Air conditioning cooling load prediction method and system

Abstract

The present application relates to the technical field of air conditioning cooling load prediction, and discloses a method and system for air conditioning cooling load prediction. The method comprises: collecting temperature, humidity, and CO2 data through an environmental sensor network and performing standardized processing; establishing a time slice matrix based on environmental parameters and personnel density to obtain user behavior characteristics; constructing a wall heat storage model in combination with environmental parameters to calculate instantaneous heat flow; establishing a spatial downscaling model based on heat flow and meteorological data; extracting time series features based on local meteorological and heat flow data to predict cooling load; and generating time-segmented cooling control instructions accordingly. By establishing a data processing and analysis process, the present application comprehensively considers factors such as building microclimate characteristics, user behavior patterns, and dynamic thermal characteristics of walls to achieve a more accurate prediction of air conditioning cooling load.

Description

Technical Field

[0001] The present application relates to the technical field of air conditioning cooling load prediction, and in particular to an air conditioning cooling load prediction method and system. Background Art

[0002] With increasing demands for building energy management, intelligent control of air conditioning systems has become a crucial tool for energy conservation and emission reduction. Current air conditioning cooling load forecasting methods primarily rely on historical building energy consumption data and weather forecasts, employing statistical analysis or machine learning algorithms. These methods analyze load variations in historical data, combine it with weather forecast information, and develop forecasting models to estimate cooling load demand for future periods, providing a basis for decision-making in air conditioning system operation and control.

[0003] However, existing air conditioning cooling load forecasting methods suffer from insufficient accuracy. This is primarily due to the inadequacy of these methods in considering the impact of building microclimate characteristics and user behavior patterns on cooling loads. Traditional forecasting methods directly utilize regional meteorological data, ignoring the specificities of the local meteorological environment surrounding the building. Furthermore, they lack in-depth analysis of building occupant activity patterns, resulting in significant discrepancies between forecast results and actual cooling load demand. Furthermore, existing methods inadequately consider the dynamic thermal characteristics of building walls, failing to accurately reflect the impact of building heat storage and thermal inertia on cooling loads. Summary of the Invention

[0004] The present application provides an air conditioning cooling load prediction method and system, which is used to achieve a more accurate prediction of the air conditioning cooling load by establishing a data processing and analysis process, taking into account factors such as the building microclimate characteristics, user behavior patterns, and dynamic thermal characteristics of the wall, thereby providing a reliable control basis for the energy-saving operation of the air conditioning system.

[0005] In the first aspect, the present application provides an air conditioning cooling load prediction method, which includes: collecting raw data of indoor and outdoor temperature, humidity, and CO2 concentration through an environmental sensor network, removing outliers and performing sliding window smoothing processing on the raw data to obtain standardized environmental parameters; establishing a time slice distribution matrix based on the standardized environmental parameters and personnel density data, and obtaining user behavior characteristics through hierarchical clustering calculation; constructing a wall heat storage model based on user behavior characteristics and standardized environmental parameters, and obtaining instantaneous heat flow data of the building through dynamic thermal resistance calculation; establishing a spatial downscaling model based on the instantaneous heat flow data of the building and meteorological forecast data, and obtaining local meteorological parameters through microclimate correction; constructing a prediction feature matrix based on local meteorological parameters and the instantaneous heat flow data of the building, and obtaining a cooling load prediction value through time series feature extraction and attention weight calculation; calculating the cooling power adjustment coefficient of the air-conditioning system based on the cooling load prediction value, and generating a time-divided cooling capacity control instruction.

[0006] In a second aspect, the present application provides an air conditioning cooling load prediction system, the air conditioning cooling load prediction system comprising:

[0007] The acquisition module is used to collect raw data of indoor and outdoor temperature, humidity, and CO2 concentration through the environmental sensor network, remove outliers and perform sliding window smoothing on the raw data to obtain standardized environmental parameters;

[0008] The calculation module is used to establish a time slice distribution matrix based on standardized environmental parameters and personnel density data, and obtain user behavior characteristics through hierarchical clustering calculation;

[0009] A construction module is used to build a wall thermal storage model based on user behavior characteristics and standardized environmental parameters, and obtain the instantaneous heat flow data of the building through dynamic thermal resistance calculation;

[0010] The correction module is used to establish a spatial downscaling model based on the instantaneous heat flow data of the building and the meteorological forecast data, and obtain local meteorological parameters through microclimate correction;

[0011] The prediction module is used to construct a prediction feature matrix based on local meteorological parameters and instantaneous heat flow data of the building, and obtain the cooling load prediction value through time series feature extraction and attention weight calculation;

[0012] The generating module is used to calculate the cooling power adjustment coefficient of the air-conditioning system according to the cooling load prediction value, and generate a time-divided cooling capacity control instruction.

[0013] A third aspect of the present invention provides a computer device comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned air conditioning cooling load prediction method.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, enable the computer to execute the above-mentioned air-conditioning cooling load prediction method.

[0015] In the technical solution provided by this application, raw data on indoor and outdoor temperature, humidity, and CO2 concentration are collected through an environmental sensor network, and outlier removal and sliding window smoothing are performed, thereby improving the quality and reliability of environmental data and providing an accurate data foundation for subsequent predictions. A time slice distribution matrix is ​​established based on standardized environmental parameters and occupant density data, and user behavior characteristics are obtained through hierarchical clustering calculations, enabling accurate identification of building usage patterns. A wall thermal storage model is constructed based on user behavior characteristics and standardized environmental parameters, and instantaneous building heat flow data is obtained through dynamic thermal resistance calculation, accurately reflecting the dynamic heat transfer characteristics of the building wall. A spatial downscaling model is established using building instantaneous heat flow data and meteorological forecast data, and local meteorological parameters are obtained through microclimate correction, solving the problem that regional meteorological data cannot accurately reflect the characteristics of the building's surrounding environment. A prediction feature matrix is ​​constructed using local meteorological parameters and building instantaneous heat flow data, and cooling load prediction values ​​are obtained through time series feature extraction and attention weight calculation. This fully utilizes the advantages of deep learning algorithms in time series data processing and improves prediction accuracy. The cooling power adjustment coefficient of the air-conditioning system is calculated based on the cooling load prediction value and time-based cooling capacity control instructions are generated, achieving precise control of the air-conditioning system. In the specific field of air conditioning load forecasting, this solution introduces algorithms such as hierarchical clustering and deep learning, and combines them with the physical characteristics and usage characteristics of buildings to establish a data processing and analysis process, significantly improving forecast accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0017] Figure 1 This is a schematic diagram of an embodiment of the air conditioning cooling load prediction method in the embodiment of the present application;

[0018] Figure 2 Schematic diagram of the process of obtaining the cooling load prediction value by extracting time series features and calculating attention weights in an embodiment of the present application;

[0019] Figure 3 A schematic diagram of load levels in an embodiment of the present application;

[0020] Figure 4 This is a schematic diagram of an embodiment of an air conditioning cooling load prediction system in an embodiment of the present application;

[0021] Figure 5 It is a schematic block diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The embodiments of the present application provide a method and system for predicting cooling load of an air conditioner. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or devices.

