A method, system, medium, and device for dynamic indoor environment prediction

Abstract

The application discloses an indoor environment dynamic prediction method, system, medium and equipment, and relates to the technical field of building intelligentization. The method comprises the following steps: constructing a plurality of mechanism models for indoor environment prediction; inputting indoor environment related data into the plurality of mechanism models to obtain a plurality of first time prediction values of the plurality of mechanism models; taking the first time prediction values as inputs of a neural network, performing nonlinear capturing training on the neural network, obtaining a nonlinear combiner and obtaining a first time indoor physical parameter prediction value; determining a residual error between the first time indoor physical parameter prediction value and an actual value of the indoor physical parameter, inputting the residual error into a deep layered fuzzy model for training to obtain a data compensation model; inputting the indoor environment related data into the data compensation model to obtain a data compensation amount; and adding the residual error and the data compensation amount to obtain a second time physical parameter prediction value related to the indoor environment. Through the above scheme, the environmental parameters of a complex indoor environment can be accurately predicted.

Description

Technical Field

[0001] The present invention relates to the field of intelligent building technology, and in particular to a method, system, medium and equipment for dynamic prediction of indoor environment. Background Art

[0002] As society continues to develop, people are spending more and more time indoors, and their activities are becoming more diverse, causing the indoor environment of buildings to constantly change. Since good control of the indoor environment can help improve building performance and reduce building energy consumption, a technology for predicting the indoor environment is needed to better control various parameters of the indoor environment.

[0003] Existing technologies mainly predict indoor environments through a single model. However, the model has poor learning and generalization capabilities and is limited in its use of environmental information, resulting in the inability to accurately predict the physical parameters of complex indoor environments. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, system, medium and equipment for dynamic prediction of indoor environment to address the above technical problems.

[0005] The present invention adopts the following technical solutions:

[0006] In a first aspect, the present invention provides a method for dynamic prediction of an indoor environment, the method comprising:

[0007] Constructing a plurality of mechanism models for indoor environment prediction, and inputting data related to indoor environment prediction into the plurality of mechanism models to obtain a plurality of first-moment prediction values ​​of the plurality of mechanism models;

[0008] The predicted value at the first moment is used as an input of a neural network, and the neural network is trained to capture nonlinear relationships to obtain a nonlinear fuser, and the predicted value of the indoor physical parameter at the first moment is obtained;

[0009] Determine the residual between the predicted value of the indoor physical parameter at the first moment and the collected true value of the indoor physical parameter, and input the residual into the deep hierarchical fuzzy model for training to obtain a data compensation model;

[0010] Inputting the data related to indoor environment prediction into the data compensation model to obtain a data compensation amount;

[0011] The residual is added to the data compensation amount to obtain a predicted value of a physical parameter at a second moment related to the indoor environment.

[0012] Furthermore, before constructing the multiple mechanism models for indoor environment prediction, the method further includes:

[0013] Obtaining a heat source distribution in the room within a preset time period, and determining a first number of people in the room within the preset time period based on the heat source distribution;

[0014] Determining a second number of people in the room during the preset time period based on surveillance footage of the room during the preset time period;

[0015] During the preset time period, the persons in the room are located using ultra-wideband (UWB) technology to obtain a third number of persons in the room during the preset time period;

[0016] Obtaining the fourth number of people who enter or leave the room within the preset time period;

[0017] The first number of people, the second number of people, the third number of people, and the fourth number of people are used as inputs of a deep learning network model for training to obtain a personnel distribution model for predicting the number of people in the room within the preset time period, where the number of people in the room is one of the data related to the indoor environment.

[0018] Furthermore, the data related to indoor environment prediction specifically includes:

[0019] The parameter vector of the indoor environment, the parameter vector of the outdoor environment, the number of people in the room, the physical parameter vector of the room in the room, and the terminal start and stop status and parameter settings of the indoor air conditioning and fresh air system.

[0020] Furthermore, the construction of multiple mechanism models for indoor environment prediction specifically includes:

[0021] Constructing a building thermal dynamic model based on the acquired indoor structural information, constituent material information, external environment information, heating information, ventilation information, operating status of the air conditioning system, and start / stop status of the fresh air system;

[0022] Using the obtained indoor ventilation information, personnel activity information, indoor air flow rate information, and carbon dioxide generation and emission information as boundary conditions of the constraint equation to construct a carbon dioxide concentration diffusion model;

[0023] According to the collected indoor three-dimensional architectural information, window position information, and light blocking object information, simulation parameters about the indoor environment are defined, and the simulation parameters are input into the lighting simulation software to obtain an illumination model.

[0024] Furthermore, determining the residual between the predicted value of the indoor physical parameter at the first moment and the collected true value of the indoor physical parameter specifically includes:

[0025] Calculating the residual between the predicted value of the indoor physical parameter at the first moment and the collected true value of the indoor physical parameter;

[0026] When the residual is greater than a preset threshold, the predicted values ​​of the multiple mechanism models are input into the nonlinear fuser again for training until the residual is less than the preset threshold.

