Edge calculation control method and system based on building heat supply energy-saving regulation and control

Abstract

The invention discloses an edge calculation control method and system based on building heat supply energy-saving regulation and control, and relates to the technical field of building heat supply, and the method comprises the following steps: when a heat supply system starts to operate, firstly setting an initial heat load demand prediction frequency for establishing a basic prediction mechanism; the method comprises the following steps: deploying edge nodes in each key area in a building for collecting temperature data of each edge node in real time; a fixed monitoring time window is set, and key feature extraction is carried out on temperature data collected in the monitoring time window. According to the method, temperature data are collected through edge nodes, features are extracted in a monitoring time window, a deep learning model is called, the heat supply state is divided into a stable type and a frequent fluctuation type, the prediction frequency is dynamically improved in the fluctuation type, heat supply parameters are optimized, it is ensured that the room temperature is stable, the heat supply response speed and efficiency are improved, and energy waste is reduced; start-stop frequency and abrasion of equipment are reduced, and service life of the equipment is prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of building heating, and in particular to an edge computing control method and system based on building heating energy-saving regulation. Background Art

[0002] Edge computing control based on building heating energy-saving regulation refers to the installation of edge computing equipment at the building level to monitor and analyze the operating data of the heating system in real time, such as temperature, humidity, outdoor climate conditions and user needs, and perform local computing and intelligent regulation. Through edge computing technology, it is possible to quickly respond to changes in heating demand within the building, dynamically adjust heating parameters, optimize energy efficiency, reduce energy waste, avoid overheating, and ultimately achieve the goal of energy saving and consumption reduction, while reducing dependence on central servers and improving the intelligence and autonomy of the heating system.

[0003] During the heating process of the building, edge computing devices are deployed inside the building to process data collected from sensors and smart terminal devices. With the help of built-in intelligent algorithms, the equipment analyzes and models the data locally, identifies changes in heating demand in the building in real time, and predicts short-term heat load demand. These processing operations do not need to transmit all data to the central server, but are completed quickly directly at the edge node. Based on the analysis results of edge computing, the system intelligently adjusts heating parameters, such as water temperature, valve switching, etc., to accurately match the current heat load demand, ensure the efficiency and timeliness of the heating system, and avoid excessive or insufficient heating.

[0004] The prior art has the following deficiencies:

[0005] In the prior art, a fixed heat load demand forecast frequency is usually used in the building heating process. The heat load demand forecast frequency is generally set low to ensure the effective use of resources. When the edge node in the building detects that the temperature is relatively stable, the heat load demand forecast frequency can meet the adjustment of the heating demand, because when the temperature changes slightly, the heat load demand fluctuation is limited, and the heating system does not need to be adjusted frequently. This method performs well in balancing system load and energy consumption, avoiding waste of resources caused by too frequent forecasts, and ensuring that the heating system meets basic needs.

[0006] However, when the temperature of the building edge nodes is unstable, the fixed frequency of heat load demand forecasting will cause problems. Since the temperature changes frequently and rapidly, the fixed frequency forecast cannot capture these fluctuations in time, resulting in a delayed response speed of the heating system.

[0007] For example:

[0008] When the temperature inside a building suddenly drops, the heating system fails to increase the heat load in time, which may cause the indoor temperature to drop sharply in a short period of time. Especially in cold weather, residents may feel significant cold discomfort, affecting their daily life and work, and even causing health problems such as colds or hypothermia. This hysteresis effect may have a serious impact on the comfort and health of residents in extreme cases.

[0009] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0010] The purpose of the present invention is to provide an edge computing control method and system based on building heating energy-saving regulation to solve the problems in the above-mentioned background technology.

[0011] In order to achieve the above object, the present invention provides the following technical solution: an edge computing control method based on building heating energy-saving regulation, comprising the following steps:

[0012] When the heating system starts to operate, an initial heat load demand forecast frequency is first set to establish a basic forecast mechanism;

[0013] Deploy edge nodes in key areas of the building to collect temperature data from each edge node in real time;

[0014] Set a fixed monitoring time window, extract key features from the temperature data collected within the monitoring time window, and perform anomaly analysis on the extracted features. Based on the anomaly analysis, call the pre-trained deep learning model to evaluate the temperature changes of all edge nodes in the building.

[0015] According to the evaluation results of the deep learning model, the heating status of the building is divided into two categories: overall stable type and overall frequent fluctuation type;

[0016] If the deep learning model is evaluated as overall stable, the building heating process will continue to be predicted according to the set initial heat load demand prediction frequency;

[0017] When it is detected that the temperature in the building fluctuates frequently as a whole, the frequency of heat load demand prediction is dynamically increased according to the temperature changes of all edge nodes, the heating response speed is improved, and problems caused by excessive or insufficient heating are avoided.

[0018] Preferably, the steps for setting the initial heat load demand forecast frequency are as follows:

[0019] Before setting the initial heat load demand forecast frequency, analyze the building type, building use, historical heating data and regional climate conditions. By collecting heating data from past heating seasons, understand the changing pattern of heating demand. At the same time, combine meteorological data analysis to determine the impact of different temperature ranges, wind speeds, humidity and precipitation conditions on heat load.

[0020] Evaluate the structural characteristics of the building and user needs to preliminarily determine the appropriate range of the prediction frequency, so as to meet user needs while avoiding excessive calculation and waste of resources;

[0021] After comprehensive analysis of historical data, climate conditions and building characteristics, an initial heat load demand forecast frequency range is set.

[0022] Preferably, key features are extracted from the collected temperature data, wherein the extracted key features include the diffusion speed and amplitude of the temperature difference between edge nodes in different regions, and the temperature symmetry between adjacent edge nodes. Within the monitoring time window, after abnormal analysis of the diffusion speed and amplitude of the temperature difference between edge nodes in different regions, and the temperature symmetry between adjacent edge nodes, a temperature difference diffusion quantification value and a temperature asymmetry quantification value are generated respectively. The temperature difference diffusion quantification value reflects the propagation of temperature changes in the building and the severity of temperature fluctuations in local areas. The temperature asymmetry quantification value reflects whether the heat distribution of the heating system between various regions is uniform and whether there is local temperature imbalance caused by uneven heating.

