CN114970133A - Lighting prediction model training method, lighting prediction method, device and equipment - Google Patents

Abstract

The embodiment of the invention discloses a lighting prediction model training method, a lighting prediction device and lighting prediction equipment. Acquiring meteorological data and simulation time data of a target area, transmitting the meteorological data and the simulation time data to a lighting simulation system to set lighting simulation environment parameters, transmitting multiple groups of acquired building parameters to the lighting simulation system to obtain lighting data corresponding to the building parameters, wherein the building parameters and the corresponding lighting data form sample data; creating at least one initial lighting prediction model, and performing iterative training on each initial lighting prediction model based on training sample data in the sample data to obtain a lighting prediction model corresponding to each initial lighting model; and verifying the lighting prediction model based on verification sample data in the sample data to obtain at least one evaluation index of each lighting prediction model, and determining the target lighting prediction model based on the at least one evaluation index. And the creation, training and verification of the lighting prediction model are realized based on the sample data, and the precision of the lighting prediction model is improved.

Description

Daylighting prediction model training method, daylighting prediction method, device and equipment

technical field

Embodiments of the present invention relate to the technical field of smart city construction, and in particular, to a method for training a lighting prediction model, a method, device and equipment for lighting prediction.

Background technique

With the improvement of people's material and cultural level, there is a higher demand for indoor natural lighting indicators. Therefore, the prediction of indoor lighting can largely meet the health needs of people's living environment.

At present, the prediction method of indoor lighting is mainly to use lighting simulation software, such as Ecotect, EnergyPlus, etc.

However, such software is large in size and complex in input parameters. It is necessary to strictly follow the steps of model establishment, input parameters, simulation and analysis, which is time-consuming and inefficient. It requires highly specialized personnel, high time and labor costs, and input parameters The type is limited, and personalization cannot be achieved, resulting in limited scope of application and low predictive benefit.

SUMMARY OF THE INVENTION

The invention provides a daylighting prediction model training method, a daylighting prediction method, a device and equipment, so as to solve the problem of large software volume, complex input parameters, low efficiency, high cost, and prediction benefit for predicting the existence of indoor daylighting illumination through daylighting simulation software in the prior art. lower issues.

According to an aspect of the present invention, a training method for a lighting prediction model is provided, characterized in that it includes:

Obtain the meteorological data and simulation time data of the target area, transmit them to the lighting simulation system to set the environmental parameters of the lighting simulation, and transmit the acquired sets of building parameters to the lighting simulation system to obtain the lighting data corresponding to each set of building parameters, wherein , the building parameters and the corresponding lighting data form sample data;

Create at least one initial daylighting prediction model, perform iterative training on each of the initial daylighting prediction models based on the training sample data in the sample data, and obtain a daylighting prediction model corresponding to each initial daylighting model;

The daylighting prediction model is verified based on the verification sample data in the sample data to obtain at least one evaluation index of each daylighting prediction model, and a target daylighting prediction model is determined based on the at least one evaluation index.

According to another aspect of the present invention, there is provided a daylighting prediction method, characterized in that it includes:

Reading the architectural parameters of the preset type in the architectural model, inputting the architectural parameters into the pre-trained lighting coefficient prediction model, and obtaining the lighting coefficient corresponding to the architectural model;

Inputting the building parameters and the lighting coefficient into a pre-trained lighting prediction model to obtain the natural lighting illuminance corresponding to the building model, wherein the lighting coefficient prediction model and the lighting prediction model are based on claims 1- 5. Any one of the training methods for the daylighting prediction model is obtained.

According to another aspect of the present invention, a lighting prediction model training device is characterized in that, comprising:

The sample data acquisition module is used to acquire the meteorological data and simulated time data of the target area, transmit them to the lighting simulation system to set the environmental parameters of the lighting simulation, and transmit the acquired sets of building parameters to the lighting simulation system to obtain each set of building parameters. Corresponding lighting data, wherein the building parameters and the corresponding lighting data form sample data;

The daylighting prediction model training module is used to create at least one initial daylighting prediction model, perform iterative training on each of the initial daylighting prediction models based on the training sample data in the sample data, and obtain a daylighting prediction model corresponding to each initial daylighting model;

The target daylighting prediction model determination module performs verification processing on the daylighting prediction model based on the verification sample data in the sample data, obtains at least one evaluation index of each daylighting prediction model, and determines the target daylighting prediction based on the at least one evaluation index. Model.

According to another aspect of the present invention, a lighting prediction device is characterized in that, comprising:

The daylighting coefficient prediction module is used to read the building parameters of the preset type in the building model, input the building parameters into the pre-trained daylighting coefficient prediction model, and obtain the daylighting coefficient corresponding to the building model;

The natural daylighting illuminance prediction module is used to input the building parameters and the daylighting coefficient into a pre-trained daylighting prediction model to obtain the natural daylighting illuminance corresponding to the building model, wherein the daylighting coefficient prediction model and the daylighting prediction The models are respectively obtained based on the training method of the daylighting prediction model provided in any embodiment.

According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

The memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform any of the embodiments of the present invention. Daylighting prediction model training method and/or daylighting prediction method.

According to another aspect of the present invention, a computer-readable storage medium is provided, where computer instructions are stored in the computer-readable storage medium, and the computer instructions are used to cause a processor to implement any of the embodiments of the present invention when executed. The daylighting prediction model training method and/or the daylighting prediction method.