[0023] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In one embodiment of the present application, an air conditioning cooling load prediction method includes:

[0024] Step S101: Collect raw data of indoor and outdoor temperature, humidity, and CO2 concentration through an environmental sensor network, remove outliers and perform sliding window smoothing on the raw data to obtain standardized environmental parameters;

[0025] Step S102: Establish a time slice distribution matrix based on standardized environmental parameters and personnel density data, and obtain user behavior characteristics through hierarchical clustering calculation;

[0026] Step S103: constructing a wall thermal storage model based on user behavior characteristics and standardized environmental parameters, and obtaining instantaneous heat flow data of the building through dynamic thermal resistance calculation;

[0027] Step S104: establishing a spatial downscaling model based on the instantaneous heat flow data of the building and the weather forecast data, and obtaining local meteorological parameters through microclimate correction;

[0028] Step S105: construct a prediction feature matrix based on local meteorological parameters and instantaneous heat flow data of the building, and obtain a cooling load prediction value through time series feature extraction and attention weight calculation;

[0029] Step S106: Calculate the cooling power adjustment coefficient of the air-conditioning system according to the cooling load prediction value, and generate a time-divided cooling capacity control instruction.

[0030] It is understandable that the execution subject of the present application may be an air conditioning cooling load prediction system, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0031] Specifically, an environmental sensor network collects indoor and outdoor environmental data at regular intervals. Temperature, humidity, and CO2 concentration sensors are deployed indoors and outdoors, collecting data every minute. The collected raw data is then processed for outliers by calculating the interquartile ranges for temperature, humidity, and CO2 concentration, and removing outliers that fall outside these ranges. This outlier-removed data is then subjected to a sliding average over a 5-minute window to reduce the impact of data fluctuations, resulting in standardized environmental parameter data. A time-slice distribution matrix is ​​then constructed based on the standardized environmental parameters and occupancy density data. A 24-hour period is divided into 48 time slices of 30 minutes each, and statistical analysis is performed on the environmental parameters and occupancy density data within each time slice. Hierarchical clustering is used to categorize the time-slice data into different user behavior scenarios, such as work hours and rest periods, to generate user behavior profiles. These profiles include information such as the rate of change in occupancy density and the changing trends of environmental parameters.

[0032] A wall thermal storage model is constructed based on user behavior characteristics and standardized environmental parameters. The wall thermal storage model is a mathematical model that describes the thermal conductivity characteristics of building walls. By dividing the building wall into multiple heat transfer nodes, the thermal conductivity coefficient between the nodes is calculated, and a node heat transfer network is established. The wall temperature field distribution is calculated based on the heat flow influencing factor, and the heat capacity change is calculated in combination with the wall material characteristics to obtain dynamic thermal resistance data. The instantaneous heat flow data of the building can be obtained through dynamic thermal resistance calculation. Then, a spatial downscaling model is established based on the instantaneous heat flow data of the building and the meteorological forecast data. The spatial downscaling model converts regional-scale meteorological forecast data into a spatial mapping relationship matrix of micro-meteorological data around the building. By gridding the space around the building, analyzing the correlation between heat flow data and meteorological data, establishing microclimate influence weights, and obtaining local meteorological parameters.

[0033] A prediction feature matrix is ​​constructed based on local meteorological parameters and the building's instantaneous heat flow data. The prediction feature matrix includes characteristic parameters such as the temperature change rate, heat flow change rate, and pressure gradient. Time series feature extraction is used to analyze the temporal variation of characteristic parameters, calculate feature importance indicators, and perform attention weighting to obtain a cooling load forecast. The cooling power adjustment coefficient of the air conditioning system is calculated based on the cooling load forecast. The cooling load forecast is divided into multiple time periods over a 24-hour period. The cooling load for each time period is graded, and the corresponding cooling power baseline value is calculated. Dynamic compensation is used to determine the cooling power adjustment coefficient, which is then used to generate cooling capacity control instructions for each time period.

[0034] For example, an office building uses this cooling load prediction method for air conditioning control. The raw data collected by the environmental sensor network contains an abnormal temperature value of 35°C due to a sensor failure (normal room temperature is between 20°C and 26°C). This outlier is removed using interquartile range calculation, and the remaining data is processed using a 5-minute moving average to obtain standardized environmental parameters. After dividing a 24-hour period into 48 time slices, analysis reveals that the period between 9:00 AM and 6:00 PM has a high occupancy density and significant temperature and humidity fluctuations, clustering this period as the workday. Based on these user behavior characteristics and the thermal conductivity characteristics of the multi-layer wall structure, a heat transfer network is established to calculate the dynamic thermal resistance changes during different time periods. A spatial downscaling model is used to convert regional weather forecast data into local meteorological parameters surrounding the building. Based on this data, a prediction matrix is ​​constructed, which includes features such as temperature gradients and heat flow changes. Time series analysis is used to obtain cooling load forecasts for each time period. Based on the predicted values, corresponding cooling control commands are generated, enabling intelligent adjustment of the air conditioning system.

