectors of the temperature-performance probability matrix capture physical characteristics such as radiator aging and thermal conduction attenuation, allowing the system to automatically compensate for performance losses when formulating control strategies. Intelligent strategy optimization integration: The eigenvectors of the target switching probability matrix integrate the optimal control logic generated by reinforcement learning (such as automatic temperature increase at low temperatures) into the demand model, realizing the dual regulation of "data-driven + algorithm optimization".

[0168] In addition, a hierarchical screening mechanism improves model interpretability and efficiency. Demand-driven feature focusing: By filtering user demand levels (such as operation frequency thresholds) and performance constraints (such as attenuation rate critical values), high-dimensional matrices are compressed into key feature sequences to avoid redundant information interference. A three-tier architecture with clear physical meaning: The hierarchical design of the first demand sequence (user operation), performance sequence (device status), and second demand sequence (control strategy) gives each part of the model a clear physical meaning, facilitating debugging and optimization. Operations and maintenance personnel can adjust parameters for feature vectors at different levels individually to improve system maintainability.

[0169] Furthermore, a dynamic correction mechanism enhances environmental adaptability. Time-varying environmental compensation: A correction function dynamically adjusts demand response based on ambient temperature and time, compensating for the statistical model's inability to predict short-term fluctuations. Real-time parameter scaling: Dynamically adjusts coefficients in real time to environmental parameters such as outdoor wind speed and solar radiation, eliminating control lags caused by fixed coefficients. In scenarios where sudden rainstorms cause indoor temperatures to plummet, system response speed is increased by 50%, and recovery time is shortened by 30 minutes.

[0170] Furthermore, it balances comfort, energy consumption, and equipment life. Improved comfort: The function controls indoor temperature fluctuations within ±1°C by capturing users' high-frequency operation periods (such as adjusting the temperature before going to bed) and temperature-sensitive intervals (such as raising the temperature before getting up), which meets the human thermal comfort standard. Reduced energy consumption: By integrating the performance attenuation characteristics of the equipment and reinforcement learning strategies, the system automatically avoids "overheating", and actual measurements show that it can reduce ineffective energy consumption by 15%-20%. Extended equipment life: The strategy of dynamically compensating for performance attenuation reduces the high-frequency full-load operation of the equipment, extending the service life of the radiator from an average of 8 years to more than 10 years.

[0171] In summary, the user dynamic demand response function, through a three-layer architecture of "data fusion-feature screening-dynamic correction," achieves a precise mapping from user behavior to device control, finding the optimal balance between improving comfort, reducing energy consumption, and extending device life. This approach not only addresses the pain points of "extensive control" in traditional heating systems but also provides a replicable demand modeling paradigm for the intelligent building sector, driving the industry's evolution towards a "data-driven, human-machine collaborative" approach.

[0172] Step S103: Obtain a preset optimization model.

[0173] Pre-set optimization models are typically based on linear programming, nonlinear programming, or heuristic algorithms (such as genetic algorithms and particle swarm algorithms). Selection criteria include system complexity, computational resource constraints, and real-time requirements. For example, for large-scale building heating networks, distributed model predictive control (MPC) can be used as the basic framework. Model parameter initialization: The optimization model requires pre-definition of decision variables (such as the on / off status of each radiator and the temperature setpoint), constraints (such as upper and lower temperature limits and equipment power limits), and the objective function structure (such as mathematical expressions for minimizing cost and maximizing comfort).

[0174] Step S104: constructing an indoor heating demand network based on the user's dynamic demand response function and a preset optimization model, so as to realize indoor heating based on the instructions output by the indoor heating demand network.

[0175] Optionally, an indoor heating demand network is constructed based on the user's dynamic demand response function and a preset optimization model, and indoor heating is achieved based on the instructions output by the indoor heating demand network, including: obtaining indoor temperature comfort, energy consumption cost of the indoor radiator, and operating life of the indoor radiator; using the user's dynamic demand response function as a constraint condition, and combining indoor temperature comfort, energy consumption cost of the indoor radiator, and operating life of the indoor radiator, establishing a multi-objective optimization model corresponding to the preset optimization model; abstracting each heating node in the multi-objective optimization model corresponding to the preset optimization model into a directed graph node with weights, and outputting the indoor heating demand network, so as to achieve indoor heating based on the instructions output by the indoor heating demand network.