The technical scheme of the embodiment of the present invention builds, trains and verifies a lighting prediction model based on sample data, does not require on-site measurement and a large amount of building parameter data when using the prediction function, and solves the problems of complex input parameters, low prediction efficiency, and operation in the prior art. Professional problems make the prediction of natural lighting illuminance more efficient, more accurate in prediction results, and lower in prediction cost. In addition, the forecasting method of daylighting forecasting using a forecasting model is small in size, simple in operation, clear in instructions, low in requirements for the user's professional level and operating environment, and has a wide range of applications.

It should be understood that the content described in this section is not intended to identify key or critical features of the embodiments of the invention, nor is it intended to limit the scope of the invention. Other features of the present invention will become readily understood from the following description.

Description of drawings

In order to illustrate the technical solutions in the embodiments of the present invention more clearly, the following briefly introduces the accompanying drawings used in the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained from these drawings without creative effort.

1 is a flowchart of a method for training a daylighting prediction model provided in Embodiment 1 of the present invention;

Fig. 2 is a flow chart of a lighting prediction method provided in Embodiment 2 of the present invention;

3 is a schematic structural diagram of a lighting prediction model training device provided in Embodiment 3 of the present invention;

4 is a schematic structural diagram of a lighting prediction device provided in Embodiment 4 of the present invention;

5 is a schematic structural diagram of an electronic device that can be used to implement embodiments of the present invention.

Detailed ways

In order to make those skilled in the art better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only Embodiments are part of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

It should be noted that the terms "first", "second" and the like in the description and claims of the present invention and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific sequence or sequence. It is to be understood that the data so used may be interchanged under appropriate circumstances such that the embodiments of the invention described herein can be practiced in sequences other than those illustrated or described herein. Furthermore, the terms "comprising" and "having" and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those expressly listed Rather, those steps or units may include other steps or units not expressly listed or inherent to these processes, methods, products or devices.

Example 1

FIG. 1 is a flowchart of a method for training a lighting prediction model according to Embodiment 1 of the present invention. This embodiment can be applied to the situation of training an indoor daylighting prediction model. The method can be executed by the daylighting prediction model training device provided by the implementation of the present invention. The deployment and control device can be realized by software and/or hardware. The deployment and control device can be configured in an electronic computing On the device, it includes the following steps:

S110. Acquire meteorological data and simulation time data of the target area, transmit them to a daylighting simulation system to set environmental parameters for daylighting simulation, and transmit the acquired sets of building parameters to the daylighting simulation system to obtain daylighting data corresponding to each set of building parameters , wherein the building parameters and the corresponding lighting data form sample data.

The target area is the geographic area where the building to be simulated is located. Optionally, the target area can be obtained by dividing the administrative area. For example, the target area can be divided based on the administrative level of the province/city/district; optional , the target area can also be divided based on information such as the latitude and altitude where the building is located. Since the climate, location and other factors of different areas affect the lighting data in the building, in this embodiment, the lighting simulation system is set based on the environmental data of the same target area to set the lighting simulation environment parameters, so as to obtain the lighting data used in the target area to form a sample Correspondingly, the sample data of the target area is trained to obtain a daylighting prediction model suitable for daylighting prediction of buildings in the target area, so as to improve the accuracy and pertinence of the daylighting prediction model and reduce the identification error caused by the influence of the area.

Among them, the meteorological data of the target area is determined by the area and date selected by the user. Specifically, if the target area selected by the user is Tianjin, and the simulation time data selected by the user is February 20, 2022, then the Meteorological data refers to the meteorological data of Tianjin on February 20, 2022. The meteorological data may be, but not limited to, cloudy, rainy, cloudy, etc. Further, the meteorological data may also be clear from 9:00 to 11:00. , cloudy from 11:00-15:00, light rain from 15:00-17:00, etc., which are not limited. Among them, different meteorological data can affect the lighting conditions in the building. For example, meteorological data such as cloudy days, rainy days, snow days, and foggy days lead to poor lighting in the building; meteorological data such as sunny days lead to good lighting in the building. Through the historical meteorological data in the target area, the lighting simulation environment is set to improve the authenticity and accuracy of the lighting simulation of the buildings in the target area.

The simulation time data is the time data set by the user for lighting simulation, which is associated with the obtained meteorological data. Optionally, the simulation time data can be a certain season or a certain date, because the sunshine intensity of different seasons and different days is different. , the lighting will be affected to a certain extent. Specifically, the sunshine intensity in summer is relatively high, and the corresponding sunshine intensity in winter is relatively low, so we can use the season as the division standard to train the lighting prediction model for different seasons; At the very least, the daylighting prediction model can be trained more specifically for this special date. In this embodiment, the lighting simulation system is set based on the meteorological data and the simulation time data to set the lighting simulation environment parameters, so as to obtain the lighting data used in the target area to form sample data. A daylighting prediction model for daylighting forecasting on specific dates, seasons and specific dates, to further improve the accuracy and pertinence of daylighting forecasting, and reduce the impact of climate change and seasonal replacement on daylighting forecasting.

Meteorological data and simulation time data are used to set the environmental parameters of daylighting simulation in the daylighting simulation system. The acquired sets of building parameters are transmitted to the daylighting simulation system as the input of the daylighting simulation system, and the daylighting data corresponding to each group of building parameters can be obtained.