[0035] In an embodiment of the present application, raw data of indoor and outdoor temperature, humidity, and CO2 concentration are collected through an environmental sensor network, and outliers are removed and a sliding window smoothing process is performed, thereby improving the quality and reliability of the environmental data and providing an accurate data basis for subsequent predictions. A time slice distribution matrix is ​​established based on standardized environmental parameters and personnel density data, and user behavior characteristics are obtained through hierarchical clustering calculations, thereby achieving accurate identification of building usage patterns. A wall heat storage model is constructed based on user behavior characteristics and standardized environmental parameters, and instantaneous heat flow data of the building is obtained through dynamic thermal resistance calculation, accurately reflecting the dynamic heat transfer characteristics of the building wall. A spatial downscaling model is established using the instantaneous heat flow data of the building and meteorological forecast data, and local meteorological parameters are obtained through microclimate correction, solving the problem that regional meteorological data cannot accurately reflect the characteristics of the surrounding environment of the building. A prediction feature matrix is ​​constructed using local meteorological parameters and instantaneous heat flow data of the building, and the cooling load prediction value is obtained through time series feature extraction and attention weight calculation, which fully utilizes the advantages of deep learning algorithms in time series data processing and improves prediction accuracy. The cooling power adjustment coefficient of the air-conditioning system is calculated based on the cooling load prediction value and a time-divided cooling capacity control instruction is generated, thereby achieving precise control of the air-conditioning system. In the specific field of air conditioning load forecasting, this solution introduces algorithms such as hierarchical clustering and deep learning, and combines them with the physical characteristics and usage characteristics of buildings to establish a data processing and analysis process, significantly improving forecast accuracy.

[0036] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0037] (1) Within the preset 24-hour time interval, each hour is divided into 60 sampling moments. The indoor and outdoor temperature, humidity, and CO2 concentration data are collected at each sampling moment through the environmental sensor network to generate the original environmental sequence;

[0038] (2) The original environmental sequence is segmented on an hourly basis, and the upper and lower quartile values ​​of the temperature, humidity, and CO2 concentration data in each time period are calculated. Abnormal data points that exceed the interval range are eliminated to obtain the abnormal value processing sequence;

[0039] (3) The temperature, humidity, and CO2 concentration data of each time period in the outlier processing sequence are normalized according to the maximum and minimum value differences to obtain the normalized environmental sequence;

[0040] (4) The temperature, humidity, and CO2 concentration data in the normalized environmental series are grouped according to five-minute time windows, and the cumulative average values ​​of the environmental parameters in each group are calculated to generate a smoothed environmental series;

[0041] (5) The smoothed environmental sequence is reorganized according to the 24-hour time interval and converted into a time series environmental data matrix to obtain standardized environmental parameters.

[0042] Specifically, within a 24-hour timeframe, each hour is divided into 60 sampling moments, equivalent to collecting data every minute. The temperature, humidity, and CO2 concentration sensors in the environmental sensing network simultaneously collect data at each sampling moment. The raw data includes indoor and outdoor temperature values ​​(in °C), relative humidity values ​​(in %), and CO2 concentration values ​​(in ppm). This data is arranged in chronological order to form the raw environmental sequence.

[0043] The original environmental data series is segmented and processed hourly. For each hourly time period, the interquartile range (IQR) is calculated for the temperature, humidity, and CO2 concentration data. The IQR is calculated by sorting the data from smallest to largest, identifying the Q1 value at the 25th percentile, the Q2 value (median) at the 50th percentile, and the Q3 value at the 75th percentile. The upper limit is defined as Q3 + 1.5 IQR, and the lower limit is defined as Q1 - 1.5 IQR, where IQR is the interquartile range and is equal to Q3 - Q1. Data within each time period is screened, and data points outside the upper and lower limits are marked as outliers and removed, forming an outlier processing sequence.

[0044] Normalization is performed on the data for each time period in the outlier processing sequence. This normalization uses the maximum-minimum difference method to map data of different dimensions to the interval [0,1]. Specifically, the maximum and minimum values ​​of temperature, humidity, and CO2 concentration within each time period are determined, and then the original data are linearly transformed. The transformed data maintains the relative size relationships of the original data, facilitating subsequent processing and resulting in a normalized environmental sequence. A sliding time window method is used to smooth the normalized environmental sequence. A time window length of 5 minutes is set, and the window slides backward by 1 minute at a time. The arithmetic mean of the temperature, humidity, and CO2 concentration data within each time window is calculated. This smoothing process reduces random fluctuations in the data, highlights the changing trends of environmental parameters, and generates a smoothed environmental sequence.

[0045] The smoothed environmental series is reorganized into 24-hour time intervals. The temperature, humidity, and CO2 concentration data are aligned according to time to construct a three-dimensional data matrix, where each dimension represents an environmental parameter and each time point corresponds to a set of environmental parameter values.

[0046] For example, an environmental sensor network collects data every minute for 24 hours, generating a total of 1440 sets of raw data. For example, among the 60 temperature values ​​collected within a given hour, one temperature value is abnormally high at 35°C due to a sensor failure. By calculating the quartiles of the temperature data for that hour, we obtain Q1 = 23°C, Q2 = 24°C, Q3 = 25°C, an IQR = 2°C, an upper bound of 28°C, and a lower bound of 20°C. The value of 35°C clearly exceeds the upper bound and is therefore discarded. The remaining temperature data is normalized to identify the highest temperature of 26°C and the lowest temperature of 22°C. These values ​​are then linearly transformed to the range [0, 1]. The normalized data is then averaged using a 5-minute sliding window, with each window containing five data points. The average is then calculated to produce a smoother temperature curve. The same process is performed on the humidity and CO2 concentration data. The three processed environmental parameters are reorganized in chronological order to form a standardized environmental parameter matrix, which serves as the basis for subsequent cooling load prediction. The original environmental sequence retains all the information collected by the sensors, outlier processing avoids the influence of erroneous data, normalization processing solves the problem of inconsistent dimensions of different parameters, sliding average eliminates the interference of short-term fluctuations, and data reorganization facilitates subsequent multi-parameter comprehensive analysis.

[0047] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0048] (1) Within a 24-hour time interval, the standardized environmental parameters are divided into 48 time slices of 30 minutes to generate an environmental time slice sequence;

[0049] (2) Statistically analyze the population density data according to the time slice sequence, calculate the cumulative value of the population density in each time slice, and obtain the population density distribution data;

[0050] (3) Time-align the environment time slice series with the personnel density distribution data, construct the environment-personnel joint data matrix, and obtain the time slice distribution matrix;

[0051] (4) Calculate the environmental change rate and the personnel density change rate for each time slice in the time slice distribution matrix to generate a time-varying feature sequence;

[0052] (5) The time-varying feature sequence is grouped into data according to weekdays and weekends, and the scene type identification is obtained through hierarchical division to generate user behavior characteristics.

[0053] Specifically, for the environmental parameters at time t, the mathematical expression is:

[0054]

[0055] in is the environmental parameter vector, is the normalized temperature value, is the standardized humidity value, is the standardized CO2 concentration value.