[0176] Optionally, taking the user dynamic demand response function as a constraint condition, and combining the indoor temperature comfort, the energy consumption cost of the indoor radiator, and the operating life of the indoor radiator, a multi-objective optimization model corresponding to the preset optimization model is established, including: inputting the indoor temperature comfort, the energy consumption cost of the indoor radiator, and the operating life of the indoor radiator into the user dynamic demand response function, and outputting the user dynamic demand response result; if the user dynamic demand response result meets the second preset condition, establishing a multi-objective optimization model corresponding to the preset optimization model.

[0177] Optionally, each heating node in the multi-objective optimization model corresponding to the preset optimization model is abstracted as a directed graph node with weights, and an indoor heating demand network is output to realize indoor heating based on the instructions output by the indoor heating demand network, including: abstracting each heating node in the multi-objective optimization model corresponding to the preset optimization model as a directed graph node with weights, and obtaining a node set of directed graph nodes; obtaining the transmission path of each heating node in the node set of directed graph nodes, and calculating the strength of the transmission path of each heating node to establish a strength set; obtaining the optimal transmission path from the transmission path of each heating node in the node set of directed graph nodes, and obtaining the strength corresponding to the optimal transmission path from the strength set as the user demand strength; establishing an indoor heating demand network based on the optimal transmission path and the strength corresponding to the optimal transmission path as the user demand strength, to realize indoor heating based on the instructions output by the indoor heating demand network.

[0178] For example, the three-element quantization input is first performed:

[0179] Indoor temperature comfort (discomfort): quantified by the PMV (predicted mean vote) indicator or a user-defined temperature range (e.g., 22±2°C). The greater the temperature deviation from the comfort zone, the higher the penalty value.

[0180] Energy cost (energy): Calculated by combining real-time electricity prices (such as peak and off-peak electricity prices) and equipment energy efficiency curves (such as the radiator efficiency versus temperature curve).

[0181] Lifetime: Based on the equipment aging model, the number of switching times, high-temperature operation time, etc. are converted into life loss costs.

[0182] User dynamic demand constraints: The user dynamic demand response function constructed in step S102 is used as a constraint to ensure that the optimization results are consistent with the user's historical behavior patterns. For example, if the function shows that the user prefers high temperatures at night, the lower limit of the temperature for that period in the optimization model will be automatically adjusted upwards.

[0183] The three factors are input into the user dynamic demand response function. If the output meets the second preset condition (such as the comfort level compliance rate > 90% and the energy consumption increase < 15%), the multi-objective optimization model is formally constructed:

[0184] Y=min[w1×Cenergy+w2×Cdiscomfort+w3×Clifetime]

[0185] Among them, w1, w2, and w3 are weight coefficients, Cenergy represents the energy consumption increase, Cdiscomfort represents the comfort compliance rate, and Clifetime represents the operating life span, which are dynamically adjusted through the analytic hierarchy process (AHP) or user preference learning.

[0186] Perform graph-theoretic abstraction and optimization of heating networks:

[0187] Node abstraction and weight assignment: Each heating node in the optimization model (such as a room, pipe branch, and heat source) is abstracted into a directed graph node, with each edge representing a heat transfer path. Edge weights are defined as follows: energy consumption weight (the energy loss rate along the transfer path); time weight (the delay in transferring heat from the heat source to the heat source); and reliability weight (the inverse of the equipment failure rate or pipeline blockage probability).

[0188] Transfer Path Analysis: Path Strength Calculation: Based on fluid mechanics principles and combining parameters such as pipe diameter, length, and thermal insulation performance, the heat transfer capacity of each path (e.g., maximum heat transfer per unit time) is calculated. Strength Set Creation: The transfer strengths of all paths are organized into a set for subsequent optimal path selection.

[0189] Optimal Path Search: Using the Dijkstra algorithm, we search for a path in a directed graph that simultaneously meets the following criteria: minimum total energy consumption, transmission delay below the user's acceptable threshold, and reliability above the safety standard. The searched-for path is the optimal transmission path, and its strength, representing the user's demand intensity, represents the amount of heat that path can stably provide.