[0023] Preferably, after obtaining the temperature difference diffusion quantization value and the temperature asymmetry quantization value generated by abnormal analysis of the extracted key features, the temperature difference diffusion quantization value and the temperature asymmetry quantization value are comprehensively analyzed through a pre-trained deep learning model to generate an overall temperature stability coefficient, and the temperature changes of all edge nodes in the building are evaluated through the overall temperature stability coefficient.

[0024] Preferably, within the monitoring time window, the specific steps of generating the temperature difference diffusion quantization value after abnormal analysis of the diffusion speed and amplitude of the temperature difference between edge nodes in different regions are as follows:

[0025] Calculate the instantaneous temperature difference change rate of edge nodes in adjacent areas to capture the speed of heat diffusion. The calculation expression is:

[0026]

[0027] Where, ΔT i (t) represents the instantaneous temperature difference change rate of the i-th edge node, T i (t) represents the temperature of the i-th node at time t, Δt represents the time interval between two sampling time points, and α is the time attenuation coefficient, which is used to control the influence of the time interval on the rate of change;

[0028] The temperature differences of all adjacent areas are accumulated, and a weight coefficient is introduced to reflect the diffusion characteristics of each node to quantify the cumulative diffusion amplitude of temperature diffusion. The calculation expression of the cumulative diffusion amplitude is:

[0029]

[0030] Where A(t) represents the cumulative diffusion amplitude of all regions at time t, γ i represents the weight coefficient of node i, which is used to measure the diffusion influence of different nodes, N represents the total number of edge nodes in the building, and β is the nonlinear expansion index, which is used to enhance the sensitivity to severe fluctuations;

[0031] The instantaneous temperature difference change rate and the cumulative diffusion amplitude between nodes are combined to generate the temperature difference diffusion quantization value. The calculation expression of the temperature difference diffusion quantization value is:

[0032]

[0033] Where, DI index It represents the quantitative value of temperature difference diffusion, which is used to comprehensively reflect the propagation of temperature changes in the building.

[0034] Preferably, within the monitoring time window, the specific steps of generating a temperature asymmetry quantization value after performing an abnormal analysis on the temperature symmetry between adjacent edge nodes are as follows:

[0035] Establish a temperature difference function D u，k (t), where D u，k (t) represents the temperature difference function between two adjacent edge nodes u and k at time t, which is used to quantify the temperature inconsistency between two adjacent edge nodes and measure the temperature difference between adjacent edge nodes u and k at time t. The expression of the temperature difference function is: u，k (t)=(T u (t)-T k (t)) 2 +p·|T u (t)-T k (t)|, where T u (t) and T k (t) represents the temperature of edge nodes u and k at time t, and the weight coefficient of the absolute value term p is used to balance the square term (T u (t)-T k (t)) 2 and the absolute value term |T u (t)-T k (t) relative impact;

[0036] In a monitoring time window W, the temperature differences of all adjacent edge nodes are accumulated and calculated, and the time weighted method is introduced to calculate the weighted cumulative asymmetry. The calculation expression is:

[0037]

[0038] In the formula, C uk (W) represents the weighted cumulative asymmetry of edge nodes u and k within the monitoring time window W, and ω(t) represents the time weighting coefficient, which is defined as Among them, t new is the latest time point in the monitoring time window, and W represents the length of the monitoring time window;

[0039] The results of all adjacent nodes in the building are normalized and the temperature asymmetry quantification value is generated to quantify the balance of the overall heating system. The generation expression of the temperature asymmetry quantification value is:

[0040]

[0041] In the formula, I asym It represents the quantitative value of temperature asymmetry, which is used to measure the uniformity of heat distribution in a building. P represents the set of all adjacent nodes in the building.

[0042] Preferably, the overall temperature stability coefficient generated when evaluating the temperature changes of all edge nodes in the building through the deep learning model within the monitoring time window is compared and analyzed with the preset overall temperature stability coefficient reference threshold, and the heating status of the building is divided into two categories. The specific division process is as follows:

[0043] If the overall temperature stability coefficient is less than the overall temperature stability coefficient reference threshold, the heating state of the building in the monitoring time window is classified as an overall frequent fluctuation type;

[0044] If the overall temperature stability coefficient is greater than or equal to the overall temperature stability coefficient reference threshold, the heating status of the building within the monitoring time window is classified as an overall stable type.

[0045] Preferably, the frequency of heat load demand prediction is dynamically increased according to the temperature changes of all edge nodes. The specific steps are as follows:

[0046] The overall temperature stability coefficient generated by the deep learning model is compared with the overall temperature stability coefficient reference threshold, and the stability coefficient deviation is calculated to quantify the degree of temperature fluctuation. The calculation expression is: ΔS = Overall sef -Overall Tx , where ΔS represents the temperature stability coefficient deviation, which is used to measure the difference between the current temperature fluctuation and the reference standard.sef Indicates the preset temperature stability coefficient reference threshold, Overall Tx Indicates the overall temperature stability coefficient calculated within the current monitoring time window;

[0047] According to the temperature stability coefficient deviation and the monitoring density of the edge nodes, the dynamic adjustment coefficient is calculated to guide the improvement of the prediction frequency. The calculation expression is:

[0048]

[0049] In the formula, α' represents the dynamic adjustment coefficient, which indicates the influence of the intensity of temperature fluctuation on the predicted frequency;

[0050] The new actual forecast frequency is calculated based on the initial heat load demand forecast frequency and the dynamic adjustment coefficient. The calculation expression is: new =f initial ·(1+β'·α'), where f new is the actual predicted frequency after adjustment, f initial is the initial heat load demand prediction frequency, β' is the response sensitivity coefficient, which controls the increase of the prediction frequency;

[0051] In order to avoid resource waste and system overload caused by too high prediction frequency, the upper limit of the prediction frequency is set, and the adjusted actual prediction frequency is compared with the upper limit of the prediction frequency to obtain the final application prediction frequency. The determination expression of the application prediction frequency is: final =min(f new , f max ), where f final The frequency predicted for the final application is used for practical application, f max The maximum prediction frequency allowed by the system.