The daylighting data follows the normalization principle, which is as follows:

Since there is no negative value for each physical quantity, the value range is selected as [0,1]. The original value range is mapped to [0,1], which can be transformed in two ways: normal distribution and uniform distribution. Since there is no obvious normal distribution relationship between the values of physical quantities, the uniform distribution method is used for normalization. Let the original data (such as the lighting data obtained by simulation) be data, the normalized array is data', min() is the function of taking the minimum value, max() is the function of taking the maximum value, the serial number is i, and its mathematical expression :

In order to display the modeling results more intuitively, after performing the modeling operation on the normalized values, the model output should be mapped to the original value range, and this process is denormalization. Denormalization is a process that is completely opposite to normalization, so the denormalization formula can be obtained from the above expression:

data(i)=data'(i)×(max(data)-min(data))+min(data)

The lighting simulation system may be a lighting simulation software, and the type of the lighting simulation system is not limited, as long as the lighting simulation can be performed on the input building parameters. Build a building model in the simulation software, load the local meteorological data into the model file, set the time and date, input the building parameters, set the size of the calculation grid, simulate, obtain the lighting data, and organize and analyze the lighting data. Wherein, the building parameters include window sill height, shading size, wall material reflectivity, glass light transmittance, etc.; the lighting data is the simulation and analysis of building parameters, including window sill height, shading size, wall material, etc. The independent effects of reflectivity and glass transmittance are simulated and analyzed, as well as the cross effects of glass transmittance, wall material reflectivity, and window sill height. The building parameters and the corresponding lighting data form sample data, which is the basis for training and validating the lighting prediction model.

S120. Create at least one initial lighting prediction model, perform iterative training on each of the initial lighting prediction models based on the training sample data in the sample data, and obtain a lighting prediction model corresponding to each initial lighting model.

The initial daylighting prediction model is the daylighting prediction model to be trained, and the initial daylighting prediction model is iteratively trained based on the training sample data, and the trained daylighting prediction model can be obtained. Among them, the training sample data refers to a part of the sample data that is arbitrarily selected from the sample data and used to train the lighting prediction model, and the remaining part of the sample data is used as the verification sample data to verify the lighting prediction model. In the sample data, 90% of the data is arbitrarily selected as the training sample data for the training model, and the remaining 10% of the data is used as the verification sample data for the verification model. The initial daylighting prediction model can be created by the user, the user can independently set the network structure and network depth of the model, create different initial daylighting prediction models based on different network structures and/or different network depths, and obtain different types of daylighting prediction models or different networks In-depth daylighting prediction model, the user can select the optimal daylighting prediction model according to the evaluation index from the obtained daylighting prediction model. By iterative training of the initial daylighting prediction model, the prediction accuracy of the daylighting prediction model can be improved.

S130. Perform verification processing on the daylighting prediction model based on the verification sample data in the sample data, obtain at least one evaluation index of each daylighting prediction model, and determine a target daylighting prediction model based on the at least one evaluation index.

Wherein, the evaluation index obtained by the verification process is used as the evaluation standard of the prediction effect of the daylighting prediction model, and the target daylighting prediction model is determined, and the evaluation index includes at least one of the following: average relative error, maximum relative error, mean square error and fitting Spend. Specifically, the verification result is obtained by processing the input verification sample data based on the daylighting prediction model, and the at least one evaluation index described above is calculated based on the verification result. The evaluation index is an important basis for determining the target daylighting prediction model. Therefore, as many evaluation indicators as possible should be selected as the basis for determining the target daylighting prediction model. In this embodiment, based on the verification sample data, the trained lighting prediction model is verified to obtain each evaluation index, and the evaluation indicators of each lighting prediction model are compared to determine the optimal lighting prediction model. The prediction model is the target lighting prediction model. In some embodiments, multiple evaluation indexes may be weighted, and the evaluation indexes of each lighting prediction model are compared based on the weighted result.

Among them, the calculation method of each evaluation index is as follows:

The average relative error is the average value of the absolute value of the error between the model output value and the actual value in the ratio of the actual value range. The ideal value is 0. The mathematical formula is:

The maximum relative error is the maximum value of the absolute value of the error between the model output value and the actual value in the actual value range. The ideal value is 0. The mathematical formula is:

The mean square error is the root mean square value of the ratio of the absolute value of the error between the model output value and the actual value to the actual value range. The ideal value is 0. The mathematical formula is:

The degree of fit, also known as the goodness of fit, describes the relationship between the dependent variable and the independent variable of the regression equation, that is, the percentage of the variability of the dependent variable that the model can reveal. The ideal value is 1. The mathematical formula:

In the above formula, n is the number of output values, i is the ith model output value, X _c is the model output value, X _c-mean is the model output average, X _r is the actual value, and X _r-max is the actual maximum value. value, X _r-min is the actual minimum value.

In the technical solution provided by this embodiment, the environmental parameters of the lighting simulation system are set by the meteorological data and the simulation time data of the target area, and the lighting data used in the target area is obtained to form sample data, and the sample data of the target area can be trained to obtain suitable lighting for the target area. The lighting forecasting model for buildings, weather, seasons or specific dates in the target area solves the problem of inaccurate model forecasting in the prior art, improves the accuracy and pertinence of the lighting forecasting model, reduces regional, weather changes, The impact of seasonal change on the training model; the daylighting prediction model is trained by iterative training based on the training sample data, which further improves the accuracy of the model prediction; The daylighting prediction model with the best prediction performance further improves the accuracy of daylighting prediction.

