Air conditioner control method and device based on temperature field, equipment and medium

Abstract

The application discloses an air conditioner control method and device based on a temperature field, equipment and a medium. The method comprises the following steps: dividing a target space according to a preset standard grid to obtain grid position coordinates of a plurality of grids; acquiring temperature data and heat source data of each grid, and constructing a three-dimensional temperature field model according to the temperature data, the heat source data and the grid position coordinates; inputting the grid temperature output by the three-dimensional temperature field model into a pre-trained space-time convolution network model to obtain a predicted temperature corresponding to the grid; and acquiring current air supply parameters of an air conditioner, and adjusting the air supply parameters according to the predicted temperature. The application realizes intelligent dynamic adjustment of air supply parameters of the air conditioner by establishing a three-dimensional temperature field model and predicting temperature changes by using a space-time convolution network, effectively solves the problem of temperature control lag of a traditional air conditioner, and significantly improves temperature control accuracy and comfort.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of air conditioner technology, and in particular to an air conditioning control method, device, equipment and medium based on temperature field. Background Technology

[0002] Currently, most air conditioning systems on the market collect indoor temperature data at a single point using a temperature sensor. This data is then combined with preset control logic, such as PID control, to adjust operating parameters and achieve temperature regulation. Airflow strategies are mostly fixed, such as uniform airflow throughout the room or unidirectional airflow. However, this type of control has significant limitations: limited by the ability to collect temperature data at a single point or in a coarse area, it cannot accurately capture temperature differences between different areas of the room. Especially in rooms with high ceilings or localized heat sources (such as appliances), temperature detection is prone to delays, leading to sluggish air conditioning temperature control response. Simultaneously, the airflow direction and intensity cannot be dynamically adapted to specific temperature difference areas, easily resulting in insufficient airflow to areas requiring temperature control and excessive airflow to areas not requiring temperature control. This causes untimely temperature control, poor user comfort, and increased energy consumption due to ineffective operation, making it difficult to balance temperature control accuracy, comfort, and energy efficiency. In summary, existing air conditioning systems suffer from technical problems such as delayed temperature control response, poor comfort, and high energy consumption due to a lack of refined temperature difference sensing and matching capabilities. Summary of the Invention

[0003] This invention provides an air conditioning control method, device, equipment, and medium based on a temperature field, aiming to solve the problems of delayed temperature control response and poor comfort in existing air conditioning systems.

[0004] In a first aspect, embodiments of the present invention provide an air conditioning control method based on a temperature field, the method comprising:

[0005] The target space is divided into multiple grids based on a preset standard grid to obtain the grid position coordinates.

[0006] Acquire temperature data and heat source data for each grid, and construct a three-dimensional temperature field model based on the temperature data, the heat source data, and the grid position coordinates;

[0007] The temperature of each grid point output by the three-dimensional temperature field model is input into a pre-trained spatiotemporal convolutional network model to obtain the predicted temperature corresponding to the grid point.

[0008] Obtain the current air supply parameters of the air conditioner, and adjust the air supply parameters according to the predicted temperature.

[0009] Secondly, the present invention also provides an air conditioning control device based on a temperature field, including a unit for performing the above-described method.

[0010] Thirdly, embodiments of the present invention also provide a computer device, the computer device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the above-described method.

[0011] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the above-described method.

[0012] This invention provides an air conditioning control method, device, equipment, and medium based on a temperature field. The method includes: dividing a target space into multiple grids according to a preset standard grid; acquiring temperature data and heat source data for each grid, and constructing a three-dimensional temperature field model based on the temperature data, heat source data, and grid position coordinates; inputting the temperature of each grid output from the three-dimensional temperature field model into a pre-trained spatiotemporal convolutional network model for prediction to obtain the predicted temperature corresponding to the grid; acquiring the current air supply parameters of the air conditioner, and adjusting the air supply parameters according to the predicted temperature. This invention combines a three-dimensional temperature field model and a spatiotemporal convolutional network model for temperature prediction, and dynamically adjusts the air supply parameters based on the predicted temperature to achieve proactive prediction and advance control of the target space temperature, shortening the air conditioning temperature control response time, avoiding discomfort caused by temperature fluctuations, and balancing temperature control effectiveness with energy efficiency. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of an air conditioning control method based on a temperature field according to an embodiment of the present invention;

[0015] Figure 2 for Figure 1 A flowchart illustrating the sub-steps of step S120;

[0016] Figure 3 for Figure 1 A flowchart illustrating another sub-step of step S120;

[0017] Figure 4 for Figure 1 A flowchart illustrating the sub-steps of step S130;

[0018] Figure 5 for Figure 1A flowchart illustrating the sub-steps of step S140;

[0019] Figure 6 for Figure 1 A flowchart illustrating another sub-step of step S140;

[0020] Figure 7 This is a flowchart illustrating the feedback optimization steps of the air conditioning control method based on temperature field according to an embodiment of the present invention.

[0021] Figure 8 A schematic block diagram of an air conditioning control device based on a temperature field provided in an embodiment of the present invention;

[0022] Figure 9 This is a schematic block diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0025] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0026] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0027] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrases "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0028] Traditional air conditioning control systems generally suffer from problems such as delayed temperature regulation and insufficient comfort. This is mainly because existing technologies rely primarily on temperature sensors at single locations, which cannot comprehensively perceive changes in the spatial temperature distribution, resulting in significant system response delays. When ambient temperature or heat source conditions change, air conditioners often require a considerable amount of time to react, causing large temperature fluctuations and increased energy consumption. This problem is particularly pronounced in large spaces or scenarios with multiple heat sources.

[0029] To address this, this invention proposes an air conditioning control method, device, equipment, and medium based on a temperature field. By dividing the target space into grids, constructing a three-dimensional temperature field model by combining grid temperature and heat source data, predicting the grid temperature using a pre-trained spatiotemporal convolutional network model, and then dynamically adjusting the air conditioning supply parameters based on the predicted temperature, the invention achieves proactive prediction and advance control of the target space temperature. This effectively solves the problem of temperature control lag in traditional air conditioning systems, improves temperature control accuracy and human comfort, and also considers energy efficiency. Details are as follows:

[0030] Please see Figure 1 , Figure 1 The flowchart of the air conditioning control method based on temperature field provided in the embodiment of the present invention is shown. The method includes steps S110-S140.

[0031] S110. Divide the target space into multiple grids according to a preset standard grid to obtain the grid position coordinates;

[0032] In this embodiment, the preset standard grid refers to a pre-defined, uniform three-dimensional grid unit used to divide the target space. Its size needs to be determined based on the size of the target space and the temperature control accuracy requirements. The target space is the area where the air conditioner needs to control the temperature, such as a bedroom, living room, or other enclosed or semi-enclosed space. The grid position coordinates refer to the specific coordinate values ​​of each grid unit in the three-dimensional coordinate system of the target space. This coordinate system can be established by setting a fixed point (such as a corner of the wall) in the target space as the origin, and includes the x-axis (horizontal length direction), y-axis (horizontal width direction), and z-axis (vertical height direction). Specifically, firstly, the actual three-dimensional dimensions of the target space, including length, width, and height, are measured by sensors or input by the user; then, according to the size of the preset standard grid (such as 1m×1m×0.5m), the target space is divided sequentially along the x-axis, y-axis, and z-axis directions, decomposing the entire target space into multiple continuous and non-overlapping grid units; finally, with the preset origin as the reference, the coordinate values ​​of the geometric center or vertex of each grid unit in the three-dimensional coordinate system are calculated to obtain the grid position coordinates corresponding to each grid. For a simple example, a standard room measuring 6m × 8m × 3m can be divided into 288 grid cells using a grid size of 1m × 1m × 0.5m. Alternatively, if the target space is a bedroom with a length of 4m, a width of 3m, and a height of 2.5m, and the default standard grid size is set to 1m × 1m × 0.5m, then 4 grids can be divided along the x-axis, 3 along the y-axis, and 5 along the z-axis, forming a total of 4 × 3 × 5 = 60 grid cells. Taking the bottom left corner of the bedroom floor as the origin, the grid cells located in the intervals of x = 1m - 2m, y = 0m - 1m, and z = 2.0m - 2.5m can be represented by the coordinates (1.5m, 0.5m, 2.25m). By refining the target space into multiple grids and determining the coordinates, it is possible to achieve a refined modeling of the temperature field of the target space. This lays the foundation for subsequent accurate collection of temperature data from various regions and the construction of a three-dimensional temperature field model that reflects the differences in spatial temperature distribution. It avoids the problem that traditional single sensors cannot capture local temperature differences, improves the spatial accuracy of temperature field modeling, and ensures the accuracy of temperature control and the comprehensiveness of spatial coverage. It effectively solves the problem of inaccurate temperature control caused by insufficient spatial sampling in traditional air conditioning systems.