[0052] The edge computing control method based on building heating energy-saving regulation includes an initial prediction frequency setting module, a temperature data acquisition module, a feature extraction and abnormality analysis module, a heating state recognition module, a stable prediction module, and a frequent fluctuation response module;

[0053] The initial prediction frequency setting module sets an initial heat load demand prediction frequency when the heating system starts to operate, which is used to establish a basic prediction mechanism;

[0054] The temperature data collection module deploys edge nodes in key areas of the building to collect temperature data from each edge node in real time;

[0055] The feature extraction and anomaly analysis module sets a fixed monitoring time window, extracts key features from the temperature data collected within the monitoring time window, and performs an anomaly analysis on the extracted features. Based on the anomaly analysis, the pre-trained deep learning model is called to evaluate the temperature changes of all edge nodes in the building.

[0056] The heating status recognition module divides the building heating status into two categories based on the evaluation results of the deep learning model: overall stable type and overall frequent fluctuation type;

[0057] The stable prediction module, when the deep learning model is evaluated as overall stable, continues to predict the building heating process according to the set initial heat load demand prediction frequency;

[0058] The frequent fluctuation response module dynamically increases the frequency of heat load demand prediction based on the temperature changes of all edge nodes when it detects that the temperature in the building is frequently fluctuating as a whole, thereby improving the heating response speed and avoiding problems caused by excessive or insufficient heating.

[0059] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0060] The present invention utilizes the temperature data collected in real time by the edge node, extracts key features and performs abnormal analysis within a fixed monitoring time window, and divides the heating status of the building into overall stable type and overall frequent fluctuation type by calling a pre-trained deep learning model. In the stable type with small temperature changes, the system maintains the initial prediction frequency to avoid excessive calculation and energy waste; in the frequent fluctuation type, the system dynamically increases the prediction frequency and quickly adjusts the heating parameters to ensure that the indoor temperature is always maintained in a comfortable range, avoiding the cold discomfort caused by insufficient heating and the energy waste caused by excessive heating. It not only optimizes the response speed and heating efficiency of the system, but also reduces the frequent start and stop and excessive wear of the equipment, and increases the service life of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0062] Figure 1 This is a method flow chart of the edge computing control method based on building heating energy-saving regulation of the present invention.

[0063] Figure 2 This is a module schematic diagram of the edge computing control system based on building heating energy-saving regulation of the present invention. DETAILED DESCRIPTION

[0064] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0065] The present invention provides Figure 1 The edge computing control method based on building heating energy-saving regulation shown includes the following steps:

[0066] When the heating system starts to operate, an initial heat load demand forecast frequency is first set to establish a basic forecast mechanism;

[0067] The steps to set the initial heat load demand forecast frequency are as follows:

[0068] Before setting the initial heat load demand forecast frequency, you first need to analyze the building type, building use, historical heating data, and regional climate conditions. By collecting heating data from past heating seasons, you can understand the changing patterns of heating demand, such as the impact of different time periods, weather conditions, and building use on heating demand. At the same time, combined with meteorological data analysis, determine the impact of different temperature ranges, wind speeds, humidity, and precipitation on heat load. This step ensures that the setting of the initial forecast frequency is scientific and targeted, laying a data foundation for subsequent optimization.

[0069] In addition to climate and historical data, the structural characteristics of the building and user needs must also be evaluated. The building's insulation performance, orientation, floor height, and residents' heating habits will affect the demand for heating. For example, newly built buildings may have better insulation performance and are insensitive to temperature changes, while old buildings require frequent adjustments to the heating. In addition, the demand for heating is greater during peak hours such as mornings and evenings. Based on these factors, the system can preliminarily determine the appropriate range of the prediction frequency to meet user needs while avoiding excessive calculations and waste of resources.

[0070] After comprehensively analyzing historical data, climate conditions and building characteristics, a reasonable initial heat load demand forecast frequency interval is set. Usually, when the temperature is relatively stable, the forecast frequency can be set to once every 1 to 2 hours to reduce the system load. In special cases, such as the initial stage of heating or under extreme weather conditions, the forecast interval can be shortened to once every 30 minutes or 15 minutes. This interval setting ensures that the system can be flexibly adjusted in different situations, improves forecast efficiency, and provides a reference for the subsequent adjustment of the dynamic forecast frequency.

[0071] Deploy edge nodes in key areas of the building to collect temperature data from each edge node in real time;

[0072] Edge nodes can monitor multiple parameters, including indoor and outdoor temperatures, heating pipe temperatures, and user temperature control behaviors. These data are directly transmitted to the edge computing platform through the Internet of Things (IoT) devices, forming a comprehensive temperature monitoring network for the building. This link ensures that the system is always aware of the latest changes in the thermal environment, providing comprehensive data support for subsequent analysis and prediction.

[0073] The key areas in the building include public areas, user living spaces, equipment rooms, and pipeline nodes. The temperature of public areas such as stairwells, corridors, underground garages, and lobbies will affect the overall heat loss; user living spaces include rooms facing different directions, top and bottom floors, and due to different lighting and insulation conditions in different areas, the temperature changes are also different; equipment rooms such as heating stations and boiler rooms are used to monitor the temperature status of heating equipment operation to ensure the stability of the heating system; pipeline nodes include heating pipeline terminals, branch nodes, and heat exchange stations to monitor the loss and distribution of heat energy in the pipeline network. The temperature data of these key areas together constitute a complete monitoring network, providing data support for real-time prediction and precise heating.