On the basis of the above embodiment, before transmitting the acquired sets of building parameters to the lighting simulation system to obtain the lighting data corresponding to each set of building parameters, optionally, the method further includes: obtaining the target area Multiple types of building type information in the system are transmitted, and the building type information is transmitted to the lighting simulation system, so that the lighting simulation system can create corresponding multiple types of building models.

Obtain the unit type information of various types of units used in the target area, perform statistical analysis on the unit type information, obtain the usage ratio of each type of information, select several types of units with higher usage ratios for lighting simulation, and create corresponding lighting in the internal system. Among them, the standard with a higher proportion of use can be the top 5 or top 10 units in the proportion of use, or it can be the type of apartment with a proportion of use exceeding a certain percentage, and the specific percentage is determined by the user according to actual needs. In this embodiment, based on a variety of building type information, the lighting simulation system creates a variety of corresponding building models, and then obtains each group of sample data corresponding to each group of building types. Correspondingly, each group of sample data is trained to obtain suitable The lighting forecasting model for lighting forecasting of various building types in the target area further improves the pertinence of the forecasting model.

Optionally, transmitting the acquired sets of building parameters to the daylighting simulation system to obtain lighting data corresponding to each set of building parameters, including: transmitting the acquired sets of building parameters to the daylighting simulation system, so that the The daylighting simulation system updates the various building models based on the building parameters, and performs daylighting simulation on the updated various building models respectively, to obtain the daylighting data output by the daylighting simulation system, wherein based on the building parameters and daylighting The daylighting prediction model obtained by the data training is used to predict the daylighting of the buildings in the target area.

For each building model, various building parameters are input respectively to realize the lighting simulation of different building parameters under the building model, which improves the efficiency of lighting simulation; The lighting simulation of building parameters is carried out to obtain the corresponding sets of lighting data, and the lighting prediction model is trained based on the multiple sets of building parameters and lighting data, so as to obtain a lighting prediction model suitable for the entire target area, and realize the lighting prediction of the buildings in the target area. , greatly improve the efficiency of daylighting prediction model prediction, reduce labor costs, and reduce the waste of resources caused by repeated creation and training of models.

Based on the above technical solution, the building parameters include window sill height, sunshade size, wall material reflectivity, and glass light transmittance; the lighting data includes at least one of lighting coefficient and natural lighting illuminance.

Optionally, performing iterative training on each of the initial lighting prediction models based on the training sample data in the sample data to obtain a lighting prediction model corresponding to each initial lighting model includes: using the building parameters as input data, The daylighting coefficient is used as the training sample data formed by the label, and the initial daylighting prediction model is iteratively trained to obtain the daylighting coefficient prediction model corresponding to each initial daylighting model; and/or, based on the building parameters and the daylighting coefficient as The input data and the natural lighting illuminance are used as training sample data formed by labels, and the initial lighting prediction model is iteratively trained to obtain a lighting illuminance prediction model corresponding to each initial lighting model.

Taking the daylighting coefficient prediction model as the neural network model as an example, the training process of the daylighting prediction model is as follows:

(1) Input a set of building parameters in the sample data into the initial neural network lighting prediction model, and the weights in the initial neural network lighting prediction model are random quantities or initial parameters;

(2) Calculate the predicted value of the initial lighting prediction model along the direction of signal propagation;

(3) Calculate the loss function through the error principle based on the predicted value and the lighting data in the sample data;

(4) Adjust the weights along the reverse direction of signal propagation based on the principle of loss function and weight adjustment amount, and obtain an updated neural network lighting prediction model;

(5) Repeat the process of 1-4 for each remaining group of sample data until the error does not exceed a certain range, and obtain a trained lighting prediction model.

The following is the error principle followed by the neural network prediction model and the weight adjustment principle followed by the neural network prediction model.

The error principle followed by the neural network prediction model is:

期望输出向量为

误差向量为

因此第l次训练的误差：Taking the three-layer BP neural network as an example, the number of nodes in the input layer is n, the number of nodes in the hidden layer is p, and the number of nodes in the output layer is m. At the same time, let the r-th node of the input layer be x _r , the q-th node of the hidden layer be k _q , the t-th node of the output layer to be y _t , the weight between each layer is ω _qt , and the subscripts represent the two The sequence number of the layer node. At the same time, let the number of training (ie the number of iterations) be l, the input of each layer is u, and the output is v. The output vector of the network model is

The expected output vector is

The error vector is

So the error of the lth training:

e _t (l)=d _t (l)-y _t (l)

The first is the forward propagation of the signal. The output signal of the input layer is the input signal of the network model. Let the activation function be g( ), then the input of the qth node of the hidden layer:

The output of the qth node of the hidden layer:

The output of the t-th node of the output layer:

The error of the t-th node of the output layer:

The principle of weight adjustment amount followed by the neural network prediction model is:

The weights are adjusted in the opposite direction of signal propagation. According to the gradient descent method, adjust the gradient of ω _qt in the opposite direction: (η is the learning rate)

ω _qt (l)=Δω _qt (l)+ω _qt (l-1)

In accordance with the chain rule of differential equations, the gradient value can be found:

The adjustment amount of its weight:

又被称为局部梯度。对于三层的BP神经网络，总结上面各式可得出如下结论，即权值的调节量是学习率、局部梯度和上一层信号输出强度的乘积。对于更为复杂多层BP神经网络，将上面计算过程进行延伸即可。in

Also known as local gradient. For the three-layer BP neural network, the following conclusions can be drawn from summarizing the above formulas, that is, the adjustment amount of the weight is the product of the learning rate, the local gradient and the signal output intensity of the previous layer. For more complex multi-layer BP neural networks, the above calculation process can be extended.