[0033] S120. Obtain temperature data and heat source data for each grid, and construct a three-dimensional temperature field model based on the temperature data, the heat source data, and the grid position coordinates.

[0034] In this embodiment, the temperature data of the grid refers to the real-time temperature information corresponding to each grid cell in the target space, which is used to reflect the current temperature status of each local area; the heat source data refers to the relevant information of objects that generate heat in the target space, including the type, location and heat dissipation intensity of the heat source, such as the human body and electrical equipment, which are typical heat sources; the three-dimensional temperature field model refers to the model that can intuitively reflect the temperature distribution in the three-dimensional space by associating the temperature of each grid in the target space with the corresponding grid position coordinates through data modeling.

[0035] Specifically, if it is the first time the air conditioner is turned on, it is necessary to force the heating mode and run it at the highest fan speed for 10 minutes to quickly raise the overall temperature of the target space before turning off the air conditioner. During this process, the room temperature decay curve is recorded. Combined with the material properties of the target space's enclosure structure (such as walls and roof), the room heat transfer coefficient is calculated using the least squares method (this coefficient characterizes the material's ability to transfer heat, and can be verified by looking up a table, providing key basic parameters for subsequent temperature field modeling). Subsequently, the real-time temperature of each grid cell is collected one by one using a temperature acquisition device (such as a distributed temperature sensor) matched to the grid position coordinates, obtaining the corresponding temperature data. Then, the location information of heat sources and their heat dissipation characteristic parameters are obtained by using infrared thermal imaging equipment or position sensors, identifying all heat sources in the target space, recording the location information and heat dissipation-related parameters of each heat source, and forming heat source data. Then, using the grid position coordinates as the basis for spatial positioning, the temperature data and heat source data of each grid are associated and matched with the coordinate information of the corresponding grid to establish a correspondence between data and spatial location. Finally, based on this correspondence, a temperature field modeling algorithm (such as interpolation algorithm, numerical simulation algorithm, etc.) is used to construct a three-dimensional temperature field model that can reflect the temperature change law of each grid with spatial coordinates. This model can output the temperature value of each grid in the current state. If it is not the first power-on, there is no need to repeat the forced heating and heat transfer coefficient calculation steps. The historical heat transfer coefficient can be directly called, and the three-dimensional temperature field model can be updated by combining the real-time collected grid temperature data and heat source data to ensure that the model can quickly adapt to the current spatial temperature state. For example, if the coordinates of a grid in the target space are (1.5m, 0.5m, 2.25m), and the collected temperature data for that grid is 26℃, and there is a computer with a heat dissipation intensity of 100W (heat source data) 1m away, then during modeling, the "26℃ temperature data," "computer heat source information," and "coordinates (1.5m, 0.5m, 2.25m)" will be associated. This will then present the grid's temperature and the potential impact of surrounding heat sources on it at the corresponding location in the 3D model. By acquiring grid temperature and heat source data and combining them with coordinates to construct a 3D temperature field model, a refined and spatial representation of the temperature distribution in the target space is achieved. This provides a reliable model foundation for subsequent accurate prediction of temperature changes and adjustment of air conditioning supply parameters, avoiding the temperature distribution judgment bias caused by relying solely on single-point temperature data in traditional systems. This effectively solves the problem of insufficient control accuracy caused by incomplete temperature sensing in traditional air conditioning systems.

[0036] In one embodiment, such as Figure 2 As shown, step S120 includes: S121-S123.

[0037] S121. Calculate the global trend term temperature for each grid based on the heat source location coordinates and heat source heat dissipation parameters, wherein the heat source data includes the heat source location coordinates and the heat source heat dissipation parameters, and the global trend term temperature represents the overall temperature influence of all heat sources in the target space on the grid.

[0038] S122. Based on the temperature data, the global trend term temperature, and the grid position coordinates, calculate the local perturbation term temperature for each grid, wherein the local perturbation term temperature represents the local temperature deviation correction of a single heat source to its surrounding grids.

[0039] S123. Construct a three-dimensional temperature field model based on the global trend term temperature and the local perturbation term temperature to output the grid temperature of each grid, wherein the grid temperature is the superposition value of the global trend term temperature and the local perturbation term temperature of the corresponding grid.

[0040] In this embodiment, the global trend temperature (Tg) refers to the overall temperature impact value of each grid under the combined action of all heat sources in the target space (the comprehensive temperature impact of all heat sources on grid points), reflecting the temperature distribution trend at the macro level. The local perturbation temperature (Tl) refers to the local temperature deviation correction value generated by a single heat source on its surrounding nearby grids, used to refine the temperature differences at the micro level. The grid temperature (T0) is the superposition result of the global trend temperature and the local perturbation temperature, ultimately reflecting the actual temperature state of each grid and is also the final output temperature field data. The heat source location coordinates refer to the specific coordinate values ​​of the heat source in the three-dimensional coordinate system of the target space, such as (Xk, Yk, Zk). The heat source heat dissipation parameters are parameters characterizing the heat release capacity of the heat source, that is, the magnitude of the heat transfer capacity of the heat source, which includes the heat source coefficient Ak and the diffusion scale σk, where Ak characterizes the heat generation intensity of the heat source, and σk represents the diffusion range of the heat source temperature influence.

[0041] Specifically, the first step is to obtain complete heat source data for all heat sources within the target space: This involves identifying the locations of heat sources such as humans and electrical appliances using infrared thermal imaging equipment to obtain the heat source coordinates (Xk, Yk, Zk); and then determining the heat dissipation parameters of each heat source based on experimental data from the research and development phase. Subsequently, a pre-defined global trend term temperature calculation formula is used for calculation. The formula is as follows:

[0042]

[0043] Among them, T g(X,Y,Z) represents the global trend term temperature of the current computational grid, an output variable that directly reflects the overall temperature influence of all heat sources on the grid without local bias interference; (X,Y,Z) represents the grid position coordinates of the current computational grid, an input variable derived from the coordinate values ​​determined after dividing the target space into grids; K is the total number of heat sources in the target space, an input variable determined by the actual number of identified heat sources (e.g., K=2 if 1 person + 1 computer); A k Let A be the heat source coefficient of the k-th heat source (characterizing the heat dissipation intensity of the heat source; a larger value indicates stronger heat dissipation). This is a constant and needs to be obtained during the R&D phase through experimental data fitting using the least squares method. Specifically, this involves testing the impact of different types of heat sources (such as a 200W computer or a resting human body) on various temperature points in the room, collecting multiple sets of temperature data, and then using the least squares method to fit the A value corresponding to each heat source. k The values ​​are preset in the system; (Xk,Yk,Zk) are the coordinates of the k-th heat source location, which are input variables and are consistent with the heat source location identified by the infrared device; σ k σ is the diffusion radius of the k-th heat source (characterizing the effective radius of heat diffusion; a larger value indicates a wider heat diffusion range). It is a constant obtained through testing different heat sources during the R&D phase. For example, testing the temperature change of a 100W desk lamp at different distances determines its effective heat diffusion radius. k B is the linear trend constant coefficient (correcting for the linear change in target space temperature along the x, y, and z axes, such as the temperature difference between the ceiling and the floor of a room). It is a constant and was also obtained by fitting experimental data from the R&D phase using the least squares method. The fitting logic is the same as A. k Consistent; Lx, Ly, and Lz are the total lengths of the target space along the x, y, and z axes, respectively, and are constants determined by measuring the actual dimensions of the target space (e.g., Lx = 4m if the room length is 4m); x, y, and z have the same meaning as (X, Y, Z), all being the position coordinates of the current grid and are input variables. This formula obtains a base temperature value reflecting the global temperature distribution law by superimposing the overall heat dissipation effect of all heat sources on the grid and combining it with the linear change trend of the space temperature, avoiding overall temperature judgment deviations caused by missing heat sources or ignoring the linear temperature difference in space.