[0074] Set a fixed monitoring time window (such as 30 minutes or 1 hour), extract key features of the temperature data collected within the monitoring time window, and perform anomaly analysis on the extracted features. Based on the anomaly analysis, call the pre-trained deep learning model to evaluate the temperature changes of all edge nodes in the building;

[0075] Key features are extracted from the collected temperature data, where the extracted key features include the diffusion speed and amplitude of the temperature difference between edge nodes in different regions, and the temperature symmetry between adjacent edge nodes. Within the monitoring time window, the diffusion speed and amplitude of the temperature difference between edge nodes in different regions, and the temperature symmetry between adjacent edge nodes are analyzed for abnormalities, and then temperature difference diffusion quantification values ​​and temperature asymmetry quantification values ​​are generated respectively. The temperature difference diffusion quantification values ​​reflect the propagation of temperature changes in the building and the severity of temperature fluctuations in local areas. The temperature asymmetry quantification values ​​reflect whether the heat distribution of the heating system between regions is uniform and whether there is local temperature imbalance caused by uneven heating.

[0076] After obtaining the temperature difference diffusion quantification value and temperature asymmetry quantification value generated by abnormal analysis of the extracted key features, the temperature difference diffusion quantification value and the temperature asymmetry quantification value are comprehensively analyzed through a pre-trained deep learning model to generate an overall temperature stability coefficient. The temperature changes of all edge nodes in the building are evaluated through the overall temperature stability coefficient.

[0077] A pre-trained deep learning model is a model that is trained using a large amount of historical data and simulation data before it is officially put into use, so that it has the ability to predict and judge. In the application of building heating systems, such models will be used to analyze characteristic data such as the quantization value of temperature difference diffusion and the quantization value of temperature asymmetry, and establish complex nonlinear mapping relationships by learning the patterns of previous temperature changes, abnormal conditions and system response effects. Commonly used models include time series models such as support vector machines (SVM), decision tree models, and long short-term memory networks (LSTM). They can process high-dimensional, multivariate data, identify the laws of temperature changes, and accurately generate the overall temperature stability coefficient. The model continuously optimizes its parameters and reduces errors during the training process, so that it can efficiently and accurately predict the temperature fluctuation trends in different regions in practical applications.

[0078] After training, the deep learning model can quickly conduct a comprehensive assessment of the temperature stability in the building when receiving the real-time temperature difference diffusion quantification value and temperature asymmetry quantification value. Through these feature data inputs, the model will identify whether the current building temperature change deviates from the normal range based on previous patterns, and generate an overall temperature stability coefficient. The overall temperature stability coefficient reflects whether the heating system is running smoothly. The trained model also has adaptive learning capabilities. With the continuous input of new data, the model can gradually optimize its analysis strategy to maintain the accuracy and timeliness of the prediction. Using pre-trained deep learning models in the heating process not only improves the intelligence level of the system, but also reduces the need for manual intervention, ensuring that the heating system can be flexibly adjusted according to actual conditions, providing residents with more stable and efficient heating services.

[0079] When the temperature difference between edge nodes in different areas of a building spreads rapidly and with large amplitude, it usually indicates that the temperature stability of all edge nodes in the building is poor. This phenomenon means that the heat is not transferred evenly between areas, which may be caused by multiple factors, such as sudden changes in outside temperature, differences in insulation performance between floors, lags in the regulation of the heating system, or unbalanced heat distribution in pipes. Large and rapid temperature difference spreads can lead to underheating in some areas and overheating in others, causing discomfort to residents and increasing energy waste.

[0080] The specific steps for generating the temperature difference diffusion quantization value after abnormal analysis of the diffusion speed and amplitude of the temperature difference between edge nodes in different regions within the monitoring time window are as follows:

[0081] Calculate the instantaneous temperature difference change rate of edge nodes in adjacent areas to capture the speed of heat diffusion. The calculation expression is:

[0082]

[0083] Where, ΔT i(t) represents the instantaneous temperature difference change rate of the i-th edge node, T i (t) represents the temperature of the i-th node at time t, Δt represents the time interval between two sampling time points, β is the time attenuation coefficient, which is used to control the effect of the time interval on the rate of change, usually 1<β<2, reflecting the sensitivity of temperature changes in a short time;

[0084] This step reveals preliminary features of heat diffusion, especially its propagation speed and the sensitivity of node responses, by calculating the temperature changes between adjacent time points.

[0085] The temperature differences of all adjacent areas are accumulated, and a weight coefficient is introduced to reflect the diffusion characteristics of each node to quantify the cumulative diffusion amplitude of temperature diffusion. The calculation expression of the cumulative diffusion amplitude is:

[0086]

[0087] Where A(t) represents the cumulative diffusion amplitude of all regions at time t, γ i represents the weight coefficient of node i, which is used to measure the diffusion influence of different nodes, N represents the total number of edge nodes in the building, and α is the nonlinear expansion index (usually 1<α<3), which is used to enhance the sensitivity to severe fluctuations;

[0088] In a heating system, the location and function of different nodes will have different effects on the propagation speed and distribution of temperature. For example, the end node of a pipe and the node in the middle of the pipe may have different contributions and sensitivities to heat diffusion. The introduction of weight coefficients allows the model to take these differences into account, thereby more accurately reflecting the actual impact of each node when calculating the quantitative value of temperature difference diffusion. Ultimately, these weight coefficients can help the system better identify which nodes need priority attention in temperature control and optimize the control strategy of the heating system;

[0089] This step captures the global magnitude of temperature diffusion, ensuring that the system can identify which areas experience the most dramatic temperature fluctuations, thereby improving the control accuracy of the heating system.

[0090] The instantaneous temperature difference change rate and the cumulative diffusion amplitude between nodes are combined to generate the temperature difference diffusion quantization value. The calculation expression of the temperature difference diffusion quantization value is:

[0091]

[0092] Where, DI index It represents the quantitative value of temperature difference diffusion, which is used to comprehensively reflect the propagation of temperature changes in the building;

[0093] This step combines the diffusion speed, amplitude and node weight to reflect the severity of temperature change in a nonlinear form. If the temperature difference diffusion quantization value is high, it means that the temperature diffusion in the building is severe and uneven, and the heating system needs to increase the response frequency.