The daylighting prediction model includes a daylighting coefficient prediction model and a daylighting illuminance prediction model. According to different input and output, it is divided into the following two prediction models:

(1) Daylighting coefficient prediction model: the input data is the building parameters, the building parameters include: window sill height, shade size, wall material reflectivity, glass light transmittance, and the output data is the daylighting coefficient;

(2) Lighting illuminance prediction model: the input data is the lighting coefficient and building parameters, such as the height of the window sill, the size of the sunshade, the reflectivity of the wall material, and the light transmittance of the glass, and the output data is the lighting illuminance;

The lighting coefficient prediction model is used to predict the lighting coefficient, and the lighting coefficient can be used as one of the input data of the lighting illuminance prediction model to predict the lighting illuminance.

Predicting the lighting coefficient by training the lighting coefficient prediction model can improve the accuracy of the lighting coefficient. Then, using the lighting coefficient as the input data to train the lighting illuminance prediction model, the prediction accuracy of the lighting illuminance prediction model can be further improved, and the prediction result can be further improved. The accuracy of lighting illuminance.

Based on the above technical solutions, optionally, the network structures of the at least one initial daylighting prediction model are different, or the network depths of the at least one initial daylighting prediction model are different. Wherein, the daylighting prediction model includes, but is not limited to, a neural network prediction model, a combined prediction model, a Kalman filter prediction model, and the like. Create at least one initial daylighting prediction model for training, and the network structure and/or network depth of the initial daylighting prediction model are different. Taking the neural network prediction model as an example, the following models can be created:

(1) Model 1: The training model is set to 3 hidden layer structures, the number of neurons in each layer is [12, 12, 1], and the corresponding excitation functions of each layer are 'purelin', 'logsig', 'purelin' ';

(2) Model 2: The training model is set to 4 hidden layer structures, the number of neurons in each layer is [12, 12, 12, 1], and the excitation functions of each layer are 'purelin', 'tansig', ' logsig', 'purelin', retrain the model;

Among the above two models, model 1 and model 2 have different network structures and network depths, and the performance and prediction accuracy of daylighting prediction models with different network structures and network depths are different. Create and train as many daylighting prediction models as possible. The daylighting prediction model with better performance can be screened and the accuracy of daylighting prediction can be improved.

The iterative training of each initial lighting prediction model based on the training sample data in the sample data to obtain a lighting prediction model corresponding to each initial lighting model includes: differentiating the initial lighting prediction model based on the training sample data. Iterative training for the number of training times can obtain the lighting prediction model corresponding to each initial lighting model.

Taking the neural network prediction model as an example, the following model can be created: randomly select about 90% of the three sets of independently affected simulated data, that is, 60 sets of data, as data set 1, and train the neural network prediction model to form Train model 0 to get the prediction result of daylighting coefficient; input the prediction result of daylighting coefficient generated by dataset 1 and training model 0 into training model X to get the prediction result of natural lighting illuminance; training model X includes training model 1, training model Model 2, training model 3, training model 4; training model 1 is set to 3 hidden layer structures,

Set the sample data obtained as 68 groups, and randomly select about 90% of the data in the sample data, that is, 60 groups of data, as the training sample data, train the neural network prediction model, and form the daylighting coefficient prediction model to obtain the prediction of the daylighting coefficient. Result; input the daylighting coefficient prediction result and sample data into the following training model to get the training result. (1) Model 1: The training model is set to 3 hidden layer structures, the number of neurons in each layer is [12, 12, 1], and the excitation functions corresponding to each layer are 'purelin', 'logsig', 'purelin' ';

(2) Model 2: The model structure is unchanged, and the training times are changed;

(3) Model 3: The training model is set to 4 hidden layer structures, the number of neurons in each layer is [12, 12, 12, 1], and the excitation functions of each layer are 'purelin', 'tansig', ' logsig', 'purelin'}, retrain the model;

(4) Model 4: Keep the model structure unchanged and change the training times.

The remaining 10% of the data in the sample data, that is, 8 groups of data, are used as the verification sample data and verified, as shown in the following table. Predictive model of daylighting. By changing the training times of the daylighting prediction model, the models are made more diverse, it is convenient to select the daylighting prediction model with the highest prediction accuracy, and the prediction accuracy of the daylighting prediction model is improved.

This embodiment provides a training method for a lighting prediction model. The method simulates a lighting environment based on a lighting simulation software, takes the lighting data obtained by the simulation software and the corresponding building parameters as sample data, and creates, trains and verifies a lighting prediction based on the sample data. model, to realize the training of the lighting prediction model. The method obtains input data by simulating software, which reduces labor costs and improves the efficiency of daylighting prediction model training; at the same time, this method creates and trains various types of daylighting prediction models, and compares and selects the optimal daylighting prediction model In this way, the prediction accuracy of the daylighting prediction model is improved.