[0044] When calculating the local perturbation term temperature for each grid, it is necessary to base it on the global trend term temperature, grid location coordinates, and measured temperature data mentioned earlier: First, obtain the real-time measured temperature T of the current grid. k (i.e., temperature data), using the spatial distance formula: Calculate the distance d from the grid to each heat source. k Then, substituting the values ​​into the local disturbance term temperature calculation formula, the specific formula is as follows:

[0045]

[0046] Among them, T l (X,Y,Z) represents the local perturbation temperature of the current computational grid, which is an output variable used to correct the local temperature deviation of a single heat source on the grid; N is the total number of heat sources in the target space, consistent with the meaning of K in S121, and is an input variable; w k This is the coordinate correction coefficient for the k-th heat source (correction range for fine-tuning local deviations), a constant determined through experimental adjustments during the R&D phase. For example, it's used to adjust w when testing the temperature deviation of the grid surrounding a computer. k This ensures that the calculated local disturbance term accurately offsets the difference between the global trend term and the measured temperature; T k The real-time measured temperature value of the current grid is an input variable, derived from temperature data collected by distributed sensors; T g (xk, yk, zk) represents the global trend term temperature for the current grid, which is an intermediate output variable, specifically the T value for this grid calculated in S121. g (X,Y,Z) value; d k The distance from the current grid to the k-th heat source is an input variable, calculated from the position coordinates of the grid and the heat source. 'a' is the distance attenuation coefficient (characterizing the attenuation effect of distance on local temperature deviation; a larger value indicates a stronger attenuation of the deviation), a constant determined through experimental adjustments during the R&D phase. For example, when testing the attenuation of temperature deviations at different distances around a human body, 'a' is adjusted to ensure the formula accurately reflects the influence of distance on local deviations. This formula calculates the deviation between the measured temperature of the grid and the global trend temperature, and, combined with the distance attenuation effect, corrects for local temperature anomalies in the surrounding grid caused by a single heat source (such as excessively high temperatures in the grid near the heat source), making the temperature calculation more closely match the actual local temperature distribution in space.

[0047] When constructing a three-dimensional temperature field model and outputting the mesh temperature, it is necessary to base it on the global trend term temperature T of S121. g The temperature T of the local disturbance term of S122 l The final temperature of each grid cell is calculated using a grid temperature superposition formula, specifically:

[0048] T0 = ​​T g +T l

[0049] Where T0 is the current mesh temperature, which is an output variable, i.e., the final temperature value of each mesh output by the 3D temperature field model; T g Temperature, the global trend term for the current grid, is an input variable derived from the output of S121; T lThe temperature of the local perturbation term in the current grid is an input variable derived from the output of S122. This formula superimposes T_g, reflecting the global temperature trend, with T_l, which corrects for local biases, to obtain T0, which accurately characterizes the actual temperature of a single grid. Subsequently, T0 of all grids is associated with their corresponding grid position coordinates (X, Y, Z) to construct a three-dimensional temperature field model. This model can visualize the three-dimensional temperature distribution of the target space through heatmaps and other visual methods, and can directly output the grid temperature of any grid, providing accurate and comprehensive temperature data support for subsequent input of grid temperatures into a spatiotemporal convolutional network model for temperature prediction.

[0050] For example, taking an office with a target space of 6m×8m×3m (6m long along the x-axis, 8m wide along the y-axis, and 3m high along the z-axis, with the bottom left corner of the wall as the origin) as an example, there are two heat sources in this space: a computer with a heat dissipation power of 200W (heat source 1) and a person in a resting state (heat source 2). The space is divided into a grid of 1m×1m×0.5m (a total of 288 grid cells). Now, calculate the global trend term temperature, local disturbance term temperature, and grid temperature of the grid with coordinates (3m, 4m, 1.5m) (denoted as grid M).

[0051] The first step is to calculate the global trend term temperature T. g First, determine the formula parameters: K = 2 (2 heat sources); through experimental fitting during the R&D phase, for the computer (heat source 1), A1 = 15 (heat source coefficient, constant) and σ1 = 0.8 (diffusion scale, constant); for the human body (heat source 2), A2 = 10 (heat source coefficient, constant) and σ2 = 0.6 (diffusion scale, constant); B = 2 (linear trend constant coefficient, experimental fitting constant); Lx = 6m, Ly = 8m, Lz = 3m (total spatial length, measured constant); the position coordinates of heat source 1 (X1, Y1, Z1) = (2m, 3m, 1m) (infrared recognition input variable); the position coordinates of heat source 2 (X2, Y2, Z2) = (4m, 5m, 1.5m) (infrared recognition input variable); the coordinates of grid M (X, Y, Z) = (3m, 4m, 1.5m) (input variable). Substituting into the formula, the effect of heat source 1 on grid M is calculated as: 15 × [-((3-2)] 2 +(4-3) 2 +(1.5-1) 2 ) / (2×0.8)]=15×[-(1+1+0.25) / 1.6]=15×(-1.328)≈-19.92; Influence term of heat source 2 on grid M: 10×[-((3-4)] 2 +(4-5) 2 +(1.5-1.5) 2) / (2×0.6)]=10×[-(1+1+0) / 1.2]=10×(-1.667)≈-16.67; Linear trend term: 2×(3 / 6+4 / 8+1.5 / 3)=2×(0.5+0.5+0.5)=3; Final Tg=(-19.92)+(-16.67)+3≈-33.59 (output variable, this is a calculation example value, the actual value needs to be calibrated with temperature reference).

[0052] The second step is to calculate the temperature T of the local disturbance term. l : Determine the formula parameters: N = 2 (2 heat sources); through experiments, adjust w1 = 0.8 (heat source 1 coordinate correction coefficient, constant), w2 = 0.7 (heat source 2 coordinate correction coefficient, constant); a = 1.2 (distance attenuation coefficient, experimental adjustment constant); measured temperature of grid M T_k = 25℃ (sensor-acquired input variable); Tg(xk,yk,zk) = Tg≈-33.59 (input variable, global term output value); calculate the distance from grid M to the two heat sources: d1 = √[(3-2)] 2 +(4-3) 2 +(1.5-1) 2 ]≈1.58m (input variable), d2=√[(3-4) 2 +(4-5) 2 +(1.5-1.5) 2 ]≈1.41m (input variable). Substituting into the formula, the deviation correction term corresponding to heat source 1 is: 0.8×(25-(-33.59))×(1 / (1.58)). 1 · 2 ))≈0.8×58.59×0.57≈26.6; Deviation correction term corresponding to heat source 2: 0.7×(25-(-33.59))×(1 / (1.41) 1 · 2 ))≈0.7×58.59×0.62≈25.3; Finally, T_l=26.6+25.3≈51.9 (output variable).

[0053] Step 3: Calculate the grid temperature T0: Substitute into the formula T0 = T g +T l Therefore, T0≈-33.59+51.9≈18.31℃ (output variable, i.e., the final temperature of grid M, which conforms to the comfortable temperature range of the office).

[0054] Therefore, by using the global trend term temperature formula, the overall impact of all heat sources is combined with the spatial linear temperature trend, avoiding the global temperature judgment bias caused by traditional single heat source calculations or neglecting spatial temperature differences, thus providing an accurate overall temperature basis for temperature field modeling. The local perturbation term temperature formula corrects the local temperature anomalies of a single heat source on the surrounding grid by combining the deviation between the measured temperature and the global temperature with the distance attenuation effect, making up for the deficiency of the global trend term neglecting local details, and making the temperature calculation of each grid more consistent with the local temperature distribution of the actual space. Finally, the grid temperature superposition formula combines the global and local temperature terms, and the output temperature of each grid can accurately represent the three-dimensional temperature distribution of the target space. The constructed three-dimensional temperature field model can directly provide comprehensive and accurate input data for the temperature prediction of the subsequent spatiotemporal convolutional network, thereby providing a reliable basis for the adjustment of air conditioning supply parameters, and ultimately achieving refined and advance control of the target space temperature, reducing the discomfort caused by temperature fluctuations, and avoiding energy waste caused by inaccurate temperature judgment of the air conditioning.

[0055] In one embodiment, such as Figure 3 As shown, step S120 further includes: S124-S126.

[0056] S124. Calculate the real-time heat load of the target space according to the preset thermodynamic constraint equation;

[0057] S125. Determine whether a sudden change in heat source occurs in the target space based on the real-time heat load;

[0058] S126. If a heat source abruptly occurs in the target space, the temperature of the local disturbance term is corrected by a nonlinear compensation algorithm, and the grid temperature corresponding to the grid is updated based on the corrected local disturbance term temperature.

[0059] In this embodiment, the thermodynamic constraint equation is a preset mathematical equation used to calculate the heating and cooling load of the target space. This equation converts temperature parameters into quantified heat load values, providing an objective basis for judging sudden changes in heat sources. Real-time heat load refers to the heat exchange demand of the target space at the current moment due to temperature differences; its numerical changes directly reflect whether the heat source state is stable. Sudden heat source changes refer to the phenomenon where an unexpected heat source suddenly appears in the target space (such as the addition of a high-power electrical appliance or the rapid entry of multiple people) or the heat dissipation intensity of an existing heat source changes abruptly, causing significant fluctuations in heat load over a short period.

