Human factor illumination regulation and control method with rhythm learning function

Abstract

The invention discloses a human factor illumination regulation and control method with a rhythm learning function, and the method comprises the steps: storing a preset light color value set in a sunrise and sunset time period into a sunrise and sunset time period table, and constructing a neural network model; obtaining the geographic position and time data of the user; obtaining a sunrise and sunset time period corresponding to the geographic position according to the geographic position data; loading a preset light color value set in a sunset time period table according to the sunrise and sunset time periods; acquiring a preset light color value in the sunset time period table in the time period according to the time data; continuously recording human factor light color values input by the user in time data every day, and establishing mapping of the preset light color values and the human factor light color values as a light color sample set; when the sample size of the light color sample set reaches a sample threshold value, assigning a neural network model to the light color sample set; the neural network model updates the trained predicted light color data to a sunrise and sunset period table to form a new preset light color value; and controlling the LED illumination system to perform illumination regulation and control according to the new preset light color value.

Description

Technical Field

[0001] The present invention relates to the field of human-caused lighting technology and engineering, and in particular to a human-caused lighting control method with a rhythm learning function. Background Art

[0002] In recent years, smart lighting furniture has attracted much attention due to its ability to dynamically adjust lighting parameters according to environmental changes and user needs. With the rapid development of AI algorithms and IoT technology, the research on smart lighting furniture has entered a new stage. At present, the human body's adaptation to the light environment is closely related to health, and traditional lighting systems are difficult to meet personalized health needs.

[0003] According to the China Engineering Construction Standardization Association standard T / CECS1424-2023, lighting control needs to be designed. However, among the numerous lighting control designs, there are few cases where control strategy design can be achieved. Especially in home design, even if the design is completed and meets the design standards, the effect is still limited due to different habits and preferences of different people.

[0004] Therefore, there is an urgent need for a human-oriented intelligent lighting control method that combines an AI model and can meet the lighting needs of different users to solve the above problems. Summary of the invention

[0005] The present invention aims to preset the dynamic parameters of theoretical rhythms in combination with the AI ​​model through the learning function of different human preferences, so that when people enter any lighting space, the artificial light generated is a scene that has been theoretically demonstrated and learned many times in accordance with personal preferences. The patent of this invention not only helps to promote technological innovation in the field of smart homes, but also provides people with a healthier, more comfortable and intelligent lighting environment.

[0006] First, a preset light color parameter for theoretical rhythmic health is embedded in the sunrise and sunset time period table in advance. The light color parameter is to divide different time periods according to sunrise and sunset, and preset divisions are made for color temperature and illumination. Then, the longitude and latitude data and time data in the lighting space where the user is located are obtained through the positioning system. The sunrise and sunset time data in the lighting space where the user is located are calculated through the longitude and latitude data and time data. When the user first enters the lighting space, the color temperature value and illumination value output by the LED lighting system are the healthy preset color temperature value data and preset illumination value data corresponding to the sunrise and sunset time data in the sunrise and sunset time period table. If the user is not used to the current preset color temperature value and preset illumination value, they can be adjusted manually. The data recording system will continuously record the human-caused color temperature value and human-caused illumination value entered by the user at the current sunrise and sunset time every day, and establish the human-caused color temperature value at the current sunrise and sunset time and the current sunrise and sunset time. The mapping of preset color temperature values ​​within the segment, and the mapping of the human-factor illuminance value at the current sunrise and sunset time and the preset illuminance value within the current sunrise and sunset time period are established, the human-factor color temperature value and the corresponding preset color temperature value are used as the first sample data, and the human-factor illuminance value and the corresponding preset illuminance value are used as the second sample data. When the number of collected human-factor color temperature values ​​exceeds the color temperature threshold and / or the number of human-factor illuminance values ​​exceeds the illuminance threshold, the first sample data is input into the color temperature neural network model for training, and the second sample data is input into the illuminance neural network model for training. After the color temperature neural network model is trained, the final predicted color temperature value is output, and after the illuminance neural network model is trained, the final predicted illuminance value is output; at the current sunrise and sunset time, the LED lighting system uses the final predicted color temperature value and the final predicted illuminance value for lighting; the final predicted color temperature value and the final predicted illuminance value are biased towards the human-factor color temperature value and the human-factor illuminance value preferred by the user.