[0027] Furthermore, after constructing the plurality of mechanism models for indoor environment prediction, the method further includes:

[0028] Comparing the predicted values ​​output by the multiple mechanism models with the collected true values ​​corresponding to the multiple mechanism models to obtain a comparison result;

[0029] According to the comparison results, the parameters of the multiple mechanism models are adjusted.

[0030] Furthermore, after inputting the first moment prediction value into the nonlinear fuser to obtain the first moment prediction value of the indoor physical parameter, the method further includes:

[0031] Calculating an indoor energy consumption value associated with the predicted value of the indoor physical parameter at the first moment;

[0032] According to the predicted value of the indoor physical parameter at the first moment and the indoor energy consumption value, the actual value of the indoor physical parameter is adjusted to reduce the indoor energy consumption value.

[0033] In a second aspect, the present invention provides a system for dynamic prediction of indoor environment, comprising:

[0034] A first acquisition module is configured to construct a plurality of mechanism models for indoor environment prediction, input data related to indoor environment prediction into the plurality of mechanism models, and obtain a plurality of first-moment prediction values ​​of the plurality of mechanism models;

[0035] A training module is configured to: use the predicted value at the first moment as an input to a neural network, perform training on the neural network to capture nonlinear relationships, obtain a nonlinear fuser, and obtain a predicted value of the indoor physical parameter at the first moment;

[0036] a determination module, configured to determine a residual between a predicted value of the indoor physical parameter at the first moment and a collected true value of the indoor physical parameter, and input the residual into a deep hierarchical fuzzy model for training to obtain a data compensation model;

[0037] A second acquisition module is configured to input the data related to indoor environment prediction into the data compensation model to obtain a data compensation amount;

[0038] The calculation module is used to add the residual and the data compensation amount to obtain a predicted value of a physical parameter at a second moment related to the indoor environment.

[0039] The present invention provides a computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the indoor environment dynamic prediction method is implemented.

[0040] The present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the program, the indoor environment dynamic prediction method is implemented.

[0041] The at least one technical solution adopted by the present invention can achieve the following beneficial effects: first, construct multiple mechanism models for indoor environment prediction, input indoor environment related data into the multiple mechanism models, and obtain multiple first-moment prediction values ​​of the multiple mechanism models; then use the first-moment prediction value as the input of the neural network, perform nonlinear capture training on the neural network, obtain a nonlinear fusion device and obtain the first-moment indoor physical parameter prediction value; then determine the residual between the first-moment indoor physical parameter prediction value and the true value of the indoor physical parameter, and input the residual into the deep hierarchical fuzzy model for training to obtain a data compensation model; input the indoor environment related data into the data compensation model to obtain the data compensation amount; finally, add the residual and the data compensation amount to obtain the second-moment physical parameter prediction value related to the indoor environment. Through the above scheme, multiple mechanism models are combined with the data compensation model to dynamically predict indoor physical parameters. It is possible to comprehensively utilize the different forms of expression and characterization focus information of different models to dynamically predict the strongly coupled, nonlinear and highly uncertain indoor environment during operation. It effectively solves the problems of poor generalization ability of a single model in learning data and low utilization rate of external environmental information, so that the parameters of the complex indoor environment at a certain moment or several moments in the future can be accurately predicted, and a more accurate dynamic prediction of the indoor environment can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0043] Figure 1 A schematic flow chart of a method for dynamic prediction of indoor environment provided by the present invention;

[0044] Figure 2 A schematic diagram of a dynamic prediction modeling of a building indoor environment provided by the present invention;

[0045] Figure 3A flow chart of a multi-model fusion method provided by the present invention;

[0046] Figure 4 A schematic flow chart of the method for establishing a thermal dynamic model provided by the present invention;

[0047] Figure 5 A carbon dioxide concentration diffusion model diagram provided by the present invention;

[0048] Figure 6 A schematic diagram of an indoor environment dynamic prediction system provided by the present invention;

[0049] Figure 7 A schematic diagram of a computer device for implementing a method for dynamic prediction of indoor environment provided by the present invention. DETAILED DESCRIPTION

[0050] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention and corresponding drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0051] The server mentioned in the present invention can be a server installed on a business platform, or a device such as a desktop computer or laptop computer capable of executing the solution of the present invention. For ease of explanation, the following description will only use the server as the execution entity. The following, combined with the accompanying drawings, details the technical solutions provided by various embodiments of the present invention.

[0052] Figure 1 The following is a flow chart of a method for dynamic prediction of indoor environment in the present invention, which specifically includes the following steps:

[0053] S10: constructing a plurality of mechanism models for indoor environment prediction, and inputting data related to indoor environment prediction into the plurality of mechanism models to obtain a plurality of first-moment prediction values ​​of the plurality of mechanism models.