[0094] It can be seen from the temperature difference diffusion quantization value that within the monitoring time window, the larger the performance value of the temperature difference diffusion quantization value generated after the abnormal analysis of the diffusion speed and amplitude of the temperature difference between edge nodes in different regions, it usually means that within the monitoring time window, the temperature difference diffusion speed and amplitude between edge nodes in different regions are large, indicating that the temperature fluctuation in the building is more drastic and the temperature stability is poor. Conversely, the smaller the performance value of the temperature difference diffusion quantization value, the slower the diffusion speed and smaller the amplitude of the temperature difference between regions, and the temperature tends to be more balanced and stable. Therefore, the temperature difference diffusion quantization value can be used as an effective measurement indicator to reflect the temperature changes of all edge nodes in the building. The larger the performance value, the higher the temperature instability, and the smaller the performance value, the more stable the temperature.

[0095] When heating a building, the worse the temperature symmetry between adjacent edge nodes in the building, the worse the temperature stability of all edge nodes in the building. Temperature symmetry reflects the balance of heating in adjacent areas and the coordination of heat distribution. When the symmetry is poor, it means that some areas may be underheated or overheated, resulting in drastic temperature fluctuations in local areas. Such imbalances will make it difficult to maintain a stable temperature in the entire building, increase the difficulty of adjusting the heating system, and reduce energy efficiency. It may also cause discomfort to residents and affect the heating experience. Therefore, the temperature symmetry of adjacent nodes is an important indicator for measuring the temperature stability of a building.

[0096] The specific steps of generating a temperature asymmetry quantization value after abnormal analysis of the temperature symmetry between adjacent edge nodes within the monitoring time window are as follows:

[0097] In order to capture the temperature difference between adjacent edge nodes, we first establish a temperature difference function D u，k (t), where D u，k (t) represents the temperature difference function between two adjacent edge nodes u and k at time t, which is used to quantify the temperature inconsistency between two adjacent edge nodes and measure the temperature difference between adjacent edge nodes u and k at time t. The expression of the temperature difference function is: u，k (t)=(T u (t)-T k (t)) 2 +p·|T u (t)-T k (t)|, where T u (t) and T k(t) represents the temperature of edge nodes u and k at time t, and the weight coefficient of the absolute value term p is used to balance the square term (T u (t)-T k (t)) 2 and the absolute value term |T u (k)-T k The relative impact of (t)|

[0098] This temperature difference function combines the square term and the absolute value, which increases the sensitivity to sharp fluctuations and subtle changes and prevents a single feature from affecting the overall evaluation.

[0099] Within a monitoring time window W (e.g., within 30 minutes), the temperature differences of all adjacent edge nodes are accumulated and calculated, and a time weighting method is introduced to ensure that more recent temperature changes have a greater impact on the results. The weighted cumulative asymmetry is calculated, and the calculation expression is:

[0100]

[0101] In the formula, C uk (W) represents the weighted cumulative asymmetry of edge nodes u and k within the monitoring time window W, and ω(t) represents the time weighting coefficient, which is defined as Where (β is the time decay coefficient, which is the same as the time decay coefficient above)t new is the latest time point in the monitoring time window, and W represents the length of the monitoring time window (such as 30 minutes or 1 hour);

[0102] This step ensures that the temperature change closer to the current time has a greater weight in the calculation of the final temperature asymmetry quantization value, thereby enhancing the ability to respond to sudden changes.

[0103] The results of all adjacent nodes in the building are normalized and the temperature asymmetry quantification value is generated to quantify the balance of the overall heating system. The generation expression of the temperature asymmetry quantification value is:

[0104]

[0105] In the formula, I asym It represents the quantitative value of temperature asymmetry, which is used to measure the uniformity of heat distribution in a building. P represents the set of all adjacent nodes in a building (N represents the total number of edge nodes in a building);

[0106] It can be seen from the temperature asymmetry quantization value that within the monitoring time window, the larger the performance value of the temperature asymmetry quantization value generated after the abnormal analysis of the temperature symmetry between adjacent edge nodes, the more significant the temperature difference between adjacent nodes, indicating that the heat distribution in the building is uneven and the temperature stability of the system is poor; this may cause local areas to be overheated or overcooled, increase the discomfort of residents, and reduce the heating efficiency. On the contrary, when the temperature asymmetry quantization value is small, it means that the temperature difference between adjacent nodes is small, the heating system runs smoothly, and the temperature distribution in each area is uniform, indicating that the temperature stability in the building is good.

[0107] The deep learning model is not specifically limited here, and can realize the quantization of temperature difference diffusion value DI index and temperature asymmetry quantification value I asym Perform comprehensive analysis to generate the overall temperature stability coefficient Overall Tx In order to realize the technical solution of the present invention, the present invention provides a specific implementation method; Overall temperature stability coefficient Overall Tx The generated calculation formula is: Overall Tx =m1*DI index -1 +m2*I asym -1 , where m1 and m2 are the temperature difference diffusion quantization values ​​DI index and temperature asymmetry quantification value I asym The preset proportional coefficient, and m1 and m2 are both greater than 0.

[0108] It can be seen from the overall temperature stability coefficient that within the monitoring time window, the greater the performance value of the temperature difference diffusion quantization value generated after the abnormal analysis of the diffusion speed and amplitude of the temperature difference between edge nodes in different regions, the greater the performance value of the temperature asymmetry quantization value generated after the abnormal analysis of the temperature symmetry between adjacent edge nodes. That is, the smaller the performance value of the overall temperature stability coefficient generated when the temperature changes of all edge nodes in the building are evaluated by the deep learning model within the monitoring time window, the worse the temperature stability of the temperature in the building, and vice versa.