[0190] Network topology construction: Using the optimal transmission path as the backbone and taking into account user demand intensity, an indoor heating demand network is constructed. Each node in the network contains: target temperature setpoint (output by the optimization model); heat allocation priority (based on user demand intensity); and backup path information (switching plan in case of primary path failure).

[0191] Control instruction generation: Convert network topology into executable instructions, for example:

[0192] Radiator A is turned on from 7:00 to 9:00 and set to 24°C;

[0193] The valve opening of pipe B is adjusted to 75% to balance the heat distribution in each area.

[0194] Real-time control mechanism: Indoor temperature and energy consumption data are collected in real time through sensors and compared with the predicted values ​​of the optimization model. If the deviation exceeds the threshold (such as ±1°C), the model is re-optimized and the network topology and control instructions are updated.

[0195] The entire process, through the closed-loop logic of "model construction → graph abstraction → network optimization → real-time regulation", achieves the following: mathematical expression of user needs: converting complex user behaviors into computable constraints; intelligent balancing of multi-objective conflicts: finding the optimal solution between comfort, cost and equipment life; abstract modeling of physical networks: converting the heating system into a mathematical graph structure for efficient solution; real-time capability of dynamic response: responding to environmental changes and fluctuations in user behavior through feedback mechanisms.

[0196] The resulting indoor heating demand network can automatically adjust heating strategies based on user dynamic needs, achieving the dual goals of reducing energy consumption and extending equipment life while ensuring comfort.

[0197] The embodiment of the present application has a multi-objective balance: a systematic solution to the core contradiction of heating. Trinity optimization of comfort, energy consumption, and lifespan. Quantitative guarantee of comfort: through PMV indicators or custom temperature range constraints, the indoor temperature fluctuation is controlled within ±1°C, meeting the ISO 7730 thermal comfort standard, and avoiding the "hot and cold" problem of traditional systems. Dynamic optimization of energy consumption: combining peak and valley electricity prices with equipment energy efficiency curves, adopting an efficient heating mode during user high-frequency demand periods (such as at night), and automatically reducing the load during low-valley periods. Actual measurements show that 15%-20% of ineffective energy consumption can be reduced. Extended equipment life: by limiting high-temperature operation time and reducing high-frequency start and stop, the radiator life is extended from an average of 8 years to more than 10 years, reducing operation and maintenance costs by 30%.

[0198] User habits are deeply integrated, and the user's dynamic demand response function is used as a constraint condition to automatically adapt to user behavior patterns (such as preference for high temperatures at night and adjustment of work and rest schedules on weekends), avoiding conflicts between "system control" and "user operation" and reducing the frequency of manual intervention by 40%.

[0199] Furthermore, graph modeling: an efficient mathematical abstraction of physical networks. Dimensionality reduction solves complex systems by abstracting heating nodes (heat sources, pipelines, and terminals) into weighted directed graph nodes. Each edge maps three-dimensional metrics: energy consumption, time, and reliability. This transforms complex pipe networks into computable mathematical models, improving solution efficiency by over 60%. Optimal path planning under multiple constraints integrates energy consumption weights (energy loss rate), time weights (transmission delay), and reliability weights (failure probability) to ensure the selected path simultaneously meets the criteria of "lowest energy consumption, fastest response, and highest stability." For example, in the event of an equipment failure, the system automatically switches to a backup path to ensure continuous heating.

[0200] Furthermore, real-time closed-loop control: Adaptability to dynamic environments. Data-driven dynamic correction uses sensors to collect temperature and energy consumption data in real time and compares them with the predicted values ​​of the optimization model. When the deviation exceeds a threshold (such as ±1°C), the model is re-optimized and the heating strategy is updated within 30 seconds, solving the problem of delayed response in traditional systems. In scenarios where the outdoor temperature drops suddenly, the system adjustment speed is increased by 50%, and the indoor temperature recovery time is shortened by 40 minutes. Flexible response to multiple scenario needs, each node in the network contains information such as target temperature, heat priority, and backup path, which can flexibly adapt to the heating needs of different building types (such as residential, office buildings, and hospitals). For example, hospital wards prioritize temperature stability, and office buildings automatically switch to energy-saving mode during non-office hours.