[0056] The calculation of population density distribution data is as follows:

[0057]

[0058] in is the personnel density value of the kth time slice, is the number of people at the i-th sampling point in the time slice, is the effective area of ​​the building, is the time slice length (30 minutes), and n is the number of sampling points in the time slice.

[0059] The expression of the environment-person joint data matrix is:

[0060]

[0061] in is the matrix element, For the i-th time slice at time The environmental parameter values, is the corresponding personnel density value.

[0062] The calculation formulas for the environmental change rate and the population density change rate are:

[0063]

[0064]

[0065] in is the environmental change rate of the kth time slice, is the rate of change of population density.

[0066] Scene type feature vector Defined as:

[0067]

[0068] in The working day identifier of the time slice (working day is 1, rest day is 0).

[0069] For example, in the time slice of 8:00-8:30 in the morning, the temperature change rate is 0.4℃ / 30min, and the population density change rate is 20 people / ( By calculating the eigenvectors of 48 time slices and combining data grouping for weekdays and weekends, we generate a user behavior profile. This profile reflects the dynamic changes in environmental parameters and human activity during building use.

[0070] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0071] (1) The indoor and outdoor temperature difference in the standardized environmental parameters and the change rate of the population density in the user behavior characteristics are combined in time series to generate the heat flow influencing factor;

[0072] (2) The heat flow influencing factors are mapped hierarchically according to the building structure to generate multi-layer heat transfer nodes. The inter-node heat conduction coefficient is calculated based on the time series data of the multi-layer heat transfer nodes to construct a node heat transfer network.

[0073] (3) Correlate the node heat transfer network with the building material parameters to generate a wall heat storage model. Calculate the temperature distribution gradient of each layer of the building wall based on the wall heat storage model and the heat flow influencing factor to obtain the wall temperature field data.

[0074] (4) The wall temperature field data is divided into time steps based on the wall thermal storage model, and the heat capacity change of the wall material in each time step is calculated to obtain the thermal storage characteristic parameters;

[0075] (5) The time-varying coefficients of the heat storage characteristic parameters are calculated in combination with the wall heat storage model, and the thermal resistance base value is calculated in combination with the wall material thickness to obtain the initial thermal resistance sequence;

[0076] (6) Dynamically correct the initial thermal resistance sequence and the temperature field change rate to obtain the thermal resistance distribution sequence;

[0077] (7) The thermal resistance distribution sequence and the indoor and outdoor temperature difference are converted into heat flux density to obtain the instantaneous heat flux data of the building.

[0078] Specifically, the combined effects of indoor and outdoor temperature differences and human activities on heat flow are considered. The heat flow influencing factor is calculated using the formula:

[0079]

[0080] in is the heat flow influencing factor at time t, is the difference between indoor and outdoor temperature (℃), is the rate of change of population density ( h), and is the weight coefficient. The temperature factor and the human activity factor are weighted together to reflect their combined impact on wall heat transfer. Based on the heat flow influencing factor, the building wall is analyzed in layers. Heat transfer nodes are set for each layer of material, and the heat conduction coefficient between nodes is calculated:

[0081]

[0082] in is the heat transfer coefficient between the nodes in the i-th and j-th layers, is the thermal conductivity of the i-th layer material, is the heat transfer area, is the distance between nodes (m). It reflects the heat transfer capacity between different material layers.

[0083] According to the heat conduction characteristics between nodes, the wall temperature field distribution equation is established:

[0084]

[0085] in is the temperature field distribution at position x at time t (℃), is the heat flux density, is the thermal conductivity of the material. The temperature field distribution directly affects the heat storage characteristics of the wall.

[0086] Calculate the heat storage characteristics of the wall based on the temperature field distribution:

[0087]

[0088] in is the heat storage capacity at time t (J), is the material density, is the specific heat capacity, is the material volume, The thermal storage characteristic characterizes the ability of the wall to store and release heat.

[0089] Consider the thermal resistance of each layer of wall material:

[0090]

[0091] in is the thermal resistance value at time t, in units of , is the thickness of the i-th layer of material (m), is the thermal conductivity, in units of The thermal resistance value reflects the wall's ability to hinder heat transfer.

[0092] Dynamically correct thermal resistance according to temperature field changes:

[0093]

[0094] in is the corrected thermal resistance value, is the temperature change rate influence coefficient. Dynamic correction takes into account the effect of temperature change on thermal resistance. Calculate the instantaneous heat flow of the building:

[0095]

[0096] in is the heat flux density at time t, reflecting the real-time heat transfer status of the wall.

[0097] For example, a building's exterior wall consists of three layers (from outside to inside: insulation, masonry, and plaster). When the indoor and outdoor temperature difference is 5°C, heat flow influencing factor calculations reveal that human activity contributes approximately 20% of the heat flow. Temperature gradients at each layer node reveal that the insulation layer has the greatest temperature drop, accounting for 60% of the total temperature difference. Dynamic thermal resistance calculations indicate that when the indoor temperature changes rapidly, the equivalent thermal resistance of the wall increases by 15-20%, directly impacting the instantaneous heat flow calculation results.

[0098] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0099] (1) The instantaneous heat flow data of the building is divided into multiple regional data segments according to the building orientation to obtain the heat flow regional distribution;

[0100] (2) Decompose the temperature, humidity, and wind speed data in the weather forecast data into time series to obtain the weather change trend;

[0101] (3) Map the regional distribution of heat flow and meteorological change trends in time and space to generate regional impact factors;

[0102] (4) Construct a heat island intensity calculation matrix based on regional influencing factors and obtain the spatial correlation coefficient;

[0103] (5) Combining the spatial correlation coefficient with the building geometric parameters to generate a spatial downscaling model, where the spatial downscaling model converts regional scale meteorological forecast data into a spatial mapping relationship matrix of micro-meteorological data around the building;

[0104] (6) Scale the meteorological forecast data according to the spatial downscaling model to obtain the initial microclimate data;

[0105] (7) Correct the initial microclimate data based on the building shading effect to obtain the microclimate correction coefficient;

[0106] (8) Perform spatial weighted calculation on the microclimate correction coefficient and the initial microclimate data to generate local meteorological parameters.

[0107] Specifically, the microclimate correction process needs to process the heat flow data of each building orientation. The instantaneous heat flow of the building is divided into four zones according to the four directions of east, south, west, and north. The heat flow distribution of each zone can be expressed as:

[0108]

[0109] in For spatial points Heat flow value at (W / ), is the orientation weight coefficient, is the heat flux distribution function in each direction, reflecting the heat flux distribution characteristics of the building in different directions.