[0060] Specifically, when calculating the real-time heat load of the target space, it is necessary to call the preset thermodynamic constraint equation based on the parameters obtained in the process of constructing the three-dimensional temperature field model mentioned above. The equation is as follows:

[0061] Q = KF[(t d -t N )]

[0062] Where Q is the real-time heat load of the target space (usually in W or kW), an output variable that directly represents the current heat exchange demand of the space and is the core quantitative indicator for judging sudden changes in heat source; K is the heat transfer coefficient of the building materials used in the enclosure structure (such as walls, roof, and floor) of the target space (usually in W / (m²)). 2 The K-value (·℃) is a constant and can be obtained by consulting the standard table of heat transfer coefficients for building materials. The specific value needs to be determined based on the actual materials of the building envelope (such as concrete, glass, and insulation boards). For example, the K-value for ordinary concrete walls is approximately 1.5 W / (m²). 2 ·℃), pre-set in the system; F is the heat exchange area of ​​the target space enclosure structure (usually in m²). 2 The heat exchange area (F) is a constant, calculated by measuring the actual dimensions of each enclosure structure in the target space. For example, the heat exchange area of ​​a rectangular wall is the product of its length and height. If a south-facing wall in a room is 4m long and 2.8m high, then the F value for that wall is 11.2m. 2 The calculation is preset in the system; t d The temperature used to calculate the room's cooling (heating) load is a constant and a simplified representation of the outdoor temperature. It summarizes the overall impact of the outdoor environment (such as solar radiation and outdoor air temperature) on the indoor heat load. It can be obtained by consulting the "Room Cooling (Heating) Load Calculation Temperature Table." Its value is mainly related to the orientation of the room's walls (e.g., south-facing or north-facing) and time (e.g., daytime or nighttime). For example, the temperature for a south-facing wall in summer is... d The value is usually higher than that of the north-facing wall and is preset in the system; t N The mean temperature T0 of all grids in the target space is an input variable. We need to first calculate T0 for each grid in the output of the aforementioned 3D temperature field model, and then take the arithmetic mean of T0 for all grids to obtain t. N For example, if a room has 288 grids, and the sum of the temperatures T0 of all grids is 5184℃, then t N =5184℃ ÷ 288 = 18℃. This formula combines the characteristics of building materials (K, F) and the influence of the outdoor environment (t). d Compared with the average indoor temperature t N By combining this approach, the real-time heat load Q of the current target space is quantitatively calculated, avoiding the subjectivity of judging the heat source status solely based on temperature changes, and providing a precise quantitative basis for subsequent judgment of heat source abrupt changes. When determining whether a heat source abrupt change has occurred in the target space based on the real-time heat load, a "reference heat load Q without heat source abrupt changes" must first be determined. Then, the judgment of heat source abrupt changes is achieved by comparing the deviation between the real-time heat load Q and the reference Q to see if it exceeds a preset threshold. The calculation logic for the reference Q is consistent with that for Q, still using the formula Q... 基准 =KF[(t d -t N基准[], only the input variable t needs to be changed. N Replace with t N基准 , t N基准 The mean value of T0 for all grid cells is a constant when the target space is in a normal state without sudden changes in heat sources. This value is calculated by repeatedly collecting data under normal conditions (e.g., only one person resting in the room, with no additional electrical appliances on). N The average value is then taken, for example, t calculated three times consecutively under normal conditions. N The temperatures are 18.2℃, 17.8℃, and 18℃ respectively, then t N基准 =18℃, K, F, t d All parameters are consistent with those used in real-time heat load calculation (all are constants), and Q is the output variable, preset in the system as the judgment criterion.

[0063] The specific judgment process is as follows: First, calculate the absolute value of the deviation between the real-time heat load Q and the reference heat load Qreference |Qreference|. 基准 Then, the absolute value of this deviation is compared with the preset heat source mutation judgment threshold ΔQ (which is a constant, determined through experiments during the R&D phase; for example, for a room of 10-20㎡, ΔQ is usually set to 10%-15% of the Q baseline, i.e., if Q_baseline = 1000W, then ΔQ = 100-150W); if |QQ 基准 If ||≤ΔQ, it indicates that the current heat load is basically consistent with the normal state, and there is no sudden change in the heat source; if |QQ 基准 If |>ΔQ, then a sudden change in heat source has occurred in the target space, triggering the subsequent nonlinear compensation process. The specific temperature correction formula for the local disturbance term used in the nonlinear compensation algorithm is as follows:

[0064]

[0065] Among them, T h This is the correction increment for the temperature of the local disturbance term, which is an output variable used to add to the original temperature T of the local disturbance term. l Above, to achieve T l The correction; N is the number of heat sources that undergo sudden changes in the target space, which is an input variable and is determined by the number of newly added or suddenly changed heat sources identified by the system (e.g., if one electric heater is suddenly turned on, then N = 1); β k β is the exponential coefficient of the k-th abrupt heat source (characterizing the rate of temperature change caused by the abrupt heat source; the larger the value, the faster the temperature change), and is a constant. Its value is negatively correlated with the distance from the abrupt heat source to the corresponding grid (the closer the distance, the higher the β). k The larger the value, the more it is determined through testing experiments on different types of sudden heat sources (such as a 500W electric heater, or 5 people entering at the same time) during the research and development phase. For example, the grid β within 1m of the sudden heat source. k =0.8, grid spacing 2-3m β k=0.4, preset in the system; d k The distance from the current computational grid to the k-th abrupt heat source (usually in meters) is an input variable, obtained by calculating the spatial distance between the grid location coordinates and the abrupt heat source location coordinates; α is the distance decay exponent (characterizing the distance-to-correction increment T). h The attenuation effect increases with distance T as the value increases. h The stronger the weakening (α), the more constant it is, determined through experimental adjustments during the R&D phase. For example, for heat sources like electric heaters, α is usually set to 1.5, while for heat sources like the human body, α is usually set to 1.2; T l The original local disturbance term temperature (T before correction) l This refers to the input variable, specifically the temperature of the local perturbation term calculated in S122. Using this formula, based on the characteristics β of the abrupt heat source... k The distance d between the grid and the sudden heat source k and the original local disturbance term temperature T l Calculate the amount of T required l The incremental T for correction h This ensures that the corrected local perturbation temperature accurately reflects the impact of abrupt heat source changes on the local temperature of the surrounding mesh, laying the foundation for subsequent mesh temperature updates. By quantifying the real-time heat load of the target space using thermodynamic constraint equations, and combining this with a preset baseline heat load and deviation threshold, it can objectively and accurately determine whether a heat source change has occurred, avoiding the subjectivity and inaccuracy of traditional methods relying on temperature fluctuations. If a heat source change is determined to have occurred, the local perturbation temperature is corrected using a nonlinear compensation algorithm, and the mesh temperature is then updated based on the corrected local perturbation temperature. This allows for rapid adaptation to the impact of abrupt heat source changes on the space temperature, ensuring that the 3D temperature field model always accurately reflects the actual temperature distribution of the target space. This provides reliable data support for subsequent air conditioning supply parameter adjustments, effectively avoiding temperature control lag and temperature distribution imbalance caused by heat source changes, ensuring user comfort, and reducing additional energy consumption caused by temperature judgment errors in air conditioning.

[0066] S130. Input the temperature of each grid output by the three-dimensional temperature field model into a pre-trained spatiotemporal convolutional network model to predict the predicted temperature corresponding to the grid.