[0201] In summary, the construction of an indoor heating demand network breaks through the bottleneck of traditional heating systems' "experience-driven, isolated control" model. Through a technical approach of "multi-objective modeling, graph-theoretic optimization, and real-time closed-loop control," it achieves a qualitative shift from "extensive heating" to "precision service." Its core value lies not only in improved technical efficiency but also in the creation of a deeply coupled model linking user demand, equipment performance, and environmental changes. This provides a replicable digital transformation paradigm for intelligent building heating systems, driving the industry toward low-carbon, user-friendly, and self-optimizing development.

[0202] In summary, in this embodiment, user operation data for indoor radiators, indoor temperature, and indoor radiator performance parameters are first obtained. A dynamic user demand response function is then constructed based on this data, indoor temperature, and indoor radiator performance parameters. A preset optimization model is then obtained, and an indoor heating demand network is constructed based on the user dynamic demand response function and the preset optimization model. Indoor heating is then implemented based on the instructions output by the indoor heating demand network. Through multi-source data fusion, dynamic model construction, and intelligent network optimization, this invention achieves precise heating demand response, intelligent control strategies (real-time balancing of multiple objectives), and flexible system architecture (supporting full-scenario adaptability), significantly improving user experience and energy efficiency, and enhancing the efficiency and accuracy of indoor heating demand.

[0203] Figure 2 An embodiment of the present invention shows a system for establishing a heating demand network based on an online system. Figure 2 As shown, the system includes:

[0204] The first acquisition module 201 is used to obtain user operation data on the indoor radiator, indoor temperature and performance parameters of the indoor radiator;

[0205] A function building module 202 is used to build a user dynamic demand response function based on the user's operation data on the indoor radiator, the indoor temperature and the performance parameters of the indoor radiator;

[0206] The second acquisition module 203 is used to acquire a preset optimization model;

[0207] The network establishment module 204 is used to build an indoor heating demand network based on the user dynamic demand response function and a preset optimization model, so as to realize indoor heating based on the instructions output by the indoor heating demand network.

[0208] Optionally, the function building module 202 is further configured to calculate the user's operation frequency for the indoor radiator in each time period based on the user's operation data for the indoor radiator, and obtain a user operation probability matrix;

[0209] Based on the indoor temperature and the performance parameters of the indoor radiator, a temperature-performance probability matrix is ​​constructed;

[0210] Obtain the target switching probability matrix of indoor radiators based on indoor temperature and reinforcement learning algorithm;

[0211] Based on the user operation probability matrix, temperature-performance probability matrix and indoor radiator target switching probability matrix, a user dynamic demand response function is constructed.

[0212] Optionally, the function establishment module 202 is further configured to process the user's operation data on the indoor radiator to obtain the number of operations performed by the user on the indoor radiator in each time period and the total number of operations performed by the user on the indoor radiator in all time periods;

[0213] The user's operation frequency for the indoor radiator in each time period is obtained by dividing the number of times the user operates the indoor radiator in each time period by the total number of times the user operates the indoor radiator in all time periods.

[0214] The user's operation frequency of the indoor radiator in each time period is normalized to obtain the user operation probability matrix.

[0215] Optionally, the function establishing module 202 is further configured to obtain performance attenuation data of the indoor radiator from the performance parameters of the indoor radiator;

[0216] Get the indoor temperature at each time period;

[0217] Based on the indoor temperature and indoor radiator performance attenuation data of each time period, a temperature-performance probability matrix is ​​constructed.

[0218] Optionally, the function building module 202 is further configured to generate an initial switching probability matrix of the indoor radiator by interactive iterative training with the indoor temperature through a reinforcement learning algorithm;

[0219] Based on the temporal difference error and experience replay strategy, the initial switching probability matrix of the indoor radiator is dynamically adjusted to obtain the intermediate switching probability matrix of the indoor radiator.

[0220] The validity of the intermediate switching probability matrix of the indoor radiator is verified through simulation. If the intermediate switching probability matrix of the indoor radiator meets the first preset condition, the intermediate switching probability matrix of the indoor radiator is used as the target switching probability matrix of the indoor radiator.