[0054] In this embodiment, schematically, Figure 2 This is a schematic diagram of dynamic prediction modeling of building indoor environment provided by the present invention.

[0055] refer to Figure 2The multiple mechanism models refer to mathematical models based on basic principles of physics, chemistry, biology, etc., which simulate the dynamic behavior and performance of the system by establishing mathematical equations, algorithms or logical relationships, reveal the internal laws of the system, and provide a basis for predicting and controlling the behavior of the system. The multiple mechanism models include but are not limited to: building thermal dynamic model, carbon dioxide concentration diffusion model and illumination model. The first moment can be the kth moment, where the kth moment can be any second in any hour. The data related to indoor environment prediction include but are not limited to: indoor environment parameter vector W i (k), outdoor environmental parameters W o (k), room physical parameter vector P, air conditioning and fresh air terminal start and stop status u c (k) and parameter setting y c (k) The data related to indoor environment prediction is obtained using sensors deployed during the operation of buildings and campuses.

[0056] S11: Using the predicted value at the first moment as the input of the neural network, performing training on the neural network to capture nonlinear relationships, obtaining a nonlinear fuser, and obtaining the predicted value of the indoor physical parameter at the first moment.

[0057] In this embodiment, before using the predicted value at the first moment as the input of the neural network, the predicted value of the mechanism model associated with the first moment needs to be normalized.

[0058] Optionally, in this embodiment, the normalization method may adopt a maximum-minimum normalization method, which can normalize data such as temperature, carbon dioxide concentration, and light intensity to the [0, 1] interval, as shown in formula (1):

[0059]

[0060] Among them, X n represents the nth data obtained by normalizing X in the data set, X represents the data to be processed, X min Represents the minimum value in the data set, X max Represents the maximum value in the data set.

[0061] refer to Figure 2 , the predicted value of the indoor physical parameter at the first moment is Figure 2 The predicted value of the medium mechanism model, Figure 2The true value in is the indoor physical parameter at the first moment, collected by pre-deployed sensors, including but not limited to: a carbon dioxide (CO2) concentration sensor, a light intensity sensor, an infrared sensor, a humidity sensor, and a temperature sensor. The residual refers to the error between the predicted value of the mechanism model and the collected true value of the indoor environment, and is used to measure the prediction error of the mechanism model.

[0062] In this embodiment, the normalized predicted value of the mechanism model associated with the first moment is input into a multilayer perceptron (MLP) for training to obtain an optimized predicted value of the predicted value of the mechanism model associated with the first moment.

[0063] Indicative, Figure 3 This is a flow chart of the multi-model fusion method provided by the present invention.

[0064] In this embodiment, the normalized predicted value of the mechanism model associated with the first moment is input into the neural network to train a nonlinear fuser, such as Figure 3 The model fuser in

[15] is a nonlinear fuser that can adaptively fuse the outputs based on the complex relationships between multidimensional data.

[0065] Optionally, the process of building a model fusion and data compensation model requires sufficient measured data and prediction data from the mechanism model for training and verification. Simultaneously, model selection and parameter adjustment are also key steps, and the appropriate model type and training algorithm can be selected based on the specific situation.

[0066] refer to Figure 3 , Figure 3 The neural network used in this paper is a multilayer perceptron. The objective function of this neural network is f(x, θ), where x is the set of input variables and θ is the set of unknown parameters, θ = {w1, w2, w3, b}. w1, w2, and w3 are the weight parameters for the building thermal dynamics model, the carbon dioxide concentration diffusion model, and the illumination model, respectively, and b is the bias term of the neural network. The goal of this neural network is to train the parameter set so that the predictions of the three models are most similar to the actual environmental values ​​collected. If the residual is greater than a set value, the neural network is trained again. Finally, the residual is added to the output of the data compensation model to obtain the final predicted output.

[0067] S12: Determine the residual between the predicted value of the indoor physical parameter at the first moment and the collected true value of the indoor physical parameter, and input the residual into the deep hierarchical fuzzy model for training to obtain a data compensation model.

[0068] In this embodiment, reference Figure 3 , Figure 3 The predicted value of the mechanism model shown in is the predicted value of the indoor physical parameter at the first moment. Figure 3 The data model shown in is the data compensation model.

[0069] In this embodiment, for data containing time series, the data compensation model can predict and compensate for environmental parameters at future moments based on the historical trends of each data.

[0070] In this embodiment, a deep neural network is designed as a feature extractor, and a fully connected neural network can be used for feature extraction, which can identify the data type and change trend of data such as temperature, carbon dioxide concentration, and light intensity.

[0071] In this embodiment, a fuzzy system is used to handle the uncertainty of physical quantities in complex environments. Decisions about physical quantities in the environment are made using fuzzy sets and fuzzy rules, and fuzzy logic is embedded in a deep neural network. The weights of the fuzzy rules can be dynamically updated through the training process of the neural network, forming an adaptive fuzzy inference system and a hybrid model.