[0110] Weather forecast data is processed through time series decomposition using the following expression:

[0111]

[0112] in is the time series of meteorological parameters, is the trend item, is the periodic term, is a random term. Decomposition helps to identify the main characteristics of meteorological changes.

[0113] The spatial and temporal mapping relationship between the regional distribution of heat flow and the meteorological change trend is expressed by the regional impact factor:

[0114]

[0115] in is the regional impact factor, is the meteorological trend adjustment coefficient. It reflects the coupling relationship between local heat flow and meteorological changes. The heat island intensity calculation matrix is ​​constructed using:

[0116]

[0117] in is the heat island intensity matrix element, is the impact factor of the kth region, is the weight coefficient, is the spatial distance. It describes the spatial distribution of the urban heat island effect. The mathematical expression of the spatial downscaling model is:

[0118]

[0119] in is the downscaled local meteorological parameter, is the spatial basis function, M is the number of discrete points in the horizontal grid direction, and N is the number of discrete points in the vertical grid direction. These two parameters determine the accuracy and computational complexity of spatial downscaling.

[0120] The mapping relationship from regional scale to local scale is described. The shading effect correction of the initial microclimate data is adopted:

[0121]

[0122] in is the corrected microclimate data, is the occlusion weight, is the occlusion function. The calculation expression of local meteorological parameters is:

[0123]

[0124] is the local meteorological parameter, is the spatial weighting coefficient. Take an office building as an example: the solar radiation heat flux received by the south-facing exterior wall of the building during the summer noon period reaches 800W / , while the north-facing exterior wall is only 200W / Meteorological data decomposition shows that the daily temperature fluctuation in the region is 10°C, with obvious periodic characteristics. Spatial downscaling model analysis found that the actual temperature around buildings is 2-3°C lower than the regional meteorological forecast, which is mainly affected by building shading and greening. This refined microclimate analysis provides more accurate local meteorological parameters for air conditioning load forecasting. The microclimate correction process reflects the scale conversion law from regional meteorology to local meteorology, fully considering the impact of building orientation, heat island effect, and shading effect on the local meteorological environment.

[0125] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0126] (1) Perform time synchronization processing on local meteorological parameters and building instantaneous heat flow data to obtain time series combined data;

[0127] (2) Sliding split the time series data into 24-hour cycles to generate a time window sequence;

[0128] (3) Extracting the temperature change rate, heat flow change rate, and air pressure gradient from the time window sequence to construct a prediction feature matrix, wherein the prediction feature matrix is ​​a set of correlation data between the building's surrounding environment and the cooling load change;

[0129] (4) Calculate the correlation between each feature based on the predicted feature matrix and generate the feature importance index;

[0130] (5) Sort the feature importance indicators according to their relevance to obtain the attention weight coefficient;

[0131] (6) Perform weighted operations on the predicted feature matrix and attention weight coefficient of each time window to obtain a weighted feature sequence;

[0132] (7) The weighted feature sequence is mapped according to the historical cooling load data to generate the cooling load prediction value.

[0133] Specifically, if Figure 2 As shown, it is a flow chart of obtaining the cooling load prediction value by time series feature extraction and attention weight calculation in an embodiment of the present application, and time synchronization processing is performed on local meteorological parameters and instantaneous heat flow data of the building. The two types of data are aligned according to the same timestamp to form time series combination data containing meteorological information and heat flow information. Time synchronization ensures the consistency of data from different sources in the time dimension, and provides a basis for subsequent analysis. The time series combination data needs to be slidingly divided according to the basic cycle of 24 hours. The length of the sliding window is set to 24 hours, and it slides back 1 hour each time, so that each time window contains a daily cycle data. This segmentation method fully takes into account the periodicity of building usage characteristics, and each time window sequence contains daily change information.

[0134] The features extracted for each time window sequence include three key indicators: temperature change rate, heat flow change rate, and pressure gradient. The temperature change rate reflects the speed of change of the ambient temperature and is calculated by the temperature difference between adjacent moments. The heat flow change rate represents the dynamic characteristics of the building's heat transfer state and is obtained by the difference in heat flux density between adjacent moments. The pressure gradient describes the spatial distribution characteristics of the local air pressure field and is calculated by the spatial difference of air pressure values ​​at different locations. These features form a prediction feature matrix, which is a collection of data related to the building's surrounding environment and changes in cooling load. When calculating the correlation between features based on the prediction feature matrix, the Pearson correlation coefficient method is used. The correlation coefficient is calculated between each pair of features to form a correlation matrix. The larger the absolute value of the correlation coefficient, the more significant the impact of the feature on the cooling load prediction, thereby generating a feature importance index.

[0135] Feature importance indices are sorted in descending order of relevance to generate attention weight coefficients that reflect the contribution of each feature. Features with larger weight coefficients receive more attention during prediction. This data-driven weighting method adaptively adjusts the importance of different features. The predicted feature matrix is ​​weighted and added to the attention weight coefficients to generate a weighted feature sequence that incorporates feature importance. This weighting process emphasizes the influence of key features, suppresses interference from secondary features, and improves the effectiveness of feature representation.

[0136] The weighted feature sequence is mapped to historical cooling load data, establishing a mapping relationship from feature space to load space to generate cooling load forecasts. This mapping process takes into account the load variation patterns in historical data, making the forecast results more consistent with actual conditions.

[0137] Taking the air conditioning load forecasting of an office building as an example, the collected local meteorological data (including temperature, humidity, and air pressure) and the building's instantaneous heat flow data are time-aligned to form a time series composite data with one data point per minute. This data is then partitioned into sliding windows with a 24-hour period, with each window containing 1,440 data points. Features are extracted from these window data, such as the hourly temperature change rate, the heat flow variation along the building's orientation, and the spatial distribution of the surrounding air pressure field. Correlation analysis reveals that the temperature change rate has the highest correlation with cooling load variation, followed by the heat flow change rate, while the pressure gradient has a relatively small impact. Based on this correlation, feature weights are determined, with the temperature change rate receiving a higher attention weight. The weighted feature series is then mapped with historical cooling load data to obtain a cooling load forecast for the next 24 hours. This prediction process fully leverages information from multiple data sources. Through feature extraction, weight calculation, and mapping analysis, accurate building cooling load prediction is achieved.