[0109] According to the evaluation results of the deep learning model, the heating status of the building is divided into two categories: overall stable type and overall frequent fluctuation type;

[0110] The overall temperature stability coefficient generated by evaluating the temperature changes of all edge nodes in the building through the deep learning model within the monitoring time window is compared with the pre-set overall temperature stability coefficient reference threshold, and the heating status of the building is divided into two categories. The specific division process is as follows:

[0111] If the overall temperature stability coefficient is less than the overall temperature stability coefficient reference threshold, the heating status of the building in the monitoring time window is classified as the overall frequent fluctuation type, which indicates that the temperature in the building changes violently and frequently, which may be caused by sudden weather changes or resident behavior;

[0112] If the overall temperature stability coefficient is greater than or equal to the overall temperature stability coefficient reference threshold, the heating state of the building under the monitoring time window is classified as an overall stable type. The overall stable type indicates that the temperature change in the building is small and stable, and the heat load demand is relatively stable.

[0113] Through this division, the system can accurately determine the heating demand status of the building and provide a basis for further adjusting the forecast frequency.

[0114] If the deep learning model is evaluated as overall stable, the building heating process will continue to be predicted according to the set initial heat load demand prediction frequency;

[0115] For the case where the deep learning model is evaluated as overall stable, it means that the temperature changes in various areas within the building are small and stable, and the heating demand does not fluctuate significantly. In this case, the system does not need to frequently adjust the heating parameters or increase the prediction frequency, so it can continue to predict the heating process according to the set initial heat load demand prediction frequency. This means that the system will perform heat load demand analysis at established time intervals (such as once an hour) to ensure that the heating system saves computing resources and energy while meeting demand. This strategy can reduce the frequent start and stop of heating equipment when the temperature is stable, reduce equipment wear and maintenance costs, and improve the overall operating efficiency of the heating system.

[0116] This strategy ensures that when the temperature does not change much, the system can achieve accurate control of heating demand with the minimum computing resources, avoiding the waste of resources caused by frequent predictions. At the same time, maintaining the initial prediction frequency also avoids over-adjustment of heating equipment, which helps to extend the service life of the equipment and reduce maintenance costs.

[0117] When it is detected that the temperature in the building is fluctuating frequently as a whole, the frequency of heat load demand prediction is dynamically increased according to the temperature changes of all edge nodes, the heating response speed is improved, and the problems caused by excessive or insufficient heating are avoided;

[0118] According to the temperature changes of all edge nodes, the frequency of heat load demand prediction is dynamically increased. The specific steps are as follows:

[0119] The overall temperature stability coefficient generated by the deep learning model is compared with the overall temperature stability coefficient reference threshold, and the stability coefficient deviation is calculated to quantify the degree of temperature fluctuation. The calculation expression is: ΔS = Overall sef-Overall Tx , where ΔS represents the temperature stability coefficient deviation, which is used to measure the difference between the current temperature fluctuation and the reference standard. sef Indicates the preset temperature stability coefficient reference threshold, OVerall Tx Indicates the overall temperature stability coefficient calculated within the current monitoring time window;

[0120] According to the temperature stability coefficient deviation and the monitoring density of the edge nodes, the dynamic adjustment coefficient is calculated to guide the improvement of the prediction frequency. The calculation expression is:

[0121]

[0122] In the formula, α' represents the dynamic adjustment coefficient, which indicates the influence of the intensity of temperature fluctuation on the predicted frequency;

[0123] The new actual forecast frequency is calculated based on the initial heat load demand forecast frequency and the dynamic adjustment coefficient. The calculation expression is: new =f initial ·(1+β'·α'), where f new is the actual predicted frequency after adjustment, f initial is the initial heat load demand forecast frequency, β' is the response sensitivity coefficient, which controls the amplitude of the forecast frequency increase (adjusted according to the system strategy, such as between 0.1 and 0.5);

[0124] The new actual prediction frequency f new It will increase with the intensity of temperature fluctuations.

[0125] In order to avoid resource waste and system overload caused by too high prediction frequency, the upper limit of prediction frequency is set, and the adjusted actual prediction frequency is compared with the upper limit of prediction frequency to obtain the final application prediction frequency. The determination expression of application prediction frequency is: final =min(f new , f max ), where f final The frequency predicted for the final application is used for practical application, f max The highest prediction frequency allowed by the system, set according to hardware capabilities and computing resources;

[0126] This step ensures that the system does not exceed performance limits while increasing the frequency of predictions.

[0127] For example, the system may adjust the prediction frequency from once an hour to once every 15 minutes, predicting heating demand in real time at a higher frequency. This adjustment enables the heating system to respond quickly to drastic changes in temperature, avoiding unstable room temperature caused by heating lags, and ensuring the comfort and safety of residents, especially in cold weather.

[0128] The present invention effectively solves the problem that the fixed prediction frequency cannot respond in time when the temperature fluctuates frequently through an edge computing control method based on building heating energy-saving regulation, and realizes the intelligent management and precise control of the heating system. The scheme uses the temperature data collected in real time by the edge node to extract key features and perform abnormal analysis within the fixed monitoring time window. By calling the pre-trained deep learning model, the heating state of the building is divided into overall stable type and overall frequent fluctuation type. In the case of stable type with small temperature changes, the system maintains the initial prediction frequency to avoid over-calculation and energy waste; in the case of frequent fluctuation type, the system dynamically increases the prediction frequency and quickly adjusts the heating parameters to ensure that the indoor temperature is always maintained in the comfortable range, avoiding the cold discomfort caused by insufficient heating and the energy waste caused by excessive heating. Not only does it optimize the response speed and heating efficiency of the system, but it also reduces the frequent start and stop and excessive wear of the equipment, and increases the service life of the equipment. In addition, through this dynamic regulation, the heating system can adapt to complex climate changes and the randomness of residents' behavior, significantly improving the heating experience and comfort of residents, especially in cold weather to ensure the health and quality of life of users.