[0072] In this embodiment, the deep hierarchical fuzzy model expands the traditional fuzzy system into a multi-layer structure. Each layer can be regarded as a sub-fuzzy system. Through layer-by-layer reasoning, higher-level and more abstract features or decision information are gradually extracted. Each layer in the deep hierarchical fuzzy model can be fuzzified and reasoned through fuzzy rules, and the output is used as the input of the next layer, layer by layer. The deep hierarchical fuzzy model includes an input layer, a hidden layer, and an output layer. The input layer converts the numerical value into a membership function through fuzzification. The fuzzy rules in the hidden layer describe the relationship between the input variables. The output of each layer can be regarded as the input of the next layer. The output layer includes a defuzzification process to convert the fuzzy output into a clear value.

[0073] S13: Inputting the data related to indoor environment prediction into the data compensation model to obtain a data compensation amount.

[0074] In this embodiment, reference Figure 2 , inputting the data related to indoor environment prediction into the data compensation model, and the output of the data compensation model is used to perform data compensation on the residual.

[0075] S14: Add the residual to the data compensation amount to obtain a predicted value of a physical parameter at a second moment related to the indoor environment.

[0076] In this embodiment, reference Figure 2, add the residual to the data compensation amount output by the data compensation model to obtain the predicted value of the physical parameter at the second moment related to the indoor environment. Wherein, the second moment refers to the moment after the first moment, such as Figure 2 The K+1 to K+L sampling periods in the .

[0077] based on Figure 1 A dynamic indoor environment prediction method is shown. Multiple mechanism models for indoor environment prediction are first constructed, and data related to indoor environment prediction is input into the multiple mechanism models to obtain multiple first-time prediction values ​​for the multiple mechanism models. The first-time prediction values ​​are then used as inputs to a neural network, which is trained to capture nonlinear relationships, resulting in a nonlinear fusion device and predicted values ​​of indoor physical parameters at the first time. The residual between the predicted values ​​of the indoor physical parameters at the first time and the collected true values ​​of the indoor physical parameters is then determined, and the residual is input into a deep hierarchical fuzzy model for training to obtain a data compensation model. Finally, the data related to indoor environment prediction is input into the data compensation model to obtain a data compensation amount, and the residual is added to the data compensation amount to obtain predicted values ​​of the physical parameters related to the indoor environment at the second time. The technical solution of this embodiment effectively addresses the problems of poor generalization ability of single models for data learning and low utilization of external environmental information. This allows for accurate prediction of complex indoor environment parameters at a specific future time or several future times, enabling building managers to better manage the building based on the predicted environmental parameters. This can also reduce building energy consumption and improve the working environment for indoor occupants.

[0078] When applying the indoor environment dynamic prediction method provided by the present invention, it is not necessary to Figure 1 The steps are executed in the order shown. The specific execution order of the steps can be determined according to needs, and the present invention does not limit this.

[0079] In addition, in one or more embodiments of the present invention, before constructing multiple mechanism models for indoor environment prediction, the method further includes:

[0080] S001: Obtain heat source distribution in the room within a preset time period, and determine a first number of people in the room within the preset time period based on the heat source distribution.

[0081] In this embodiment, the method of obtaining the heat source distribution includes but is not limited to: obtaining the heat source distribution in the room within the preset time period through an infrared sensor or a temperature sensor.

[0082] S002: Determine a second number of people in the room within the preset time period based on monitoring images of the room within the preset time period.

[0083] In this embodiment, the method of determining the second number of people in the room includes but is not limited to: determining the second number of people in the room by analyzing the number of portraits of people in the monitoring screen; or determining the second number of people in the room by analyzing the number of palms and / or soles of feet of people in the monitoring screen.

[0084] S003: Within the preset time period, the persons in the room are located by using ultra-wideband technology to obtain a third number of persons in the room within the preset time period.

[0085] In this embodiment, the ultra-wideband technology can be used to track the positions of people indoors in real time. By analyzing data at different time points and spatial positions, the estimation of the number of people indoors can be corrected and optimized.

[0086] S004: Acquire the fourth number of people who enter or leave the room within the preset time period.

[0087] In this embodiment, the method of obtaining the fourth number of people who enter or leave the room within the preset time period includes but is not limited to: counting the number of people who enter and leave the room through a sensor installed at the door.

[0088] S005: The first number of people, the second number of people, the third number of people, and the fourth number of people are used as inputs of a deep learning network model for training to obtain a personnel distribution model for predicting the number of people in the room within the preset time period, where the number of people in the room is one of the data related to the indoor environment.

[0089] In this embodiment, based on the first number of people, the second number of people, the third number of people, and the fourth number of people, the personnel distribution model is obtained by training a deep learning model through big data.