[0138] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0139] (1) Divide the cooling load forecast value into multiple time periods according to the 24-hour time interval to obtain the time-sharing cooling load sequence;

[0140] (2) Discretize the cooling load data in the time-sharing cooling load sequence and generate cooling load level labels;

[0141] (3) Calculate the percentage of the rated cooling capacity of the air-conditioning system according to the cooling load level mark to obtain the cooling power reference value;

[0142] (4) Perform dynamic compensation calculation on the refrigeration power reference value to generate the refrigeration power adjustment coefficient;

[0143] (5) Match the cooling power adjustment coefficient with the operating status of the air conditioning system to obtain the cooling capacity adjustment parameter;

[0144] (6) The air-conditioning system is controlled in time periods according to the cooling capacity adjustment parameters, and cooling capacity control instructions are generated in time periods.

[0145] Specifically, the 24-hour cooling load forecast value is divided into multiple time periods. According to the usage characteristics of the building, 1 hour is used as the basic unit for division to form 24 continuous time periods. Each time period contains the predicted cooling load value within that hour. These values ​​are arranged in chronological order to form a time-sharing cooling load sequence. The time-sharing cooling load sequence needs to be discretized. The discretization process maps the continuous cooling load values ​​to preset level intervals, such as dividing the cooling load values ​​into three levels: low load, medium load, and high load. The specific division method is based on the statistical analysis of historical operating data to determine the threshold range of different levels. Through this discretization process, the cooling load level mark corresponding to each time period is generated.

[0146] When converting the cooling load level mark into the cooling capacity control parameter of the air conditioning system, it is necessary to calculate the percentage relative to the rated cooling capacity. Get the rated cooling capacity value of the air conditioning system, and then determine the corresponding cooling capacity percentage range according to different load levels. Figure 3 The figure shows a schematic diagram of the load levels in the embodiment of the present application. The air conditioning system load levels are divided into three levels: high, medium, and low. Each level corresponds to a different operating parameter configuration. At the high load level, the cooling capacity is 80%-100% of the rated value, the compressor runs at 50-60Hz, and the fan maintains high speed; at the medium load level, the cooling capacity is 50%-80% of the rated value, the compressor runs at 35-50Hz, and the fan maintains medium speed; at the low load level, the cooling capacity is controlled at 30%-50% of the rated value, the compressor runs at 20-35Hz, and the fan maintains low speed.

[0147] Low load levels correspond to 30%-50% of the rated cooling capacity, medium load levels correspond to 50%-80% of the rated cooling capacity, and high load levels correspond to 80%-100% of the rated cooling capacity. This percentage division yields a preliminary cooling power baseline value. The cooling power baseline value also requires dynamic compensation to account for the impact of the actual operating environment. The compensation calculation adjusts the baseline value based on factors such as outdoor temperature and the length of time the air conditioning system is running. When the outdoor temperature is high, the compensation factor is increased; when the system is operating continuously for a long time, the compensation factor is appropriately reduced. This dynamic compensation calculation generates a more accurate cooling power adjustment factor.

[0148] The cooling power adjustment coefficient must match the actual operating status of the air conditioning system. Operating status includes parameters such as the compressor operating frequency and fan speed, which correspond to cooling power. Through table lookup or interpolation, the cooling power adjustment coefficient is converted into specific equipment operating parameters to obtain the cooling capacity adjustment parameter. Time-based cooling capacity control instructions are generated based on the cooling capacity adjustment parameters. These control instructions contain specific control variables such as the compressor frequency setpoint and fan speed setpoint for each time period. These instructions are executed in chronological order, enabling time-based operation control of the air conditioning system.

[0149] Take the air conditioning control system in an office building as an example. A forecast indicates a gradual increase in cooling load between 9:00 AM and 11:00 AM on weekdays. This period is divided into hourly intervals, generating cooling load data for the 9:00-10:00 AM and 10:00-11:00 AM periods. Discretization reveals that 9:00-10:00 AM falls within the medium load category, corresponding to 65% of the rated cooling capacity; 10:00-11:00 AM falls within the high load category, corresponding to 85% of the rated cooling capacity. Considering the expected outdoor temperature increase from 28°C to 30°C during these two periods, the baseline cooling power value is compensated by increasing the adjustment coefficients by 5% and 8%, respectively. The generated control instructions set the compressor frequency to 45 Hz and the fan speed to medium speed during the 9:00-10:00 AM period; and increase the compressor frequency to 52 Hz and the fan speed to high speed during the 10:00-11:00 AM period. This prediction-based, time-based control approach fully accounts for the building's load fluctuations and equipment operating conditions, ensuring efficient operation of the air conditioning system.

[0150] The above describes the air conditioning cooling load prediction method in the embodiment of the present application. The following describes the air conditioning cooling load prediction system in the embodiment of the present application. Figure 4 In one embodiment of the present application, an air conditioning cooling load prediction system includes:

[0151] The acquisition module is used to collect raw data of indoor and outdoor temperature, humidity, and CO2 concentration through the environmental sensor network, remove outliers and perform sliding window smoothing on the raw data to obtain standardized environmental parameters;

[0152] The calculation module is used to establish a time slice distribution matrix based on standardized environmental parameters and personnel density data, and obtain user behavior characteristics through hierarchical clustering calculation;

[0153] A construction module is used to build a wall thermal storage model based on user behavior characteristics and standardized environmental parameters, and obtain the instantaneous heat flow data of the building through dynamic thermal resistance calculation;

[0154] The correction module is used to establish a spatial downscaling model based on the instantaneous heat flow data of the building and the meteorological forecast data, and obtain local meteorological parameters through microclimate correction;

[0155] The prediction module is used to construct a prediction feature matrix based on local meteorological parameters and instantaneous heat flow data of the building, and obtain the cooling load prediction value through time series feature extraction and attention weight calculation;

[0156] The generating module is used to calculate the cooling power adjustment coefficient of the air-conditioning system according to the cooling load prediction value, and generate a time-divided cooling capacity control instruction.