Embodiment 2

FIG. 2 is a flowchart of a lighting prediction method provided in Embodiment 2 of the present invention, and the embodiment of the present invention may be combined with each optional solution in the foregoing embodiment. The daylighting prediction method specifically includes the following steps:

S210, reading the architectural parameters of the preset type in the architectural model, and inputting the architectural parameters into a pre-trained lighting coefficient prediction model to obtain the lighting coefficient corresponding to the architectural model;

S220. Input the building parameters and the lighting coefficient into a pre-trained lighting prediction model to obtain the natural lighting illuminance corresponding to the building model, wherein the lighting coefficient prediction model and the lighting prediction model are based on the present invention respectively The training method of the daylighting prediction model described in any embodiment is obtained.

Among them, the building model refers to the model corresponding to the building type for which lighting prediction is performed. Based on the area corresponding to the building model and the building parameters read in the building model, the corresponding daylighting coefficient prediction model and daylighting prediction model are automatically matched to reduce model waste. time, speeding up the speed of daylighting prediction. There are many types of building parameters, including building size, glass transmittance, wall material reflectivity, window sill height, building orientation, wall thickness, and the location and material of doors, windows, interior and exterior walls, roofs, indoor and outdoor floors, and floors etc., but not all types of building parameters are required for daylighting prediction. Presetting the type of building parameters required before reading the building model can avoid reading useless parameters and secondary screening of the read parameters. , improve the efficiency of building parameters acquisition, and then improve the efficiency of daylighting prediction. In addition, obtaining the architectural parameters of the architectural model by reading the architectural model can also improve the efficiency of daylighting prediction. Taking the neural network prediction model as an example, the following is a specific implementation of a lighting prediction method.

Select a typical apartment type A in a certain area, read the building model of apartment type A, and its architectural parameters include building size, glass transmittance, wall material reflectivity, window sill height, building direction, as well as doors, windows, interior and exterior walls, roofs, interiors The location and material of the outer ground and floor slab; select the area where the unit type A is located, match the lighting coefficient prediction model suitable for the building parameters and the selected area, input the building parameters into the lighting coefficient prediction model, and predict the lighting coefficient; similarly, Match the daylighting prediction model suitable for the building parameters and the selected area, and input the predicted daylighting coefficient and building parameters into the daylighting prediction model to obtain the natural daylighting illuminance. Select other units to perform the above operations, and you can also get the natural lighting illumination of this unit.

The daylighting prediction method provided in this embodiment is realized based on the daylighting prediction model obtained by the daylighting prediction model training method, and the daylighting prediction method includes all the lighting prediction models in the daylighting prediction model training method provided by any of the above embodiments. beneficial effect.

Embodiment 3

FIG. 3 is a schematic structural diagram of a lighting prediction model training device provided in Embodiment 3 of the present invention. The device includes: a sample data acquisition module 310 , a lighting prediction model training module 320 , and a target lighting prediction model determination module 330 .

The sample data acquisition module 310 is used to acquire the meteorological data and simulation time data of the target area, transmit them to the daylighting simulation system to set the environmental parameters of the daylighting simulation, and transmit the acquired sets of building parameters to the daylighting simulation system to obtain each group of buildings. The lighting data corresponding to the parameters, wherein the building parameters and the corresponding lighting data form sample data.

The daylighting prediction model training module 320 is configured to create at least one initial daylighting prediction model, perform iterative training on each of the initial daylighting prediction models based on the training sample data in the sample data, and obtain a daylighting prediction model corresponding to each initial daylighting model.

The target daylighting prediction model determination module 330 performs verification processing on the daylighting prediction model based on the verification sample data in the sample data, obtains at least one evaluation index of each daylighting prediction model, and determines the target daylighting based on the at least one evaluation index prediction model.

Optionally, before transmitting the acquired multiple sets of building parameters to the lighting simulation system to obtain lighting data corresponding to each set of building parameters, the device further includes:

The multiple building model creation unit is used for acquiring multiple building type information in the target area, and transmitting the building type information to the lighting simulation system, so that the lighting simulation system creates corresponding multiple building models .

The acquired multiple sets of building parameters are transmitted to the lighting simulation system to obtain lighting data corresponding to each set of building parameters, including:

The daylighting data acquisition unit is used to transmit the acquired sets of building parameters to the daylighting simulation system, so that the daylighting simulation system updates the various building models based on the building parameters, and updates the various building models after updating. The lighting simulation is performed respectively to obtain the lighting data output by the lighting simulation system, wherein the lighting prediction model obtained by training based on the building parameters and the lighting data is used to predict the lighting of the buildings in the target area.

Optionally, the architectural parameters include window sill height, shade size, wall material reflectivity, and glass light transmittance; the lighting data includes at least one of a lighting coefficient and natural lighting illuminance.

The iterative training is performed on each of the initial lighting prediction models based on the training sample data in the sample data to obtain a lighting prediction model corresponding to each initial lighting model, including:

The daylighting coefficient prediction model training unit is configured to perform iterative training on the initial daylighting prediction model based on the building parameters as input data and the daylighting coefficient as training sample data formed by labels, and obtain the daylighting coefficient predictions corresponding to each initial daylighting model Model. and / or,

The lighting illuminance prediction model training unit performs iterative training on the initial lighting prediction model based on the building parameters and the lighting coefficient as input data, and the natural lighting illuminance as the training sample data formed by the label, and obtains the corresponding initial lighting models. The lighting illuminance prediction model.

Optionally, the network structure of the at least one initial daylighting prediction model is different, or the network depth of the at least one initial daylighting prediction model is different.