[0067] In this embodiment, the pre-trained Spatial-Temporal Convolutional Neural Network (STCN) model refers to a neural network model that has been trained in advance using historical grid temperature data, historical heat source data, and historical environmental parameters of the target space, and is capable of capturing the spatiotemporal variation patterns of temperature. The predicted temperature corresponding to each grid refers to the temperature value output by the model for each grid within a preset future time period, used to reflect the temperature change trend. First, all grid temperatures output by the three-dimensional temperature field model constructed above are obtained to ensure that each grid temperature is completely associated with its position coordinates. Then, these grid temperatures are used as input data and fed into the pre-trained Spatial-Temporal Convolutional Neural Network model. This model has been trained with a large amount of data and can identify the spatial distribution and temporal variation patterns of temperature. After processing the input current grid temperature, the model outputs the temperature of each grid within a preset future time period, which is the predicted temperature corresponding to each grid. The preset future time period can be set according to actual temperature control needs, typically 5 minutes, to match the response cycle of air conditioning air supply parameter adjustments. For a simple example, if the target space is an office measuring 6m × 8m × 3m, and it has been divided into a 3D grid of 1m × 1m × 0.5m (a total of 288 grid cells), the current temperature of each grid cell output by the 3D temperature field model is as follows: the grid cell with coordinates (3m, 4m, 1.5m) has a temperature of 25℃, and the grid cell with coordinates (4m, 5m, 1.5m) has a temperature of 25.5℃. After inputting the current temperatures of these 288 grid cells into a pre-trained spatiotemporal convolutional network model, the model, based on the historical data learned pattern that "the temperature rises slowly due to heat dissipation from the human body and electrical appliances in an office setting," outputs the predicted temperature of each grid cell within the next 5 minutes. The predicted temperature of the grid cell (3m, 4m, 1.5m) is 26.2℃, and the predicted temperature of the grid cell (4m, 5m, 1.5m) is 26.8℃. By inputting the precise grid temperature output from the three-dimensional temperature field model into a pre-trained spatiotemporal convolutional network model, the future temperature of each grid in the target space can be predicted in advance. This breaks through the limitation of traditional air conditioning relying solely on real-time temperature for regulation, and provides key data support for proactively adjusting the air supply parameters based on the predicted temperature. It avoids the lag in regulation after temperature fluctuations. At the same time, the pre-training characteristics of the model ensure prediction efficiency and accuracy, and can quickly adapt to the temperature change patterns of the target space.

[0068] In one embodiment, such as Figure 4 As shown, step S130 includes: S131-S134.

[0069] S131. Extract the spatiotemporal features of all grid temperatures at the current moment from the output of the three-dimensional temperature field model to obtain a grid temperature feature matrix that includes spatial location correlation features and time series change features.

[0070] S132. The grid temperature feature matrix is ​​concatenated with the historical grid temperature feature sequence within a preset time window to form a spatiotemporal input vector;

[0071] S133. Input the spatiotemporal input vector into the pre-trained spatiotemporal convolutional network model, capture the temperature correlation between different grids through the spatial convolutional layer of the model, and capture the temperature change trend of the same grid at different times through the temporal convolutional layer.

[0072] S134. The fully connected layer of the spatiotemporal convolutional network model outputs the temperature change sequence of each grid within a preset time period in the future, which is used as the predicted temperature of the grid.

[0073] In this embodiment, the current temperature of all grids refers to the real-time temperature of each grid at the same time node output by the 3D temperature field model. Spatiotemporal feature extraction refers to separating and extracting spatial features reflecting the positional correlation between different grids and temporal features reflecting temperature changes over time from the temperature data. Spatial positional correlation features refer to the temperature correlation patterns formed by the positional differences between different grids (e.g., adjacent grids have similar temperatures). Time series change features refer to the temperature change trend of the same grid at different times (e.g., the temperature of a certain grid gradually increases over time). The grid temperature feature matrix refers to the data structure formed by organizing the extracted spatiotemporal features in matrix form. The preset time window refers to the time range set in advance for selecting historical temperature data. The historical grid temperature feature sequence refers to the dataset formed by arranging the spatiotemporal features of each grid temperature in chronological order within the preset time window. The spatiotemporal input vector refers to the data that conforms to the model input format formed by concatenating the current grid temperature feature matrix and the historical grid temperature feature sequence. The spatial convolutional layer is a network layer in the spatiotemporal convolutional network model used to capture the temperature correlation between different grids. The temporal convolutional layer is a network layer in this model used to capture the temperature change trend of the same grid at different times. The fully connected layer is a network layer in the model used to integrate the features from the preceding data and output the prediction results. The temperature change sequence within a preset future time period refers to the set of temperature values ​​for each grid arranged at time intervals over a future period.

[0074] Specifically, firstly, spatiotemporal features are extracted from the current grid temperatures output by the 3D temperature field model. Through feature extraction algorithms, on the one hand, the positional coordinates of each grid are analyzed to uncover temperature correlations between adjacent grids and grids within the same region, forming spatial location correlation features; on the other hand, short-term temperature data prior to the current moment is combined to analyze the magnitude and rate of temperature change of the same grid over time, forming time-series variation features. These two types of features are then integrated into a grid temperature feature matrix. Secondly, a preset time window is determined (e.g., 10 minutes, divided into 2-minute intervals, containing 5 historical time nodes). The spatiotemporal temperature features of each grid within this time window are retrieved from the historical database stored in the system, arranged chronologically to form a historical grid temperature feature sequence. This sequence is then compared with the current grid temperature data. The temperature feature matrix is ​​concatenated to form a dimension-matched spatiotemporal input vector. Next, this spatiotemporal input vector is fed into a pre-trained spatiotemporal convolutional network model. The model first uses a spatial convolutional layer to process the spatial features in the input vector, capturing the temperature correlation between different grids (e.g., the temperature difference between grids near and far from the heat source). Then, it uses a temporal convolutional layer to process the temporal features, capturing the temperature change trend of the same grid at different times (e.g., a grid whose temperature rises by 0.2℃ every 2 minutes over the past 10 minutes). Finally, the fully connected layer of the model integrates the features output from the spatial and temporal convolutional layers, outputting a temperature change sequence for each grid within a preset future time period (e.g., 5 minutes, with 1-minute intervals). This sequence represents the predicted temperature for each grid. To illustrate with a simple example, let's consider an office with a grid size of 6m × 8m × 3m, divided into 288 grids of 1m × 1m × 0.5m each. First, we extract the spatiotemporal features of the temperature in all grids at the current time (e.g., 14:00). We find that the temperature difference between adjacent grids (e.g., (3m, 4m, 1.5m) and (3m, 5m, 1.5m)) is only 0.3℃ (spatial location correlation feature), and the temperature of the (3m, 4m, 1.5m) grid at 13:58 and 13:59 is 24.8℃ and 24.9℃ respectively (time series variation feature). These are integrated to form a grid temperature feature matrix. Next, we set a preset time window of 13:50-14:00 (every 2 minutes). The system uses a clock node (with a total of 5 historical nodes) to retrieve the spatiotemporal temperature features of each grid within that time period to form a historical sequence. This sequence is then concatenated with the current feature matrix to form a spatiotemporal input vector. After inputting this vector into the model, the spatial convolutional layer captures the correlation that "the (3m, 4m, 1.5m) grid is close to the computer's heat source and has a higher temperature than grids further away," while the temporal convolutional layer captures the trend that "the temperature of this grid increases by 0.2℃ every 2 minutes." Finally, the fully connected layer outputs the temperature change sequence of this grid for the next 5 minutes (14:01-14:05): 25.1℃, 25.3℃, 25.5℃, 25.7℃, and 25.8℃, thus completing the acquisition of the predicted temperature for this grid.The prediction process using a spatiotemporal convolutional network model first extracts the spatiotemporal features of temperature and constructs an input vector. Then, different levels of the model capture spatial correlations and temporal trends, ultimately outputting a sequence of future temperature changes. This further improves the accuracy and detail of the predicted temperature. Compared to prediction methods that only input the current temperature, this process more comprehensively considers the spatial distribution differences and temporal variation patterns of temperature, making the prediction results more closely match the actual temperature changes in the target space. Simultaneously, the output temperature change sequence provides a more detailed basis for subsequent adjustments to air supply parameters (such as fine-tuning the fan speed based on minute-by-minute temperature changes), effectively ensuring the timeliness and precision of air conditioning temperature control, further improving user comfort and reducing energy consumption.

[0075] S140. Obtain the current air supply parameters of the air conditioner, and adjust the air supply parameters according to the predicted temperature.

[0076] In this embodiment, the air supply parameters refer to the key parameters used by the air conditioner to regulate the temperature of the target space during current operation. These include the temperature difference fluctuation range (the allowable deviation range between the actual air supply temperature and the set temperature), wind speed (the air outlet speed of the air conditioner), air supply angle (the deflection angle of the air outlet blades), and coverage area (the spatial area that the air supply airflow can reach). The predicted temperature refers to the temperature change sequence of each grid in the target space within a preset time period, output by the pre-trained spatiotemporal convolutional network model mentioned above. It not only reflects the overall temperature trend but also accurately reflects the temperature differences caused by the location differences of different grids (such as grids near heat sources, grids where users are located, and corner grids), providing a basis for regional air supply adjustment.