[0157] Through the collaborative efforts of these components, an environmental sensor network collects raw data on indoor and outdoor temperature, humidity, and CO2 concentration. Outlier removal and sliding window smoothing are performed, improving the quality and reliability of the environmental data and providing an accurate data foundation for subsequent predictions. A time-slice distribution matrix is ​​established based on standardized environmental parameters and occupancy density data. User behavior characteristics are derived through hierarchical clustering, enabling accurate identification of building usage patterns. A wall thermal storage model is constructed based on user behavior characteristics and standardized environmental parameters. Dynamic thermal resistance calculations are used to obtain instantaneous building heat flux data, accurately reflecting the dynamic heat transfer characteristics of the building walls. A spatial downscaling model is constructed using building instantaneous heat flux data and meteorological forecast data. Local meteorological parameters are derived through microclimate correction, addressing the problem that regional meteorological data cannot accurately reflect the characteristics of the building's surrounding environment. A prediction feature matrix is ​​constructed from local meteorological parameters and building instantaneous heat flux data. Cooling load predictions are derived through time series feature extraction and attention weight calculation, leveraging the advantages of deep learning algorithms in time series data processing to improve prediction accuracy. The cooling power adjustment coefficient of the air-conditioning system is calculated based on the cooling load predictions, generating time-based cooling capacity control commands, enabling precise control of the air-conditioning system. In the specific field of air conditioning load forecasting, this solution introduces algorithms such as hierarchical clustering and deep learning, and combines them with the physical characteristics and usage characteristics of buildings to establish a data processing and analysis process, significantly improving forecast accuracy.

[0158] Reference Figure 5 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 5 As shown. The computer device includes a processor, memory, display screen, input device, network interface and database connected via a system bus. The processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above method is implemented.

[0159] Those skilled in the art will understand that Figure 5 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0160] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the above-described method when executed by a processor. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0161] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware using a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media provided herein and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM.

[0162] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0163] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0164] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting air conditioning cooling load, characterized in that: The air conditioning cooling load prediction method comprises: The raw data of indoor and outdoor temperature, humidity, and CO2 concentration are collected through the environmental sensor network, and outliers are eliminated and the sliding window smoothing process is performed on the raw data to obtain standardized environmental parameters. A time slice distribution matrix is ​​established based on standardized environmental parameters and personnel density data, and user behavior characteristics are obtained through hierarchical clustering calculations; A wall thermal storage model is constructed based on user behavior characteristics and standardized environmental parameters, and the instantaneous heat flow data of the building is obtained through dynamic thermal resistance calculation, including: combining the indoor and outdoor temperature difference in the standardized environmental parameters and the population density change rate in the user behavior characteristics in a time series manner to generate a heat flow influencing factor; mapping the heat flow influencing factor in layers according to the building structure to generate multi-layer heat transfer nodes, and calculating the heat conduction coefficient between nodes based on the time series data of the multi-layer heat transfer nodes to construct a node heat transfer network; correlating the node heat transfer network with the building material parameters to generate a wall thermal storage model, and according to the wall thermal storage model and the thermal The flow impact factor is used to calculate the temperature distribution gradient of each layer of the building wall to obtain the wall temperature field data; the wall temperature field data is divided into time steps based on the wall thermal storage model, and the heat capacity change of the wall material within each time step is calculated to obtain the thermal storage characteristic parameters; the thermal storage characteristic parameters are combined with the wall thermal storage model to calculate the time-varying coefficient, and the thermal resistance base value is calculated based on the wall material thickness to obtain the initial thermal resistance sequence; the initial thermal resistance sequence is dynamically corrected with the temperature field change rate to obtain the thermal resistance distribution sequence; the thermal resistance distribution sequence is converted into heat flux density with the indoor and outdoor temperature difference to obtain the instantaneous heat flow data of the building; A spatial downscaling model is established based on the instantaneous heat flow data of the building and the meteorological forecast data, and local meteorological parameters are obtained through microclimate correction, including: dividing the instantaneous heat flow data of the building into multiple regional data segments according to the building orientation to obtain the regional distribution of heat flow; performing time series decomposition on the temperature, humidity and wind speed data in the meteorological forecast data to obtain the meteorological change trend; performing time-space correspondence mapping on the regional distribution of heat flow and the meteorological change trend to generate regional impact factors; constructing a heat island intensity calculation matrix based on the regional impact factors to obtain the spatial correlation coefficient; combining the spatial correlation coefficient with the building geometric parameters to generate a spatial downscaling model, wherein the spatial downscaling model converts the regional scale meteorological forecast data into a spatial mapping relationship matrix of the microscopic meteorological data around the building; performing scale conversion on the meteorological forecast data according to the spatial downscaling model to obtain initial microclimate data; performing correction calculation on the initial microclimate data based on the building shielding effect to obtain the microclimate correction coefficient; performing spatial weighted calculation on the microclimate correction coefficient and the initial microclimate data to generate local meteorological parameters; A prediction feature matrix is ​​constructed based on local meteorological parameters and instantaneous heat flow data of the building, and a cooling load prediction value is obtained through time series feature extraction and attention weight calculation, including: performing time synchronization processing on local meteorological parameters and instantaneous heat flow data of the building to obtain time series combination data; slidingly dividing the time series combination data into 24-hour cycles to generate a time window sequence; extracting the temperature change rate, heat flow change rate, and air pressure gradient from the time window sequence to construct a prediction feature matrix, wherein the prediction feature matrix is ​​a set of related data between the surrounding environment of the building and the cooling load change; calculating the correlation between each feature based on the prediction feature matrix to generate a feature importance index; sorting the feature importance index according to the size of the correlation to obtain an attention weight coefficient; performing a weighted operation on the prediction feature matrix and the attention weight coefficient of each time window to obtain a weighted feature sequence; feature mapping the weighted feature sequence according to historical cooling load data to generate a cooling load prediction value; The cooling power adjustment coefficient of the air-conditioning system is calculated according to the cooling load prediction value, and a time-division cooling capacity control instruction is generated.

2. The air conditioning cooling load prediction method according to claim 1, characterized in that: The raw data of indoor and outdoor temperature, humidity, and CO2 concentration are collected through the environmental sensor network, and outliers are eliminated and the sliding window smoothing process is performed on the raw data to obtain standardized environmental parameters, including: Within a preset 24-hour time interval, each hour is divided into 60 sampling moments. The environmental sensor network collects indoor and outdoor temperature, humidity, and CO2 concentration data at each sampling moment to generate an environmental original sequence. The original environmental sequence is segmented per hour, and the upper and lower quartile values ​​of the temperature, humidity, and CO2 concentration data in each time period are calculated. Abnormal data points that exceed the interval range are eliminated to obtain the outlier processing sequence; The temperature, humidity, and CO2 concentration data of each time period in the outlier processing sequence are normalized according to the maximum and minimum value differences to obtain the normalized environmental sequence; The temperature, humidity, and CO2 concentration data in the normalized environmental series are grouped according to five-minute time windows, and the cumulative average of the environmental parameters in each group is calculated to generate a smoothed environmental series; The smoothed environmental series is reorganized according to 24-hour time intervals and converted into a time series environmental data matrix to obtain standardized environmental parameters.