The iterative training is performed on each of the initial lighting prediction models based on the training sample data in the sample data to obtain a lighting prediction model corresponding to each initial lighting model, including:

The daylighting prediction model training unit is configured to perform iterative training on the initial daylighting prediction model with different training times based on the training sample data, and obtain a daylighting prediction model corresponding to each initial daylighting model.

Optionally, the evaluation index includes at least one of the following: average relative error, maximum relative error, mean square error, and goodness of fit.

The daylighting prediction model training device provided in this embodiment can execute the daylighting prediction model training method provided by any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

Embodiment 4

FIG. 4 is a schematic structural diagram of a lighting prediction device provided in Embodiment 4 of the present invention. The device includes: a lighting coefficient prediction module 410 and a natural lighting illumination prediction module 420 .

The daylighting coefficient prediction module 410 is configured to read the building parameters of a preset type in the building model, input the building parameters into the pre-trained daylighting coefficient prediction model, and obtain the daylighting coefficient corresponding to the building model.

The natural daylighting illuminance prediction module 420 is configured to input the building parameters and the daylighting coefficient into a pre-trained daylighting prediction model to obtain the natural daylighting illuminance corresponding to the building model, wherein the daylighting coefficient prediction model and the daylighting The prediction models are respectively obtained based on the training method of the daylighting prediction model described in any embodiment of the present invention.

The daylighting prediction device provided in this embodiment can execute the daylighting prediction method provided by any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

Embodiment 5

FIG. 5 shows a schematic structural diagram of an electronic device 10 that can be used to implement embodiments of the present invention. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smart phones, wearable devices (eg, helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are by way of example only, and are not intended to limit implementations of the inventions described and/or claimed herein.

As shown in FIG. 5, the electronic device 10 includes at least one processor 11, and a memory, such as a read only memory (ROM) 12, a random access memory (RAM) 13, etc., connected in communication with the at least one processor 11, wherein the memory stores There is a computer program executable by at least one processor, and the processor 11 can be executed according to a computer program stored in a read only memory (ROM) 12 or loaded from a storage unit 18 into a random access memory (RAM) 13. Various appropriate actions and processes are performed. In the RAM 13, various programs and data necessary for the operation of the electronic device 10 can also be stored. The processor 11 , the ROM 12 , and the RAM 13 are connected to each other through a bus 14 . An input/output (I/O) interface 15 is also connected to the bus 14 .

Various components in the electronic device 10 are connected to the I/O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc. etc.; and a communication unit 19, such as a network card, modem, wireless communication transceiver, and the like. The communication unit 19 allows the electronic device 10 to exchange information/data with other devices through a computer network such as the Internet and/or various telecommunication networks.

The processor 11 may be various general and/or special purpose processing components having processing and computing capabilities. Some examples of processors 11 include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), various specialized artificial intelligence (AI) computing chips, various processors that run machine learning model algorithms, digital signal processing processor (DSP), and any suitable processor, controller, microcontroller, etc. The processor 11 performs the various methods and processes described above, such as the daylighting prediction model training method and/or the daylighting prediction method.

In some embodiments, the daylighting prediction model training method and/or the daylighting prediction method may be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as storage unit 18 . In some embodiments, part or all of the computer program may be loaded and/or installed on the electronic device 10 via the ROM 12 and/or the communication unit 19 . When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the daylighting prediction model training method and/or the daylighting prediction method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to perform the daylighting prediction model training method and/or the daylighting prediction method by any other suitable means (eg, by means of firmware).

Various implementations of the systems and techniques described herein above may be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips system (SOC), load programmable logic device (CPLD), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include being implemented in one or more computer programs executable and/or interpretable on a programmable system including at least one programmable processor that The processor, which may be a special purpose or general-purpose programmable processor, may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device an output device.

Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general purpose computer, special purpose computer or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions/operations specified in the flowcharts and/or block diagrams to be carried out. The computer program may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in connection with the instruction execution system, apparatus or device. Computer-readable storage media may include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media would include one or more wire-based electrical connections, portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), fiber optics, compact disk read only memory (CD-ROM), optical storage, magnetic storage, or any suitable combination of the foregoing.

To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having a display device (eg, a CRT (cathode ray tube) or an LCD (liquid crystal display)) for displaying information to the user monitor); and a keyboard and pointing device (eg, a mouse or trackball) through which a user can provide input to the electronic device. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (eg, visual feedback, auditory feedback, or tactile feedback); and can be in any form (including acoustic input, voice input, or tactile input) to receive input from the user.

The systems and techniques described herein may be implemented on a computing system that includes back-end components (eg, as a data server), or a computing system that includes middleware components (eg, an application server), or a computing system that includes front-end components (eg, a user's computer having a graphical user interface or web browser through which a user may interact with implementations of the systems and techniques described herein), or including such backend components, middleware components, Or any combination of front-end components in a computing system. The components of the system may be interconnected by any form or medium of digital data communication (eg, a communication network). Examples of communication networks include: Local Area Networks (LANs), Wide Area Networks (WANs), blockchain networks, and the Internet.

A computing system can include clients and servers. Clients and servers are generally remote from each other and usually interact through a communication network. The relationship of client and server arises by computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also known as a cloud computing server or a cloud host. It is a host product in the cloud computing service system to solve the traditional physical host and VPS services, which are difficult to manage and weak in business scalability. defect.

It should be understood that steps may be reordered, added or deleted using the various forms of flow shown above. For example, the steps described in the present invention can be performed in parallel, sequentially or in different orders, and as long as the desired results of the technical solutions of the present invention can be achieved, no limitation is imposed herein.