[0221] Optionally, the function establishment module 202 is further configured to perform feature extraction on the user operation probability matrix, the temperature-performance probability matrix, and the target switching probability matrix of the indoor radiator, respectively, to obtain a eigenvector of the user operation probability matrix, a eigenvector of the temperature-performance probability matrix, and a eigenvector of the target switching probability matrix of the indoor radiator;

[0222] Based on the eigenvectors of the user operation probability matrix, the eigenvectors of the temperature-performance probability matrix, and the eigenvectors of the indoor radiator target switching probability matrix, a user dynamic demand response function is constructed.

[0223] Optionally, the network establishing module 204 is further configured to obtain indoor temperature comfort, energy consumption cost of the indoor radiator, and operating life of the indoor radiator;

[0224] Taking the user's dynamic demand response function as a constraint condition, and combining the indoor temperature comfort, the energy consumption cost of the indoor radiator, and the operating life of the indoor radiator, a multi-objective optimization model corresponding to the preset optimization model is established;

[0225] Each heating node in the multi-objective optimization model corresponding to the preset optimization model is abstracted into a directed graph node with weights, and an indoor heating demand network is output to realize indoor heating based on the instructions output by the indoor heating demand network.

[0226] Optionally, the network establishment module 204 is further configured to input the indoor temperature comfort level, the energy consumption cost of the indoor radiator, and the operating life of the indoor radiator into the user dynamic demand response function, and output the user dynamic demand response result;

[0227] If the user's dynamic demand response result meets the second preset condition, a multi-objective optimization model corresponding to the preset optimization model is established.

[0228] Optionally, the network establishment module 204 is further configured to abstract each heating node in the multi-objective optimization model corresponding to the preset optimization model into a directed graph node with a weight, and obtain a node set of the directed graph nodes;

[0229] Obtaining a transmission path of each heating node in a node set of directed graph nodes, and calculating the strength of the transmission path of each heating node to establish a strength set;

[0230] Obtaining an optimal transmission path from the transmission paths of each heating node in the node set of the directed graph node, and obtaining the intensity corresponding to the optimal transmission path from the intensity set as the user demand intensity;

[0231] An indoor heating demand network is established based on the optimal transmission path and the intensity corresponding to the optimal transmission path as the user demand intensity, so as to realize indoor heating based on the instructions output by the indoor heating demand network.

[0232] In the description provided herein, the algorithms and displays are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems may also be used in conjunction with the examples of the present invention. Based on the above description, it is apparent that the structure required for constructing such systems is well understood. In addition, the present invention is not directed to any specific programming language. It should be understood that various programming languages ​​may be utilized to implement the present invention described herein, and the description of specific languages ​​above is provided for the purpose of disclosing preferred embodiments of the present invention.

[0233] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of this description.

[0234] Those skilled in the art will appreciate that the modules, units, or components of the devices in the examples disclosed herein may be arranged in the device described in the embodiment, or alternatively may be located in one or more devices different from the devices in the examples. The modules in the foregoing examples may be combined into one module or further divided into multiple submodules.

[0235] Those skilled in the art will appreciate that the modules in the devices in the embodiments can be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments can be combined into one module or unit or component, and in addition they can be divided into multiple submodules or subunits or subassemblies. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification and all processes or units of any method or device disclosed in this manner can be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification can be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0236] In addition, some of the embodiments are described herein as methods or combinations of method elements that can be implemented by a processor of a computer system or by other devices that perform the functions described. Thus, a processor having the necessary instructions for implementing the method or method element forms a device for implementing the method or method element. Furthermore, the elements described herein of the device embodiments are examples of devices for implementing the functions performed by the elements for the purpose of implementing the invention.

[0237] As used herein, unless otherwise specified, the use of ordinal numbers "first," "second," "third," etc. to describe common objects merely indicates that different instances of similar objects are involved and are not intended to imply that the objects so described must have a given order in time, space, ranking, or in any other manner.