[0090] Through the technical solution of this embodiment, the indoor heat source distribution, the number of people in the monitoring video screen, the real-time location information of the indoor people and the number of people entering or leaving the room can be comprehensively evaluated, and the number of people in the room can be estimated more accurately.

[0091] Furthermore, in one or more embodiments of the present invention, the data related to indoor environment prediction specifically includes:

[0092] The parameter vector of the indoor environment, the parameter vector of the outdoor environment, the number of people in the room, the physical parameter vector of the room in the room, and the terminal start and stop status and parameter settings of the indoor air conditioning and fresh air system.

[0093] In this embodiment, the indoor environmental parameter vector includes but is not limited to: indoor temperature, indoor humidity, indoor volatile organic compounds (VOCs), indoor carbon dioxide concentration, indoor light intensity and indoor air pressure, and the outdoor environmental parameter vector includes but is not limited to: outdoor temperature, outdoor humidity, outdoor carbon dioxide concentration, outdoor light intensity and outdoor air pressure. The room physical parameter vector includes but is not limited to: indoor space pressure change and indoor air flow rate. The terminal start-stop status and parameter settings of the air conditioning and fresh air system include but are not limited to: the operating status of the air conditioning and fresh air system, air volume changes, energy consumption information and terminal start-stop status.

[0094] Furthermore, in one or more embodiments of the present invention, the construction of multiple mechanism models for indoor environment prediction specifically includes:

[0095] S101: Constructing a building thermal dynamic model based on the acquired indoor structural information, constituent material information, external environment information, heating information, ventilation information, operating status of the air-conditioning system, and start / stop status of the fresh air system.

[0096] In this embodiment, a building thermal dynamics model is established using building simulation software (such as EnergyPlus or TRNSYS) to describe processes such as heat transfer, heat storage, and heat radiation within a building. This model can take into account the building's structure, materials, external environmental conditions, and the operation of heating, ventilation, and air conditioning systems to predict temperature changes within the building.

[0097] Indicative, Figure 4 A schematic flow chart of a method for establishing a thermal dynamic model provided by the present invention includes:

[0098] S20: Determine the simulation objectives and scope.

[0099] S21: Collect building-related data.

[0100] S22: Create a geometric model to define building partitions.

[0101] S23: Input material and component properties.

[0102] S24: Set internal load.

[0103] S25: Input environmental and weather data and set simulation parameters.

[0104] S26: Run the simulation and analyze the results.

[0105] In this embodiment, a possible implementation scheme is to use EnergyPlus software to establish a building thermal dynamic model. First, it is necessary to determine the modeling objectives, collect detailed information about the building, draw the building geometry model, divide different thermal zones, and define the building shell and building materials. Secondly, set the internal load, define the parameters of the heating, ventilation and air conditioning (HVAC) system, set the meteorological data, and define the operation schedule. Finally, perform the simulation calculation. The simulation engine calculates the temperature, humidity, energy consumption and other parameters of the building at each time step by numerically solving the dynamic equations (such as the finite difference method or the finite element method).

[0106] Building thermal dynamic models are usually based on the heat balance equation to describe the heat transfer and storage process in the building. For a building's thermal zone, the heat balance equation is shown in formula (2):

[0107]

[0108] Where C is the heat capacity of the building, T(t) is the temperature of the hot zone at time t, ∑Q i (t) is the total heat entering the hot zone, ∑Q o (t) is the total heat flowing out of the hot zone. The heat entering and leaving the hot zone includes heat through conduction, convection, radiation, and mechanical ventilation. The conduction heat transfer formula is shown in formula (3):

[0109]

[0110] Among them, Q cond is the heat transfer, U is the heat transfer coefficient, A is the heat transfer area, R is the thermal resistance, T out (t) and T in (t) are the outdoor and indoor temperatures respectively. The convective heat transfer formula is shown in formula (4):

[0111] Q conv (t) = mc p [T air (t)-T in (t)] (4)

[0112] where Q conv is the convective heat, m is the air mass flow rate, C p is the specific heat capacity of air, T air (t) is the incoming air temperature. The radiation heat transfer formula is shown in formula (5):

[0113] Q rad (t) = A g S g (t)τ g(5)

[0114] where Q rad is the heat conducted by radiation, A g is the window area, S g (t) is the solar radiation intensity, τ g is the projection coefficient of the window.

[0115] S102: Using the obtained indoor ventilation information, personnel activity information, indoor air flow rate information, and carbon dioxide generation and emission information as boundary conditions of a constraint equation to construct a carbon dioxide concentration diffusion model.

[0116] In this embodiment, schematically, Figure 5 A diagram schematically shows a carbon dioxide concentration diffusion model.