[0077] Specifically, firstly, the air supply parameters of the current operation are retrieved through the air conditioner's control system, including the real-time temperature difference fluctuation range, wind speed, air supply angle, and coverage area, to ensure that the obtained parameters are consistent with the actual operating status of the air conditioner. The air supply parameters can be determined in different ways, which are not limited here. Secondly, combined with the predicted temperatures of each grid obtained above, not only is the overall temperature change trend of the target space analyzed, but also the predicted temperature differences of different functional grids (such as the core grid where users are active, the edge grid where there is no human activity, and the high-heat grid with heat sources) are analyzed in detail. For example, if the predicted temperature of the core grid (e.g., the grid with coordinates (3m, 4m, 1.5m)) where the user is located will rise from 25℃ to 27℃ (higher than the comfortable range of 24-26℃ in summer), while the predicted temperature of the edge grid (e.g., the grid with coordinates (1m, 1m, 0.5m)) will only rise to 25.5℃, then it is determined that the cooling effect needs to be enhanced in the area where the core grid is located. If the predicted temperature of the core grid where the user is located will drop from 21℃ to 19℃ (lower than the comfortable range of 20-22℃ in winter), while the predicted temperature of the grid near the heat source (e.g., the grid near the computer at (5m, 3m, 1m)) will remain at 20℃, then it is determined that the heating effect needs to be enhanced in the core grid area. Finally, based on the differences in predicted temperature between different grids and the user's... At the grid location, the current air supply parameters are adjusted accordingly: for example, if the predicted temperature of the core grid where the user is located is too high, in addition to reducing the overall temperature difference fluctuation range (e.g., from ±1℃ to ±0.5℃) and increasing the wind speed (e.g., from 1.2m / s to 1.8m / s), the air supply angle needs to be precisely adjusted to the direction of the core grid (e.g., from the original 30° to 45°) and the coverage area needs to be reduced to focus on the core grid area (e.g., from a radius of 2m to 1.5m) to avoid wasting airflow on the edge grids where the temperature has reached the standard; if the predicted temperature of the core grid where the user is located is too low, the air supply angle and coverage area are adjusted in the opposite direction to ensure that the airflow acts on the core grid first, while not causing overheating of grids with suitable temperatures near the heat source. By combining the precise predicted temperature of each grid with the user's spatial location to adjust the air conditioner's current air supply parameters, it breaks through the traditional air conditioner's rough control mode of "only based on overall temperature without grid differentiation" and the lagging mode of "adjusting after temperature deviation". It achieves precise and proactive control based on future grid-level temperature trends. Grid-level differentiated adjustment can effectively avoid the imbalance problem of "some grids being too hot / cold and some grids being suitable" in the target space, reduce user discomfort caused by local temperature fluctuations, and avoid airflow waste caused by the air conditioner's overall rough control, further reducing energy consumption and more accurately balancing overall comfort and energy saving.

[0078] In one embodiment, such as Figure 5 As shown, step S140 includes: S141-S142.

[0079] S141. Identify the activity status of the user within the target space, the activity status including sleep state, resting state and movement state;

[0080] S142. Determine the air supply parameters that match the identified activity state according to the preset mapping relationship between the activity state and the air supply parameters. The air supply parameters include the temperature difference fluctuation range, wind speed, air supply angle and coverage range.

[0081] In this embodiment, the user's activity state refers to the user's current level of physical activity within the target space, specifically including sleep state (the user is resting or sleeping, with a low metabolic rate and sensitivity to temperature and wind), resting state (the user is in a low-activity state such as sitting or reading, with stable body heat production), and exercise state (the user is in an active state such as walking or exercising, with increased body heat production and a higher demand for wind). The preset mapping relationship between activity states and air supply parameters refers to the rules for corresponding to the most suitable air supply parameters under different activity states, determined in advance through a large number of user experience experiments, and stored in the air conditioner control system in the form of a table or algorithm. The specific categories of air supply parameters include temperature difference fluctuation range, wind speed, air supply angle, and coverage area. The preset air supply parameter table can be as follows:

[0082]

[0083] Specifically, firstly, the user's activity state is identified using infrared thermal imaging equipment, motion sensors, or cameras (with image recognition algorithms) within the target space. For example, if an infrared thermal imaging device detects that a user has been still for a long time and is lying flat, it is determined to be a sleep state; if a camera detects that a user has been sitting for a long time and has low limb movement frequency, it is determined to be a resting state; if a motion sensor detects that a user is moving quickly and has a high heart rate, it is determined to be an active state. Secondly, the air supply parameters that match the identified activity state are retrieved from the system's stored mapping database. Finally, the retrieved air supply parameters are determined as the air supply parameters that the air conditioner should currently use. If the current parameters are inconsistent with the matching parameters, they are directly updated and adjusted. For a simple example, if the target space is a 15㎡ study, and the camera detects that the user is sitting at the desk reading with a limb movement frequency of less than 3 times per minute, it is determined to be in a resting state. The system's preset mapping relationship between the resting state and the air supply parameters is as follows: temperature difference fluctuation range 0℃ (actual temperature and set temperature have no deviation), wind speed 1.2m / s, air supply angle 20° (directional, to avoid direct blowing on the user), and coverage radius of 0.8m (focusing on the desk area). At this time, if the air conditioner's current wind speed is 0.8m / s (originally adapted to the sleep state), the wind speed will be adjusted to 1.2m / s according to the mapping relationship, and the other parameters will be adjusted according to the preset matching values ​​to ensure that the air supply is adapted to the user's physical needs in a resting state. By accurately identifying the user's activity state and then matching the air supply parameters according to the preset mapping relationship, personalized adaptation of "state-parameter" is achieved. This solves the problem of poor comfort caused by the fixed parameters of traditional air conditioners that adapt to all states. In the sleep state, low wind speed and wide-angle air supply can prevent the user from catching a cold. In the exercise state, high wind speed and small-area air supply can quickly dissipate heat. In the resting state, directional air supply can balance comfort and energy saving. At the same time, the mapping relationship is determined based on experimental data, which ensures the scientific nature of parameter matching and further improves user satisfaction.

[0084] In one embodiment, such as Figure 6 As shown, step S140 further includes: S143-S145.

[0085] S143. If there are multiple users in the target space, identify multiple attribute parameters of each user and obtain the air supply coefficient corresponding to each attribute parameter, wherein the attribute parameters include at least age and clothing parameters.

[0086] S144. Calculate the comprehensive air supply coefficient for each user based on the multiple air supply coefficients for each user.

[0087] S145. The user with the smallest overall air supply coefficient among all users is identified as the target user, and the air supply parameters that match the activity status of the target user are identified as the current air supply parameters.

[0088] In this embodiment, attribute parameters refer to personal characteristics that affect a user's perceived temperature, including at least age (different age groups have different sensitivities to temperature; for example, the elderly and children are more sensitive than adults) and clothing parameters (the thickness or type of clothing worn by the user, such as a thick coat, a thin coat, or short sleeves, which directly affects the body's heat dissipation efficiency). The airflow coefficient is a coefficient set based on the attribute parameters to quantify the differences in users' perceived temperature needs; the smaller the coefficient, the more sensitive the user is to temperature and airflow (and their comfort should be prioritized). The comprehensive airflow coefficient is a single coefficient calculated by using a preset algorithm (such as product or weighted sum) to comprehensively assess the user's sensitivity. The target user is the user with the smallest comprehensive airflow coefficient among multiple users, i.e., the user whose perceived temperature needs are most critically addressed. Specifically, firstly, if multiple users are detected in the target space by sensors or cameras, the attribute parameters of each user are identified one by one. The system uses image recognition algorithms to determine the user's clothing parameters (e.g., if the camera captures the user wearing short sleeves, it's determined to be thin clothing; if it captures the user wearing a down jacket, it's determined to be thick clothing). It also determines the user's age through facial feature recognition or user input (e.g., facial wrinkles indicate an elderly person, height and facial youthfulness indicate a child). Next, it retrieves the corresponding air supply coefficient for each attribute parameter from a pre-set coefficient table (e.g., thick coats correspond to 1.2, thin coats 1.0, short sleeves 0.87, elderly people 0.8, adults 1.0, children 0.7). Then, it calculates the comprehensive air supply coefficient for each user using a pre-set algorithm (usually a product, as age and clothing have an additive effect on body sensation). Finally, it compares the comprehensive air supply coefficients of all users and identifies the user with the lowest coefficient as the target user. Finally, it identifies the target user's activity state (according to the identification method in S141) and determines the air supply parameters matching that activity state (according to the mapping relationship in S142) as the current air supply parameters for the air conditioner. For a simple example, if the target space is a 25㎡ living room, and two users are detected: User A (an adult, wearing a light jacket) and User B (a child, wearing short sleeves), first, their attribute parameters are identified: User A's age is adult, and their clothing parameter is a light jacket; User B's age is child, and their clothing parameter is short sleeves. Next, the air supply coefficients are retrieved: adult coefficient 1.0, light jacket coefficient 1.0, User A's overall air supply coefficient = 1.0 × 1.0 = 1.0; child coefficient 0.7, short sleeves coefficient 0.87, User B's overall air supply coefficient = 0.7 × 0.87 ≈ 0.609. After comparison, User B's overall air supply coefficient is smaller, thus identifying them as the target user; then, it is determined that User B is in a resting state (sitting on the sofa watching TV), and the air supply parameters corresponding to the resting state are retrieved (temperature difference fluctuation range 0℃, wind speed 1.0m / s, air supply angle 25° directional, coverage area 1.0m radius), and these parameters are determined as the current air supply parameters, prioritizing User B's comfort.By identifying multi-user attribute parameters, calculating the comprehensive air supply coefficient, and determining the target user, the problem of "temperature zone conflict" in multi-user scenarios is solved, avoiding the discomfort of other users caused by prioritizing the adaptation of some users. At the same time, by taking the user with the lowest comprehensive air supply coefficient as the core adaptation target, the overall comfort can be balanced to the greatest extent (after the needs of sensitive users are met, the comfort of other users can usually also be met). In addition, by combining the target user's activity status with matching parameters, the accuracy of air supply is further ensured. Compared with the traditional average adaptation or random adaptation method, the overall user satisfaction in multi-user scenarios is greatly improved.