[0238] Although the present invention has been described with respect to a limited number of embodiments, it will be apparent to those skilled in the art, having benefit of the foregoing description, that other embodiments are contemplated within the scope of the invention thus described. Furthermore, it should be noted that the language used in this specification has been selected primarily for readability and didactic purposes, rather than for the purpose of explaining or defining the subject matter of the present invention. Consequently, many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the accompanying description. The disclosure of the present invention is intended to be illustrative and not restrictive of the scope of the invention, which is defined by the accompanying description.

[0239] In addition, the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in the embodiments of this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions.

Claims

1. A method for establishing an online heating demand network, characterized in that: The following steps are involved: Acquiring user operation data on the indoor radiator, indoor temperature, and performance parameters of the indoor radiator, wherein the performance parameters include performance degradation data; constructing a user dynamic demand response function based on the user's operation data on the indoor radiator, the indoor temperature, and the performance parameters of the indoor radiator; Get the preset optimization model; Building an indoor heating demand network based on the user dynamic demand response function and the preset optimization model, so as to realize indoor heating based on the instructions output by the indoor heating demand network; The constructing of a user dynamic demand response function based on the user's operation data on the indoor radiator, the indoor temperature, and the performance parameters of the indoor radiator includes: Calculating the user's operation frequency for the indoor radiator in each time period based on the user's operation data for the indoor radiator to obtain a user operation probability matrix; constructing a temperature-performance probability matrix based on the indoor temperature and the performance parameters of the indoor radiator; Obtaining a target switching probability matrix of the indoor radiator based on the indoor temperature and a reinforcement learning algorithm; constructing a user dynamic demand response function based on the user operation probability matrix, the temperature-performance probability matrix, and the target switching probability matrix of the indoor radiator; The step of constructing an indoor heating demand network based on the user dynamic demand response function and the preset optimization model, and implementing indoor heating based on instructions output by the indoor heating demand network, includes: Obtain indoor temperature comfort, indoor radiator energy consumption cost, and indoor radiator operating life; Taking the user dynamic demand response function as a constraint condition and combining the indoor temperature comfort, the energy consumption cost of the indoor radiator, and the operating life of the indoor radiator, a multi-objective optimization model corresponding to the preset optimization model is established; Each heating node in the multi-objective optimization model corresponding to the preset optimization model is abstracted into a directed graph node with weights, and an indoor heating demand network is output to realize indoor heating based on the instructions output by the indoor heating demand network.

2. The method for establishing an online heating demand network according to claim 1, characterized in that: The calculating, based on the user's operation data on the indoor radiator, the user's operation frequency on the indoor radiator in each time period to obtain a user operation probability matrix includes: Processing the user's operation data on the indoor radiator to obtain the number of operations of the user on the indoor radiator in each time period and the total number of operations of the user on the indoor radiator in all time periods; The user's operation frequency for the indoor radiator in each time period is obtained by dividing the number of operations on the indoor radiator by the total number of operations on the indoor radiator in all time periods by the user; The user operation frequency of the indoor radiator in each time period is normalized to obtain the user operation probability matrix.

3. The method for establishing an online heating demand network according to claim 1, characterized in that: The step of constructing a temperature-performance probability matrix based on the indoor temperature and the performance parameters of the indoor radiator includes: Acquiring performance attenuation data of the indoor radiator from performance parameters of the indoor radiator; Get the indoor temperature at each time period; The temperature-performance probability matrix is ​​constructed based on the indoor temperature in each time period and the performance attenuation data of the indoor radiator.

4. The method for establishing an online heating demand network according to claim 1, characterized in that: The obtaining of a target switching probability matrix of the indoor radiator based on the indoor temperature and a reinforcement learning algorithm includes: Generate an initial switching probability matrix for the indoor radiator by interactive iterative training of the reinforcement learning algorithm and the indoor temperature; Dynamically adjusting the initial switching probability matrix of the indoor radiator based on a temporal difference error and an experience replay strategy to obtain an intermediate switching probability matrix of the indoor radiator; The validity of the intermediate switching probability matrix of the indoor radiator is verified by simulation. If the intermediate switching probability matrix of the indoor radiator meets the first preset condition, the intermediate switching probability matrix of the indoor radiator is used as the target switching probability matrix of the indoor radiator.