[0117] like Figure 5 As shown, the model first transmits the coordinate position information to the neural network to obtain the velocity value u(x, θ) and pressure value p(x, θ) of this coordinate position. Then it is multiplied by the constraint equation of the boundary condition and further added to the special solution of the equation to obtain u H (x, θ) and p H (x, θ), and automatically differentiate the above results to predict the velocity at that point. Performing this operation for each point yields the velocity field u for the entire area. Substituting the velocity field value into the diffusion equation yields the change in carbon dioxide concentration. The boundary conditions in the carbon dioxide diffusion model shown in this embodiment account for the influence of the ventilation system, and the diffusion coefficient D takes into account factors such as human activity, indoor air flow, and carbon dioxide production and emission, enabling more accurate prediction of changes in indoor carbon dioxide concentration.

[0118] Among them, the carbon dioxide concentration diffusion model can use an algorithm that combines knowledge and neural networks to establish a turbulence model for prediction, combining the Navier-Stokes equation and the mass transfer equation (convection-diffusion equation) to describe it. Most low-speed flows can be described by the Navier-Stokes equation, as shown in formula (6):

[0119]

[0120] Where u is the velocity field vector, ρ is the fluid density, here the density of carbon dioxide in the air, p is the pressure, μ is the dynamic viscosity coefficient, and F is the external force (such as gravity). To ensure the conservation of mass during air flow, the fluid is incompressible. The change in carbon dioxide concentration is described by formula (7):

[0121]

[0122] Where C is the CO2 concentration, D is the diffusion coefficient, and u is the velocity field derived from the Navier-Stokes equations. Fluid motion requires setting boundary conditions, which determine how the gas interacts with the external environment. For CO2, boundaries can be specified as either concentration or flux.

[0123] S103: defining simulation parameters for the interior of the room according to the collected three-dimensional architectural information, window position information, and light blocking object information, and inputting the simulation parameters into illumination simulation software to obtain an illumination model.

[0124] In this embodiment, the illumination model can be established using illumination simulation software (such as Radiance) or an illumination calculation algorithm to predict the illumination level in the building interior, that is, the light intensity distribution. To establish the illumination model, it is first necessary to determine accurate building information based on the data set collected by the sensor, determine the lighting requirements of the building, and use a modeling tool to create a three-dimensional model of the building. Secondly, define the optical properties, set the simulation parameters, and run the illumination simulation program. Radiance will use the ray tracing method based on physical optics to solve the radiance equation to determine the energy exchange between different surfaces, thereby calculating the illuminance value of each point. The radiance equation is usually called the radiation transfer equation or the photometric equation, as shown in formula (8):

[0125]

[0126] Among them, L o (x,w o ) is the direction w at point x o The radiant brightness, L e (x,w o ) is the direction w at point x o The spontaneous radiation brightness (i.e. the light emitted by the light source), f r (x,w i , w0) is the bidirectional reflectance distribution function (BRDF) at point x, which describes the reflection of light from the incident direction w i To the outgoing direction w o Reflectance ratio. L i (x,w i ) is the direction w at point x i The incident radiance, w i n is the dot product of the incident direction and the surface normal, which determines the effect of light on the energy transfer of the surface. Ω is the hemispherical space of all possible incident directions. This model can account for factors such as the location of building windows, external lighting conditions, and the influence of obstructions.

[0127] This embodiment can fully utilize various complex parameters of the indoor environment and improve the prediction accuracy of the indoor environment by constructing a mechanism model including a thermal dynamic model, a carbon dioxide concentration diffusion model, and an illumination model.

[0128] Furthermore, in one or more embodiments of the present invention, determining the residual between the predicted value of the indoor physical parameter at the first moment and the collected true value of the indoor physical parameter specifically includes:

[0129] S131, calculating the residual between the predicted value of the indoor physical parameter at the first moment and the collected true value of the indoor physical parameter.

[0130] S132: When the residual is greater than a preset threshold, the predicted values ​​of the multiple mechanism models are input into the nonlinear fuser again for training until the residual is less than the preset threshold.

[0131] In this embodiment, when the residual is greater than the preset threshold, the predicted values ​​of the multiple mechanism models are input into the nonlinear fusion device again for training until the residual is less than the preset threshold. Figure 3 , Figure 3 The set value shown in is the preset threshold, and the preset threshold can be adjusted according to different indoor environments. Figure 3 The neural network shown in Figure 1 is a multilayer perceptron. Its objective function is f(x, θ), where x is the set of input variables and θ is the set of unknown parameters, θ = {w1, w2, w3, b}. w1, w2, and w3 are the weight parameters for the building thermal dynamics model, the carbon dioxide concentration diffusion model, and the illumination model, respectively, and b is the bias term for the neural network.

[0132] Through the solution described in this embodiment, the optimal weight parameters and bias can be found to reduce the errors between the predicted values ​​calculated by the three models and the collected true values ​​of the environment.

[0133] In addition, after constructing multiple mechanism models for indoor environment prediction, the method further includes:

[0134] Comparing the predicted values ​​output by the multiple mechanism models with the collected true values ​​corresponding to the multiple mechanism models to obtain a comparison result;

[0135] According to the comparison results, the parameters of the multiple mechanism models are adjusted.