[0089] In one embodiment, such as Figure 7 As shown, the air conditioning control method based on temperature field in this embodiment of the invention further includes steps S151-S153.

[0090] S151. Record the air supply parameters after each adjustment, the grid temperature output by the three-dimensional temperature field model, the predicted temperature output by the spatiotemporal convolutional network model, the actual temperature data in the target space, and user feedback information, wherein the user feedback information includes the air supply parameters manually adjusted by the user.

[0091] S152. Calculate the deviation between the predicted temperature and the actual temperature data, and adjust the network parameters of the spatiotemporal convolutional network model according to the deviation value through a preset reinforcement learning algorithm.

[0092] S153. Based on the user feedback information, correct the correlation adjustment rules between the air supply parameters and the predicted temperature.

[0093] In this embodiment, the actual temperature data within the target space refers to the real-time temperature of each grid cell collected by distributed temperature sensors, used to verify the accuracy of the predicted temperature. User feedback information refers to information actively interacted by the user through the air conditioner control panel, mobile APP, etc., and its core includes manually adjusted airflow parameters (e.g., the user manually adjusts the fan speed from 1.2 m / s to 0.8 m / s), directly reflecting the user's comfort needs. The deviation between the predicted and actual temperature data refers to the difference (absolute value or positive / negative value) between the predicted and actual temperature data at the same grid cell and the same time node, used to quantify the model's prediction error. The preset reinforcement learning algorithm refers to a pre-set machine learning algorithm that can optimize model parameters through error feedback, possessing a learning logic of "the smaller the error, the higher the reward." The network parameters of the spatiotemporal convolutional network model refer to the core parameters in the model that affect the prediction results, such as the weights of the spatial convolutional layers and the biases of the temporal convolutional layers. The correlation adjustment rules between air supply parameters and predicted temperature refer to the logic of adjusting air supply parameters based on predicted temperature (e.g., if the predicted temperature increases by 1℃, the wind speed increases by 0.2m / s), which is used to guide the direction and magnitude of parameter adjustment.

[0094] Specifically, firstly, after each adjustment of the air conditioner's air supply parameters, the system automatically records multiple types of key data, including the complete air supply parameters after the adjustment (temperature difference fluctuation range, wind speed, air supply angle, coverage area), all grid temperatures output by the 3D temperature field model at the time of adjustment, the temperature predicted by the spatiotemporal convolutional network model previously used to support this adjustment (i.e., the future temperature prediction result corresponding to this adjustment), and the actual temperature data of the target space collected by distributed sensors within a preset time period after the adjustment (consistent with the prediction time period, usually 5 minutes). Simultaneously, it records user feedback information during this period, focusing on capturing manually adjusted air supply parameters by the user (e.g., if the user feels the wind speed is too strong and manually reduces it from 1.5 m / s to 1.0 m / s, the system records the parameter changes before and after this adjustment). Secondly, for the recorded predicted and actual temperature data, the system calculates the deviation value of each grid at the corresponding time node (generally taking the absolute value of the predicted temperature minus the actual temperature to avoid positive and negative cancellation), and then processes the deviation values ​​of all grids... The overall prediction error corresponding to this adjustment is obtained by summing the data (e.g., taking the average). This overall prediction error is then input into a preset reinforcement learning algorithm. The algorithm, aiming to "reduce prediction error," iteratively adjusts the network parameters of the spatiotemporal convolutional network model. If the deviation value is greater than a preset error threshold, the algorithm increases the correction magnitude of network parameters that significantly affect the prediction deviation (e.g., adjusting the weight of a certain type of grid temperature in the spatial convolutional layer). If the deviation value is less than the threshold, the parameters are slightly fine-tuned to maintain model stability. Simultaneously, the system analyzes manually adjusted parameters in user feedback information and corrects the correlation adjustment rules between air supply parameters and predicted temperature. For example, if the predicted temperature indicates that the wind speed needs to be increased to 1.5 m / s, but the user manually reduces it to 1.0 m / s, it indicates that the matching magnitude of "predicted temperature and wind speed" in the original correlation rule is too high. The system will correct this rule, adjusting the "wind speed increase corresponding to a 1°C increase in predicted temperature" from 0.2 m / s to 0.15 m / s, ensuring that subsequent adjustments better match the user's actual sensory needs.

[0095] For a simple example, if the target space is a 30㎡ bedroom, the air supply parameters after a certain adjustment are: temperature difference fluctuation range ±0.5℃, wind speed 0.8m / s, air supply angle 30°, and coverage area 1.2m; at the time of adjustment, the temperature of the (3m, 4m, 1.5m) grid output by the three-dimensional temperature field model is 25℃, and the spatiotemporal convolutional network model previously predicted that the temperature of this grid in the next 5 minutes would be 25.8℃; 5 minutes later, the actual temperature of this grid is 25.2℃, and the calculated deviation is |25.8-25.2|=0.6℃. Moreover, the user feels that the current wind speed is too weak and manually adjusts the wind speed to 1.0m / s. At this point, the system first records all the above data, and then inputs the deviation value of 0.6℃ into the reinforcement learning algorithm. The algorithm judges that the deviation value is slightly higher than the preset threshold of 0.5℃, so it fine-tunes the network parameters related to "bedroom area temperature prediction" in the spatiotemporal convolutional network model (such as reducing the weight of the temporal convolutional layer on the short-term temperature rise trend). At the same time, based on the user's feedback that the wind speed is manually increased, the original association rule of "wind speed increases by 0.1m / s for every 0.8℃ increase in predicted temperature" is corrected and adjusted to "wind speed increases by 0.1m / s for every 0.6℃ increase in predicted temperature" to ensure that the wind speed adjustment is more in line with the user's body feeling when the predicted temperature changes. By recording multi-dimensional data, optimizing model parameters based on prediction bias, and revising adjustment rules based on user feedback, a closed-loop learning mechanism of data recording, error optimization, and rule iteration is formed. On the one hand, it can continuously improve the temperature prediction accuracy of the spatiotemporal convolutional network model and reduce adjustment bias caused by inaccurate predictions. On the other hand, it can allow the air supply parameter adjustment rules to gradually adapt to users' personalized needs, avoiding the problem of algorithm predictions being out of sync with user experience. At the same time, this mechanism can achieve system self-evolution without manual intervention. After long-term use, it can make the air conditioner more accurate and more in line with user habits, ensuring comfort while further reducing ineffective energy consumption and improving the overall system's intelligence and practicality.

[0096] Figure 8 This is a schematic block diagram of an air conditioning control device 200 based on a temperature field, provided in an embodiment of the present invention. Figure 8 As shown, corresponding to the above-described temperature field-based air conditioning control method, the present invention also provides a temperature field-based air conditioning control device 200. This temperature field-based air conditioning control device 200 includes a unit for executing the above-described temperature field-based air conditioning control method, and the device can be configured in a computer device. Specifically, please refer to... Figure 8 The temperature field-based air conditioning control device 200 includes: a division unit 201, a construction unit 202, a prediction unit 203, and an adjustment unit 204.