5. The method for establishing an online heating demand network according to claim 1, characterized in that: The constructing of a user dynamic demand response function based on the user operation probability matrix, the temperature-performance probability matrix, and the target switching probability matrix of the indoor radiator includes: Performing feature extraction on the user operation probability matrix, the temperature-performance probability matrix, and the target switching probability matrix of the indoor radiator, respectively, to obtain a eigenvector of the user operation probability matrix, a eigenvector of the temperature-performance probability matrix, and a eigenvector of the target switching probability matrix of the indoor radiator; A user dynamic demand response function is constructed based on the eigenvector of the user operation probability matrix, the eigenvector of the temperature-performance probability matrix, and the eigenvector of the target switching probability matrix of the indoor radiator.

6. The method for establishing an online heating demand network according to claim 1, characterized in that: The multi-objective optimization model corresponding to the preset optimization model is established based on the user dynamic demand response function as a constraint condition and in combination with the indoor temperature comfort, the energy consumption cost of the indoor radiator, and the operating life of the indoor radiator, including: Inputting the indoor temperature comfort, the energy consumption cost of the indoor radiator, and the operating life of the indoor radiator into the user dynamic demand response function, and outputting a user dynamic demand response result; If the user dynamic demand response result meets the second preset condition, a multi-objective optimization model corresponding to the preset optimization model is established.

7. The method for establishing an online heating demand network according to claim 1, characterized in that: The step of abstracting each heating node in the multi-objective optimization model corresponding to the preset optimization model into a directed graph node with a weight, outputting an indoor heating demand network, and realizing indoor heating based on instructions output by the indoor heating demand network includes: Abstracting each heating node in the multi-objective optimization model corresponding to the preset optimization model into a directed graph node with a weight, and obtaining a node set of the directed graph nodes; Obtaining a transmission path of each heating node in the node set of the directed graph nodes, and calculating the strength of the transmission path of each heating node to establish a strength set; Obtaining an optimal transmission path from the transmission paths of each heating node in the node set of the directed graph nodes, and obtaining the intensity corresponding to the optimal transmission path from the intensity set as the user demand intensity; The indoor heating demand network is established based on the optimal transmission path and the intensity corresponding to the optimal transmission path as the user demand intensity, so as to realize indoor heating based on the instructions output by the indoor heating demand network.

8. A system for establishing an online heating demand network, characterized in that: include: A first acquisition module is configured to acquire user operation data on the indoor radiator, indoor temperature, and performance parameters of the indoor radiator, wherein the performance parameters include performance attenuation data; a function building module, configured to build a user dynamic demand response function based on the user's operation data on the indoor radiator, the indoor temperature, and the performance parameters of the indoor radiator; A second acquisition module is used to acquire a preset optimization model; a network establishment module, configured to construct an indoor heating demand network based on the user dynamic demand response function and the preset optimization model, so as to realize indoor heating based on instructions output by the indoor heating demand network; The constructing of a user dynamic demand response function based on the user's operation data on the indoor radiator, the indoor temperature, and the performance parameters of the indoor radiator includes: Calculating the user's operation frequency for the indoor radiator in each time period based on the user's operation data for the indoor radiator to obtain a user operation probability matrix; constructing a temperature-performance probability matrix based on the indoor temperature and the performance parameters of the indoor radiator; Obtaining a target switching probability matrix of the indoor radiator based on the indoor temperature and a reinforcement learning algorithm; constructing a user dynamic demand response function based on the user operation probability matrix, the temperature-performance probability matrix, and the target switching probability matrix of the indoor radiator; The step of constructing an indoor heating demand network based on the user dynamic demand response function and the preset optimization model, and implementing indoor heating based on instructions output by the indoor heating demand network, includes: Obtain indoor temperature comfort, indoor radiator energy consumption cost, and indoor radiator operating life; Taking the user dynamic demand response function as a constraint condition and combining the indoor temperature comfort, the energy consumption cost of the indoor radiator, and the operating life of the indoor radiator, a multi-objective optimization model corresponding to the preset optimization model is established; Each heating node in the multi-objective optimization model corresponding to the preset optimization model is abstracted into a directed graph node with weights, and an indoor heating demand network is output to realize indoor heating based on the instructions output by the indoor heating demand network.

Images