[0129] The present invention provides Figure 2 The edge computing control method based on building heating energy-saving regulation shown includes an initial prediction frequency setting module, a temperature data acquisition module, a feature extraction and abnormality analysis module, a heating state identification module, a stable prediction module, and a frequent fluctuation response module;

[0130] The initial prediction frequency setting module sets an initial heat load demand prediction frequency when the heating system starts to operate, which is used to establish a basic prediction mechanism;

[0131] The temperature data collection module deploys edge nodes in key areas of the building to collect temperature data from each edge node in real time;

[0132] The feature extraction and anomaly analysis module sets a fixed monitoring time window, extracts key features from the temperature data collected within the monitoring time window, and performs an anomaly analysis on the extracted features. Based on the anomaly analysis, the pre-trained deep learning model is called to evaluate the temperature changes of all edge nodes in the building.

[0133] The heating status recognition module divides the building heating status into two categories based on the evaluation results of the deep learning model: overall stable type and overall frequent fluctuation type;

[0134] The stable prediction module, when the deep learning model is evaluated as overall stable, continues to predict the building heating process according to the set initial heat load demand prediction frequency;

[0135] Frequent fluctuation response module: When it is detected that the temperature in the building is fluctuating frequently, it dynamically increases the frequency of heat load demand prediction according to the temperature changes of all edge nodes, improves the heating response speed, and avoids problems caused by excessive or insufficient heating;

[0136] The edge computing control method based on building heating energy-saving regulation provided by an embodiment of the present invention is realized by the above-mentioned edge computing control system based on building heating energy-saving regulation. The specific methods and processes of the edge computing control system based on building heating energy-saving regulation are detailed in the above-mentioned embodiment of the edge computing control method based on building heating energy-saving regulation, which will not be repeated here.

[0137] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0138] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center containing one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0139] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0140] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0141] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0142] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0143] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0144] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. The edge computing control method based on building heating energy-saving regulation is characterized by: The following steps are involved: When the heating system starts to operate, an initial heat load demand forecast frequency is first set to establish a basic forecast mechanism; Deploy edge nodes in key areas of the building to collect temperature data from each edge node in real time; Set a fixed monitoring time window, extract key features from the temperature data collected within the monitoring time window, and perform anomaly analysis on the extracted features. Based on the anomaly analysis, call the pre-trained deep learning model to evaluate the temperature changes of all edge nodes in the building. According to the evaluation results of the deep learning model, the heating status of the building is divided into two categories: overall stable type and overall frequent fluctuation type; If the deep learning model is evaluated as overall stable, the building heating process will continue to be predicted according to the set initial heat load demand prediction frequency; When it is detected that the temperature in the building fluctuates frequently as a whole, the frequency of heat load demand prediction is dynamically increased according to the temperature changes of all edge nodes, the heating response speed is improved, and problems caused by excessive or insufficient heating are avoided.

2. The edge computing control method based on building heating energy-saving regulation according to claim 1 is characterized in that: The steps to set the initial heat load demand forecast frequency are as follows: Before setting the initial heat load demand forecast frequency, analyze the building type, building use, historical heating data and regional climate conditions. By collecting heating data from past heating seasons, understand the changing pattern of heating demand. At the same time, combine meteorological data analysis to determine the impact of different temperature ranges, wind speeds, humidity and precipitation conditions on heat load. Evaluate the structural characteristics of the building and user needs to preliminarily determine the appropriate range of the prediction frequency, so as to avoid excessive calculation and waste of resources while meeting user needs; After comprehensive analysis of historical data, climate conditions and building characteristics, an initial heat load demand forecast frequency range is set.

3. The edge computing control method based on building heating energy-saving regulation according to claim 1 is characterized in that: Key features are extracted from the collected temperature data, where the extracted key features include the diffusion speed and amplitude of the temperature difference between edge nodes in different regions, and the temperature symmetry between adjacent edge nodes. Within the monitoring time window, the diffusion speed and amplitude of the temperature difference between edge nodes in different regions, and the temperature symmetry between adjacent edge nodes are analyzed for abnormalities, and then temperature difference diffusion quantification values ​​and temperature asymmetry quantification values ​​are generated respectively. The temperature difference diffusion quantification values ​​reflect the propagation of temperature changes in the building and the severity of temperature fluctuations in local areas. The temperature asymmetry quantification values ​​reflect whether the heat distribution of the heating system between regions is uniform and whether there is local temperature imbalance caused by uneven heating.

4. The edge computing control method based on building heating energy-saving regulation according to claim 3 is characterized in that: After obtaining the temperature difference diffusion quantification value and temperature asymmetry quantification value generated by abnormal analysis of the extracted key features, the temperature difference diffusion quantification value and the temperature asymmetry quantification value are comprehensively analyzed through a pre-trained deep learning model to generate an overall temperature stability coefficient. The temperature changes of all edge nodes in the building are evaluated through the overall temperature stability coefficient.

5. The edge computing control method based on building heating energy-saving regulation according to claim 3 is characterized in that: The specific steps for generating the temperature difference diffusion quantization value after abnormal analysis of the diffusion speed and amplitude of the temperature difference between edge nodes in different regions within the monitoring time window are as follows: Calculate the instantaneous temperature difference change rate of edge nodes in adjacent areas to capture the speed of heat diffusion. The calculation expression is: In the formula, ΔT i (t) represents the instantaneous temperature difference change rate of the i-th edge node, T i (t) represents the temperature of the i-th node at time t, Δt represents the time interval between two sampling time points, and β is the time attenuation coefficient, which is used to control the influence of the time interval on the rate of change; The temperature differences of all adjacent areas are accumulated, and a weight coefficient is introduced to reflect the diffusion characteristics of each node to quantify the cumulative diffusion amplitude of temperature diffusion. The calculation expression of the cumulative diffusion amplitude is: Where A(t) represents the cumulative diffusion amplitude of all regions at time t, γ i represents the weight coefficient of node i, which is used to measure the diffusion influence of different nodes, N represents the total number of edge nodes in the building, and β is the nonlinear expansion index, which is used to enhance the sensitivity to severe fluctuations; The instantaneous temperature difference change rate and the cumulative diffusion amplitude between nodes are combined to generate the temperature difference diffusion quantization value. The calculation expression of the temperature difference diffusion quantization value is: Where, DI index It represents the quantitative value of temperature difference diffusion, which is used to comprehensively reflect the propagation of temperature changes in the building.