The above-mentioned specific embodiments do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Note that the above are only preferred embodiments of the present invention and applied technical principles. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and can also include more other equivalent embodiments without departing from the concept of the present invention. The scope is determined by the scope of the appended claims.

Claims (10)

1. A method for training a lighting prediction model, comprising:

acquiring meteorological data and simulation time data of a target area, transmitting the meteorological data and the simulation time data to a lighting simulation system to set lighting simulation environment parameters, transmitting multiple groups of acquired building parameters to the lighting simulation system to obtain lighting data corresponding to the building parameters, wherein the building parameters and the corresponding lighting data form sample data;

creating at least one initial lighting prediction model, and performing iterative training on each initial lighting prediction model based on training sample data in the sample data to obtain a lighting prediction model corresponding to each initial lighting model;

and verifying the lighting prediction model based on verification sample data in the sample data to obtain at least one evaluation index of each lighting prediction model, and determining a target lighting prediction model based on the at least one evaluation index.

2. The method of claim 1, wherein before transmitting the plurality of sets of acquired building parameters to the lighting simulation system to obtain lighting data corresponding to each set of building parameters, the method further comprises:

acquiring various building house type information in the target area, and transmitting the building house type information to the lighting simulation system so that the lighting simulation system creates various corresponding building models;

the method for acquiring the lighting data comprises the following steps of transmitting the acquired multiple groups of building parameters to the lighting simulation system to obtain the lighting data corresponding to the building parameters, wherein the lighting data comprises:

and transmitting the acquired multiple groups of building parameters to the lighting simulation system so that the lighting simulation system updates the multiple building models based on the building parameters, and respectively performs lighting simulation on the multiple updated building models to obtain lighting data output by the lighting simulation system, wherein a lighting prediction model trained based on the building parameters and the lighting data is used for lighting prediction of the buildings in the target area.

3. The method of claim 1 or 2, wherein the building parameters include sill height, shade size, wall material reflectivity, glass transmittance; the lighting data includes at least one of a lighting coefficient and a natural lighting illuminance;

the iterative training of each initial lighting prediction model based on training sample data in the sample data to obtain a lighting prediction model corresponding to each initial lighting model comprises:

performing iterative training on the initial lighting prediction model based on training sample data formed by taking the building parameters as input data and the lighting coefficient as a label to obtain a lighting coefficient prediction model corresponding to each initial lighting model; and/or the presence of a gas in the atmosphere,

and performing iterative training on the initial lighting prediction model based on training sample data formed by taking the building parameters and the lighting coefficient as input data and taking the natural lighting illumination as a label to obtain a lighting illumination prediction model corresponding to each initial lighting model.

4. The method according to claim 1, wherein the network structure of the at least one initial lighting prediction model is different, or the network depth of the at least one initial lighting prediction model is different;

the iterative training of each initial lighting prediction model based on training sample data in the sample data to obtain a lighting prediction model corresponding to each initial lighting model comprises:

and performing iterative training on the initial lighting prediction model for different training times based on training sample data to obtain the lighting prediction model corresponding to each initial lighting model.

5. The method of claim 1, wherein the evaluation index comprises at least one of: average relative error, maximum relative error, mean square error, and fitness.

6. A lighting prediction method comprising:

reading building parameters of a preset type in a building model, and inputting the building parameters into a pre-trained lighting coefficient prediction model to obtain a lighting coefficient corresponding to the building model;

inputting the building parameters and the lighting coefficient into a lighting prediction model trained in advance to obtain the natural lighting illumination corresponding to the building model, wherein the lighting coefficient prediction model and the lighting prediction model are obtained based on the training method of the lighting prediction model according to any one of claims 1 to 5.

7. A lighting prediction model training device, comprising:

the sample data acquisition module is used for acquiring meteorological data and simulation time data of a target area, transmitting the meteorological data and the simulation time data to a lighting simulation system to set lighting simulation environment parameters, transmitting multiple groups of acquired building parameters to the lighting simulation system, and acquiring lighting data corresponding to the building parameters, wherein the building parameters and the corresponding lighting data form sample data;

the lighting prediction model training module is used for creating at least one initial lighting prediction model, and performing iterative training on each initial lighting prediction model based on training sample data in the sample data to obtain a lighting prediction model corresponding to each initial lighting model;

and the target lighting prediction model determining module is used for verifying the lighting prediction model based on verification sample data in the sample data to obtain at least one evaluation index of each lighting prediction model, and determining the target lighting prediction model based on the at least one evaluation index.

8. A lighting prediction device, comprising:

the lighting coefficient prediction module is used for reading building parameters of preset types in a building model, inputting the building parameters into a lighting coefficient prediction model trained in advance, and obtaining a lighting coefficient corresponding to the building model;

and the natural lighting illumination prediction module is used for inputting the building parameters and the lighting coefficient into a lighting prediction model which is trained in advance to obtain the natural lighting illumination corresponding to the building model, wherein the lighting coefficient prediction model and the lighting prediction model are respectively obtained based on the training method of the lighting prediction model of any one of claims 1 to 5.

9. An electronic device, characterized in that the electronic device comprises:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the light prediction model training method of any one of claims 1-5 and/or the light prediction method of claim 6.

10. A computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions for causing a processor to implement the lighting prediction model training method of any one of claims 1-5 and/or the lighting prediction method of claim 6 when executed.

Images