[0097] The system includes a partitioning unit 201, which is used to partition the target space into multiple grids according to a preset standard grid to obtain the grid position coordinates; a construction unit 202, which is used to acquire the temperature data and heat source data of each grid, and construct a three-dimensional temperature field model based on the temperature data, the heat source data, and the grid position coordinates; a prediction unit 203, which is used to input the temperature of each grid output by the three-dimensional temperature field model into a pre-trained spatiotemporal convolutional network model for prediction to obtain the predicted temperature corresponding to the grid; and an adjustment unit 204, which is used to acquire the current air supply parameters of the air conditioner and adjust the air supply parameters according to the predicted temperature.

[0098] In one embodiment, the construction unit 202 is further configured to: calculate the global trend term temperature of each grid based on the heat source location coordinates and heat source heat dissipation parameters, wherein the heat source data includes the heat source location coordinates and the heat source heat dissipation parameters, and the global trend term temperature characterizes the overall temperature influence of all heat sources in the target space on the grid; calculate the local perturbation term temperature of each grid based on the temperature data, the global trend term temperature, and the grid location coordinates, wherein the local perturbation term temperature characterizes the local temperature deviation correction of a single heat source on its surrounding grids; and construct a three-dimensional temperature field model based on the global trend term temperature and the local perturbation term temperature to output the grid temperature of each grid, wherein the grid temperature is the superposition value of the global trend term temperature and the local perturbation term temperature of the corresponding grid.

[0099] In one embodiment, the construction unit 202 is further configured to: calculate the real-time heat load of the target space according to a preset thermodynamic constraint equation; determine whether a heat source abrupt change occurs in the target space based on the real-time heat load; if a heat source abrupt change occurs in the target space, correct the temperature of the local disturbance term using a nonlinear compensation algorithm, and update the mesh temperature corresponding to the mesh based on the corrected local disturbance term temperature.

[0100] In one embodiment, the prediction unit 203 is further configured to: extract spatiotemporal features from the current grid temperatures output by the three-dimensional temperature field model to obtain a grid temperature feature matrix containing spatial location correlation features and time series change features; concatenate the grid temperature feature matrix with historical grid temperature feature sequences within a preset time window to form a spatiotemporal input vector; input the spatiotemporal input vector into a pre-trained spatiotemporal convolutional network model, capture the temperature correlation between different grids through the spatial convolutional layer of the model, and capture the temperature change trend of the same grid at different times through the temporal convolutional layer; output the temperature change sequence of each grid within a preset future time period through the fully connected layer of the spatiotemporal convolutional network model, as the predicted temperature corresponding to the grid.

[0101] In one embodiment, the adjustment unit 204 is further configured to: identify the activity state of a user within the target space, the activity state including a sleep state, a resting state, and an exercise state; and determine air supply parameters matching the identified activity state according to a preset mapping relationship between the activity state and air supply parameters, the air supply parameters including temperature difference fluctuation range, wind speed, air supply angle, and coverage range.

[0102] In one embodiment, the adjustment unit 204 is further configured to: if there are multiple users in the target space, identify multiple attribute parameters of each user and obtain the air supply coefficient corresponding to each attribute parameter, wherein the attribute parameters include at least age and clothing parameters; calculate the comprehensive air supply coefficient of each user based on the multiple air supply coefficients of each user; determine the user with the smallest comprehensive air supply coefficient among all users as the target user, and determine the air supply parameter that matches the activity state of the target user as the current air supply parameter.

[0103] The aforementioned temperature field-based air conditioning control device 200 can be implemented as a computer program, which can be used in, for example... Figure 9 It runs on the computer device shown.

[0104] Please see Figure 9 , Figure 9 This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 500 may be a terminal.

[0105] See Figure 9 The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.

[0106] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions that, when executed, cause the processor 502 to perform an air conditioning control method based on a temperature field.

[0107] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0108] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute an air conditioning control method based on a temperature field.

[0109] This network interface 505 is used for network communication with other devices. Those skilled in the art will understand that... Figure 9The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0110] The processor 502 is used to run a computer program 5032 stored in a memory to implement the steps of the above method.

[0111] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0112] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0113] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein the computer program includes program instructions. When executed by a processor, the program instructions cause the processor to perform the steps of the above-described method.

[0114] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0115] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0116] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0117] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the device of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0118] 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 storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part 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 to cause a computer device to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0119] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0120] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Since these modifications and variations fall within the scope of the claims and their equivalents, this invention also intends to include these modifications and variations.

[0121] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An air conditioning control method based on a temperature field, characterized in that, The method includes: The target space is divided into multiple grids based on a preset standard grid to obtain the grid position coordinates. Acquire temperature data and heat source data for each grid, the heat source data including heat source location coordinates and heat source heat dissipation parameters; calculate the global trend term temperature for each grid based on the heat source location coordinates and heat source heat dissipation parameters, the global trend term temperature representing the overall temperature influence of all heat sources in the target space on the grid; calculate the local perturbation term temperature for each grid based on the temperature data, the global trend term temperature, and the grid location coordinates, the local perturbation term temperature representing the local temperature deviation correction of a single heat source on its surrounding grids; construct a three-dimensional temperature field model based on the global trend term temperature and the local perturbation term temperature to output the grid temperature of each grid, wherein the grid temperature is the superposition value of the global trend term temperature and the local perturbation term temperature of the corresponding grid; The real-time heat load of the target space is calculated according to the preset thermodynamic constraint equation; the real-time heat load is used to determine whether a heat source change occurs in the target space; if a heat source change occurs in the target space, the temperature of the local disturbance term is corrected by a nonlinear compensation algorithm, and the grid temperature corresponding to the grid is updated based on the corrected local disturbance term temperature. The temperature of each grid cell output by the three-dimensional temperature field model is input into a pre-trained spatiotemporal convolutional network model to obtain the predicted temperature corresponding to the grid cell. Obtain the current air supply parameters of the air conditioner, and adjust the air supply parameters according to the predicted temperature.

2. The method according to claim 1, characterized in that, The step of obtaining the current air supply parameters of the air conditioner includes: Identify the user's activity status within the target space, including sleep status, resting status, and movement status; Based on the preset mapping relationship between the activity state and the air supply parameters, the air supply parameters that match the identified activity state are determined. The air supply parameters include the temperature difference fluctuation range, wind speed, air supply angle, and coverage area.

3. The method according to claim 2, characterized in that, The step of obtaining the current air supply parameters of the air conditioner includes: If there are multiple users in the target space, identify multiple attribute parameters for each user and obtain the air supply coefficient corresponding to each attribute parameter, wherein the attribute parameters include at least age and clothing parameters; Calculate the comprehensive air supply coefficient for each user based on the multiple air supply coefficients for each user; The user with the lowest overall air supply coefficient among all users is identified as the target user, and the air supply parameters that match the activity status of the target user are identified as the current air supply parameters.

4. The method according to claim 3, characterized in that, After the step of adjusting the air supply parameters according to the predicted temperature, the method further includes: Record the air supply parameters after each adjustment, the grid temperature output by the three-dimensional temperature field model, the predicted temperature output by the spatiotemporal convolutional network model, the actual temperature data in the target space, and user feedback information, wherein the user feedback information includes the air supply parameters manually adjusted by the user. Calculate the deviation between the predicted temperature and the actual temperature data, and adjust the network parameters of the spatiotemporal convolutional network model according to the deviation using a preset reinforcement learning algorithm; The correlation adjustment rules between the air supply parameters and the predicted temperature are revised based on the user feedback information.

5. The method according to any one of claims 1-4, characterized in that, The step of inputting the temperature of each grid cell output by the three-dimensional temperature field model into a pre-trained spatiotemporal convolutional network model for prediction to obtain the predicted temperature corresponding to the grid cell includes: Spatiotemporal features are extracted from the current grid temperatures output by the three-dimensional temperature field model to obtain a grid temperature feature matrix that includes spatial location correlation features and time series change features; The grid temperature feature matrix is ​​concatenated with the historical grid temperature feature sequence within a preset time window to form a spatiotemporal input vector; The spatiotemporal input vector is input into a pre-trained spatiotemporal convolutional network model. The spatial convolutional layer of the model captures the temperature correlation between different grids, and the temporal convolutional layer captures the temperature change trend of the same grid at different times. The fully connected layer of the spatiotemporal convolutional network model outputs the temperature change sequence of each grid within a preset future time period, which is used as the predicted temperature for the corresponding grid.

6. An air conditioning control device based on a temperature field, characterized in that, The apparatus includes a unit for performing the method of any one of claims 1-5.

7. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, can implement the method as described in any one of claims 1-5.

Images