[0136] In this embodiment, the predicted values ​​output by the multiple mechanism models are compared with the collected true values ​​corresponding to the multiple mechanism models. If the difference between the predicted values ​​output by the multiple mechanism models and the collected true values ​​corresponding to the multiple mechanism models is greater than a preset difference limit, the parameters of the multiple mechanism models need to be adjusted so that the error between the predicted values ​​output by the multiple mechanism models and the true values ​​is reduced.

[0137] In addition, after inputting the first moment prediction value into the nonlinear fuser to obtain the first moment prediction value of the indoor physical parameter, the method further includes:

[0138] Calculating an indoor energy consumption value associated with the predicted value of the indoor physical parameter at the first moment;

[0139] According to the predicted value of the indoor physical parameter at the first moment and the indoor energy consumption value, the actual value of the indoor physical parameter is adjusted to reduce the indoor energy consumption value.

[0140] In this embodiment, the indoor energy consumption value refers to the total electrical energy consumed by all electrical devices in the room. After obtaining the predicted indoor physical parameter value and the indoor energy consumption value, the actual value of the indoor physical parameter can be adjusted based on the actual experience of the indoor occupants.

[0141] Through the solution shown in this embodiment, the actual values ​​of indoor physical parameters can be reasonably adjusted according to the predicted values ​​of indoor physical parameters and indoor energy consumption, thereby saving electricity consumption and improving the comfort of the indoor environment.

[0142] The above is a method for dynamic prediction of indoor environment provided by one or more embodiments of the present invention. Based on the same idea, the present invention also provides a corresponding dynamic prediction system for indoor environment, such as Figure 6 shown.

[0143] Figure 6 A schematic diagram of an indoor environment dynamic prediction system provided by the present invention includes:

[0144] A first acquisition module is configured to construct a plurality of mechanism models for indoor environment prediction, input data related to indoor environment prediction into the plurality of mechanism models, and obtain a plurality of first-moment prediction values ​​of the plurality of mechanism models;

[0145] A training module is configured to: use the predicted value at the first moment as an input to a neural network, perform training on the neural network to capture nonlinear relationships, obtain a nonlinear fuser, and obtain a predicted value of the indoor physical parameter at the first moment;

[0146] a determination module, configured to determine a residual between a predicted value of the indoor physical parameter at the first moment and a collected true value of the indoor physical parameter, and input the residual into a deep hierarchical fuzzy model for training to obtain a data compensation model;

[0147] A second acquisition module is configured to input the data related to indoor environment prediction into the data compensation model to obtain a data compensation amount;

[0148] The calculation module is used to add the residual and the data compensation amount to obtain a predicted value of a physical parameter at a second moment related to the indoor environment.

[0149] For the specific definition of a dynamic indoor environment prediction system, please refer to the definition of a dynamic indoor environment prediction method above, and will not be repeated here. The various modules in the dynamic indoor environment prediction system can be implemented in whole or in part through software, hardware, or a combination thereof. The modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0150] The present invention also provides a computer-readable storage medium, which stores a computer program, which can be used to execute the Figure 1 A dynamic prediction method for indoor environment is provided.

[0151] The present invention also provides Figure 7 A schematic diagram of a computer device for realizing a dynamic prediction method of an indoor environment is shown in FIG. Figure 7 As shown, at the hardware level, the computer device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and may also include other hardware required for the business. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the Figure 1 A dynamic prediction method for indoor environment is provided.

[0152] Those skilled in the art will appreciate that all or part of the processes in the embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the various methods described. Among them, any reference to memory, storage, database or other media used in the various embodiments provided by the present invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0153] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of the present invention.

Claims

1. A method for dynamic prediction of indoor environment, characterized in that: include: Constructing a plurality of mechanism models for indoor environment prediction, and inputting data related to indoor environment prediction into the plurality of mechanism models to obtain a plurality of first-moment prediction values ​​of the plurality of mechanism models; The predicted value at the first moment is used as an input of a neural network, and the neural network is trained to capture nonlinear relationships to obtain a nonlinear fuser, and the predicted value of the indoor physical parameter at the first moment is obtained; Determine the residual between the predicted value of the indoor physical parameter at the first moment and the collected true value of the indoor physical parameter, and input the residual into the deep hierarchical fuzzy model for training to obtain a data compensation model; Inputting the data related to indoor environment prediction into the data compensation model to obtain a data compensation amount; Adding the residual to the data compensation amount to obtain a predicted value of a physical parameter at a second moment related to the indoor environment; The multiple mechanism models for indoor environment prediction are constructed, specifically including: Constructing a building thermal dynamic model based on the acquired indoor structural information, constituent material information, external environment information, heating information, ventilation information, operating status of the air conditioning system, and start / stop status of the fresh air system; Using the obtained indoor ventilation information, personnel activity information, indoor air flow rate information, and carbon dioxide generation and emission information as boundary conditions of the constraint equation to construct a carbon dioxide concentration diffusion model; Defining simulation parameters for the interior of the room based on the collected three-dimensional architectural information, window position information, and light obstruction information, and inputting the simulation parameters into illumination simulation software to obtain an illumination model; Determining the residual between the predicted value of the indoor physical parameter at the first moment and the collected true value of the indoor physical parameter specifically includes: Calculating the residual between the predicted value of the indoor physical parameter at the first moment and the collected true value of the indoor physical parameter; When the residual is greater than a preset threshold, the predicted values ​​of the multiple mechanism models are input into the nonlinear fuser again for training until the residual is less than the preset threshold.