3. The air conditioning cooling load prediction method according to claim 1, characterized in that: The time slice distribution matrix is ​​established based on the standardized environmental parameters and personnel density data, and the user behavior characteristics are obtained through hierarchical clustering calculation, including: In a 24-hour time interval, the standardized environmental parameters are divided into 48 time slices of 30 minutes to generate an environmental time slice sequence; The personnel density data is statistically analyzed according to the time slice sequence, and the cumulative value of the personnel density in each time slice is calculated to obtain the personnel density distribution data; Align the environment time slice series with the personnel density distribution data, construct the environment-personnel joint data matrix, and obtain the time slice distribution matrix; Calculate the environmental change rate and the personnel density change rate for each time slice in the time slice distribution matrix to generate a time-varying feature sequence; The time-varying feature sequences are grouped into data according to weekdays and weekends, and the scene type identification is obtained through hierarchical division to generate user behavior features.

4. The air conditioning cooling load prediction method according to claim 1, characterized in that: The step of calculating the cooling power adjustment coefficient of the air-conditioning system according to the cooling load prediction value and generating a time-division cooling capacity control instruction includes: The cooling load forecast value is divided into multiple time periods according to the 24-hour time interval to obtain the time-sharing cooling load sequence; Discretize the cooling load data in the time-sharing cooling load sequence to generate cooling load level marks; Calculate the percentage of the rated cooling capacity of the air conditioning system according to the cooling load level mark to obtain the cooling power benchmark value; Perform dynamic compensation calculation on the refrigeration power reference value to generate the refrigeration power adjustment coefficient; Match the cooling power adjustment coefficient with the operating status of the air conditioning system to obtain the cooling capacity adjustment parameter; The air conditioning system is controlled in time periods according to the cooling capacity adjustment parameters, and time period cooling capacity control instructions are generated.

5. An air conditioning cooling load prediction system, used to implement the air conditioning cooling load prediction method according to any one of claims 1 to 4, characterized in that: The air conditioning cooling load prediction system includes: The acquisition module is used to collect raw data of indoor and outdoor temperature, humidity, and CO2 concentration through the environmental sensor network, remove outliers and perform sliding window smoothing on the raw data to obtain standardized environmental parameters; The calculation module is used to establish a time slice distribution matrix based on standardized environmental parameters and personnel density data, and obtain user behavior characteristics through hierarchical clustering calculation; The construction module is used to build a wall heat storage model based on user behavior characteristics and standardized environmental parameters, and obtain the instantaneous heat flow data of the building through dynamic thermal resistance calculation, including: combining the indoor and outdoor temperature difference in the standardized environmental parameters and the population density change rate in the user behavior characteristics in a time series to generate a heat flow influencing factor; mapping the heat flow influencing factor in layers according to the building structure to generate multi-layer heat transfer nodes, and calculating the heat conduction coefficient between nodes based on the time series data of the multi-layer heat transfer nodes to build a node heat transfer network; correlating the node heat transfer network with the building material parameters to generate a wall heat storage model, and according to the wall heat storage The model and heat flow influencing factors are used to calculate the temperature distribution gradient of each layer of the building wall to obtain the wall temperature field data. The wall temperature field data is divided into time steps based on the wall thermal storage model, and the heat capacity change of the wall material within each time step is calculated to obtain the thermal storage characteristic parameters. The thermal storage characteristic parameters are combined with the wall thermal storage model to calculate the time-varying coefficients, and the thermal resistance base value is calculated based on the wall material thickness to obtain the initial thermal resistance sequence. The initial thermal resistance sequence is dynamically corrected with the temperature field change rate to obtain the thermal resistance distribution sequence. The thermal resistance distribution sequence is converted into heat flux density with the indoor and outdoor temperature difference to obtain the instantaneous heat flow data of the building. The correction module is used to establish a spatial downscaling model based on the instantaneous heat flow data of the building and the weather forecast data, and obtain local meteorological parameters through microclimate correction, including: dividing the instantaneous heat flow data of the building into multiple regional data segments according to the building orientation to obtain the regional distribution of heat flow; performing time series decomposition on the temperature, humidity and wind speed data in the weather forecast data to obtain the meteorological change trend; performing time-space correspondence mapping on the regional distribution of heat flow and the meteorological change trend to generate regional impact factors; constructing a heat island intensity calculation matrix based on the regional impact factors to obtain the spatial correlation coefficient; combining the spatial correlation coefficient with the building geometric parameters to generate a spatial downscaling model, wherein the spatial downscaling model converts the regional scale weather forecast data into a spatial mapping relationship matrix of the microscopic weather data around the building; performing scale conversion on the weather forecast data according to the spatial downscaling model to obtain initial microclimate data; performing correction calculation on the initial microclimate data based on the building shielding effect to obtain the microclimate correction coefficient; performing spatial weighted calculation on the microclimate correction coefficient and the initial microclimate data to generate local meteorological parameters; The prediction module is used to construct a prediction feature matrix based on local meteorological parameters and instantaneous heat flow data of the building, and obtain a cooling load prediction value through time series feature extraction and attention weight calculation, including: performing time synchronization processing on local meteorological parameters and instantaneous heat flow data of the building to obtain time series combination data; slidingly dividing the time series combination data according to a 24-hour period to generate a time window sequence; extracting the temperature change rate, heat flow change rate, and air pressure gradient from the time window sequence to construct a prediction feature matrix, wherein the prediction feature matrix is ​​a set of associated data between the surrounding environment of the building and the cooling load change; calculating the correlation between each feature based on the prediction feature matrix to generate a feature importance index; sorting the feature importance index according to the size of the correlation to obtain an attention weight coefficient; performing a weighted operation on the prediction feature matrix and the attention weight coefficient of each time window to obtain a weighted feature sequence; feature mapping the weighted feature sequence according to historical cooling load data to generate a cooling load prediction value; The generating module is used to calculate the cooling power adjustment coefficient of the air-conditioning system according to the cooling load prediction value, and generate a time-divided cooling capacity control instruction.

6. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and is characterized in that when the processor executes the computer program, the air conditioning cooling load prediction method according to any one of claims 1 to 4 is implemented. 7 . A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is caused to execute the air-conditioning cooling load prediction method according to any one of claims 1 to 4.

Images