6. The edge computing control method based on building heating energy-saving regulation according to claim 5 is characterized in that: The specific steps of generating a temperature asymmetry quantization value after abnormal analysis of the temperature symmetry between adjacent edge nodes within the monitoring time window are as follows: Establish a temperature difference function D u，k (t), where D u，k (t) represents the temperature difference function between two adjacent edge nodes u and k at time t, which is used to quantify the temperature inconsistency between two adjacent edge nodes and measure the temperature difference between adjacent edge nodes u and k at time t. The expression of the temperature difference function is: u，k (t)=(T u (t)-T k (t)) 2 +p·|T u (t)-T k (t)|, where T u (t) and T k (t) represents the temperature of edge nodes u and k at time t, and the weight coefficient of the absolute value term p is used to balance the square term (T u (t)-T k (t)) 2 and the absolute value term |T u (t)-T k (t) |The impact; In a monitoring time window W, the temperature differences of all adjacent edge nodes are accumulated and calculated, and the time weighted method is introduced to calculate the weighted cumulative asymmetry. The calculation expression is: In the formula, C uk (W) represents the weighted cumulative asymmetry of edge nodes u and k within the monitoring time window W, and ω(t) represents the time weighting coefficient, which is defined as Among them, t new is the latest time point in the monitoring time window, and W represents the length of the monitoring time window; The results of all adjacent nodes in the building are normalized and the temperature asymmetry quantification value is generated to quantify the balance of the overall heating system. The generation expression of the temperature asymmetry quantification value is: In the formula, I asym It represents the quantitative value of temperature asymmetry, which is used to measure the uniformity of heat distribution in a building. P represents the set of all adjacent nodes in the building.

7. The edge computing control method based on building heating energy-saving regulation according to claim 6 is characterized in that: The overall temperature stability coefficient generated by evaluating the temperature changes of all edge nodes in the building through the deep learning model within the monitoring time window is compared with the pre-set overall temperature stability coefficient reference threshold, and the heating status of the building is divided into two categories. The specific division process is as follows: If the overall temperature stability coefficient is less than the overall temperature stability coefficient reference threshold, the heating status of the building in the monitoring time window is classified as an overall frequent fluctuation type; If the overall temperature stability coefficient is greater than or equal to the overall temperature stability coefficient reference threshold, the heating status of the building within the monitoring time window is classified as an overall stable type.

8. The edge computing control method based on building heating energy-saving regulation according to claim 7 is characterized in that: According to the temperature changes of all edge nodes, the frequency of heat load demand prediction is dynamically increased. The specific steps are as follows: The overall temperature stability coefficient generated by the deep learning model is compared with the overall temperature stability coefficient reference threshold, and the stability coefficient deviation is calculated to quantify the degree of temperature fluctuation. The calculation expression is: ΔS = Overall sef -Overall Tx , where ΔS represents the temperature stability coefficient deviation, which is used to measure the difference between the current temperature fluctuation and the reference standard. sef Indicates the preset temperature stability coefficient reference threshold, OVerall Tx Indicates the overall temperature stability coefficient calculated within the current monitoring time window; According to the temperature stability coefficient deviation and the monitoring density of the edge nodes, the dynamic adjustment coefficient is calculated to guide the improvement of the prediction frequency. The calculation expression is: In the formula, α' represents the dynamic adjustment coefficient, which indicates the influence of the intensity of temperature fluctuation on the predicted frequency; The new actual forecast frequency is calculated based on the initial heat load demand forecast frequency and the dynamic adjustment coefficient. The calculation expression is: new =f initial ·(1+β'·α'), where f new is the actual predicted frequency after adjustment, f initial is the initial heat load demand prediction frequency, β' is the response sensitivity coefficient, which controls the increase of the prediction frequency; In order to avoid resource waste and system overload caused by too high prediction frequency, the upper limit of prediction frequency is set, and the adjusted actual prediction frequency is compared with the upper limit of prediction frequency to obtain the final application prediction frequency. The determination expression of application prediction frequency is: final =min(f new , f max ), where f final The frequency predicted for the final application is used for practical application, f max The maximum prediction frequency allowed by the system.

9. The edge computing control method based on building heating energy-saving regulation is implemented by the edge computing control system based on building heating energy-saving regulation according to any one of claims 1 to 8, characterized in that: It includes an initial prediction frequency setting module, a temperature data acquisition module, a feature extraction and abnormality analysis module, a heating status identification module, a stable prediction module, and a frequent fluctuation response module; The initial prediction frequency setting module sets an initial heat load demand prediction frequency when the heating system starts to operate, which is used to establish a basic prediction mechanism; The temperature data collection module deploys edge nodes in key areas of the building to collect temperature data from each edge node in real time; The feature extraction and anomaly analysis module sets a fixed monitoring time window, extracts key features from the temperature data collected within the monitoring time window, and performs an anomaly analysis on the extracted features. Based on the anomaly analysis, the pre-trained deep learning model is called to evaluate the temperature changes of all edge nodes in the building. The heating status recognition module divides the building heating status into two categories based on the evaluation results of the deep learning model: overall stable type and overall frequent fluctuation type; The stable prediction module, when the deep learning model is evaluated as overall stable, continues to predict the building heating process according to the set initial heat load demand prediction frequency; The frequent fluctuation response module dynamically increases the frequency of heat load demand prediction based on the temperature changes of all edge nodes when it detects that the temperature in the building is frequently fluctuating as a whole, thereby improving the heating response speed and avoiding problems caused by excessive or insufficient heating.

Images