2. The method for dynamic prediction of indoor environment according to claim 1, wherein: Before constructing the multiple mechanism models for indoor environment prediction, the method further includes: Obtaining a heat source distribution in the room within a preset time period, and determining a first number of people in the room within the preset time period based on the heat source distribution; Determining a second number of people in the room during the preset time period based on surveillance footage of the room during the preset time period; During the preset time period, the persons in the room are located by using ultra-wideband technology to obtain a third number of persons in the room within the preset time period; Obtaining the fourth number of people who enter or leave the room within the preset time period; The first number of people, the second number of people, the third number of people, and the fourth number of people are used as inputs of a deep learning network model for training to obtain a personnel distribution model for predicting the number of people in the room within the preset time period, where the number of people in the room is one of the data related to the indoor environment.

3. The method for dynamic prediction of indoor environment according to claim 1, wherein: The data related to indoor environment prediction specifically includes: The parameter vector of the indoor environment, the parameter vector of the outdoor environment, the number of people in the room, the physical parameter vector of the room in the room, and the terminal start and stop status and parameter settings of the indoor air conditioning and fresh air system.

4. The method for dynamic prediction of indoor environment according to claim 1, wherein: After constructing the multiple mechanism models for indoor environment prediction, the method further includes: Comparing the predicted values ​​output by the multiple mechanism models with the collected true values ​​corresponding to the multiple mechanism models to obtain a comparison result; According to the comparison results, the parameters of the multiple mechanism models are adjusted.

5. The method for dynamic prediction of indoor environment according to any one of claims 1 to 4, characterized in that: After inputting the first moment prediction value into the nonlinear fusion device to obtain the first moment prediction value of the indoor physical parameter, the method further includes: Calculating an indoor energy consumption value associated with the predicted value of the indoor physical parameter at the first moment; According to the predicted value of the indoor physical parameter at the first moment and the indoor energy consumption value, the actual value of the indoor physical parameter is adjusted to reduce the indoor energy consumption value.

6. A dynamic prediction system for indoor environment, characterized in that: include: A first acquisition module is configured to construct a plurality of mechanism models for indoor environment prediction, input data related to indoor environment prediction into the plurality of mechanism models, and obtain a plurality of first-moment prediction values ​​of the plurality of mechanism models; A training module is configured to: use the predicted value at the first moment as an input to a neural network, perform training on the neural network to capture nonlinear relationships, obtain a nonlinear fuser, and obtain a predicted value of the indoor physical parameter at the first moment; a determination module, configured to determine a residual between a predicted value of the indoor physical parameter at the first moment and a collected true value of the indoor physical parameter, and input the residual into a deep hierarchical fuzzy model for training to obtain a data compensation model; A second acquisition module is configured to input the data related to indoor environment prediction into the data compensation model to obtain a data compensation amount; a calculation module, configured to add the residual to the data compensation amount to obtain a predicted value of a physical parameter related to the indoor environment at a second moment; The multiple mechanism models for indoor environment prediction are constructed, specifically including: Constructing a building thermal dynamic model based on the acquired indoor structural information, constituent material information, external environment information, heating information, ventilation information, operating status of the air conditioning system, and start / stop status of the fresh air system; Using the obtained indoor ventilation information, personnel activity information, indoor air flow rate information, and carbon dioxide generation and emission information as boundary conditions of the constraint equation to construct a carbon dioxide concentration diffusion model; Defining simulation parameters for the interior of the room based on the collected three-dimensional architectural information, window position information, and light obstruction information, and inputting the simulation parameters into illumination simulation software to obtain an illumination model; Determining the residual between the predicted value of the indoor physical parameter at the first moment and the collected true value of the indoor physical parameter specifically includes: Calculating the residual between the predicted value of the indoor physical parameter at the first moment and the collected true value of the indoor physical parameter; When the residual is greater than a preset threshold, the predicted values ​​of the multiple mechanism models are input into the nonlinear fuser again for training until the residual is less than the preset threshold.

7. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by a processor, the method for dynamic prediction of an indoor environment according to any one of claims 1 to 4 is implemented.

8. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor. When the processor executes the program, the method for dynamic prediction of indoor environment according to any one of claims 1 to 4 is implemented.

Images