[0007] The preset color temperature value and the preset illuminance value in the present invention are healthy color temperature values ​​and illuminance values ​​that have been scientifically demonstrated. Appropriate preset color temperature values ​​and preset illuminance values ​​will be displayed for lighting at different sunrise and sunset time periods. The human body can increase the level of serotonin in the body under the irradiation of healthy color temperature values ​​and illuminance values, improve mood and reduce the risk of depression; in addition, using soft and healthy color temperature values ​​and illuminance values ​​for irradiation in the middle of the night can help promote the secretion of melatonin in the human body and improve sleep; furthermore, using appropriate color temperature values ​​and illuminance values ​​for irradiation during the day can not only reduce eye fatigue and protect eyesight, but also enhance people's attention, thereby improving work and learning efficiency. Secondly, the present invention will make personalized adjustments according to different user-defined needs. When the user manually adjusts the color temperature value and illuminance value of the LED lighting system at the current sunrise time every day, and the number of adjustments exceeds a certain threshold, through the training of the neural network model, the color temperature value and illuminance value output by the LED lighting system will be biased towards the user's preferred human color temperature value and human illuminance value.

[0008] The present invention is mainly achieved through the following technical solutions.

[0009] The present invention provides a human-induced lighting control method with a rhythm learning function, the method comprising: Pre-establish a set of preset light color values ​​corresponding to different time periods within the sunrise and sunset time period, store them in a sunrise and sunset time period table, and construct a neural network model; S1: Use the positioning system to obtain the geographical location data and time data of the user in the lighting space; S2: Obtaining the sunrise and sunset time period corresponding to the geographic location data according to the geographic location data; S3: Loading a preset light color value set in the sunrise and sunset time period table according to the sunrise and sunset time period; S4: acquiring preset light color value data in a sunrise and sunset time period table within the time period of the time data according to the time data; S5: Use a data recording system to continuously record the human-induced light color data input by the user in the time data every day, and establish a mapping between the preset light color value data and the human-induced light color value data as a light color sample data set; S6: when the number of samples of the light color sample data set reaches the sample threshold, assigning the light color sample data set to the neural network model training; S7: The neural network model updates the final predicted light color data after training to the sunrise and sunset time period table within the time period of the time data to form a new preset light color value; S8: Control the LED lighting system to perform lighting regulation according to the new preset light color value.

[0010] Furthermore, the preset light color value set is a preset color temperature value and a preset illuminance value in different sunrise and sunset time periods.

[0011] Furthermore, in step S1, the geographical location data is the longitude and latitude data of the lighting space where the user is located, and the time data is the current moment data of the lighting space where the user is located.

[0012] Furthermore, the preset light color value data in step S5 are the preset color temperature value data and preset illuminance value data of the system in the current sunrise and sunset time period in the sunrise and sunset time period table, and the human-related light color data are the human-related color temperature value data and human-related illuminance value data preferred by the user at the current sunrise and sunset time.

[0013] Further, step S5 includes: Use the data recording system to continuously record the human-factor color temperature value data and human-factor illuminance value data input by the user at the current sunrise and sunset time every day; Match the human color temperature value data at the current sunrise and sunset time with the preset color temperature value data in the corresponding current sunrise and sunset time period and establish a mapping relationship; Match the human factor illumination value data at the current sunrise and sunset time with the preset illumination value data in the corresponding current sunrise and sunset time period and establish a mapping relationship; The matched human color temperature value data and the preset color temperature value data are used as a first sample data set; The matched human factor illumination value data and the preset illumination value data are used as the second sample data set.

[0014] Furthermore, in step S6, the sample threshold includes a color temperature sample threshold and an illuminance sample threshold, and the neural network model includes a color temperature DNN model and an illuminance DNN model.

[0015] Furthermore, the color temperature DNN model training process in step S6 includes: When the number of samples in the first sample data set reaches the color temperature sample threshold, the preset color temperature value data set is combined as the input feature X, and the human color temperature value data set is combined as the label Y; Assign the input feature X and label Y to the input layer of the color temperature DNN model for standardization and normalization; The hidden layer receives the input feature X processed by each input layer, and uses the first weighted input calculation formula to calculate the weighted input of each hidden layer; Performing nonlinear transformation on the weighted input of each hidden layer through the first activation function to obtain the output value of each hidden layer neuron; The output layer receives the output value of each hidden layer neuron and uses the weighted input calculation formula to calculate the weighted input of each output layer; The predicted color temperature value data is calculated by the second activation function, and the predicted color temperature value data is evaluated using the loss function until the evaluation is qualified; After the evaluation is qualified, the prediction ability of the color temperature DNN model is optimized; The output layer outputs the qualified predicted color temperature value data as the final color temperature value.

[0016] Furthermore, the first weighted input calculation formula is: ; in, is the weighted input of the jth neuron; is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer, is the preset color temperature value data of the i-th neuron in the input layer, is the bias of the jth neuron in the hidden layer; The formula for the first activation function is: ; in, is the output value of the jth neuron in the hidden layer; The first weighted input calculation formula is: ; in, is the weighted input to each output layer, is the weight from the ith neuron to each output layer neuron, is the output value of the i-th neuron in the hidden layer, is the bias of each output layer neuron; The calculation formula of the second activation function is: ; in, To predict the color temperature value; The loss function calculation formula is:

[0017] Where n is the number of samples, is the color temperature value data of the i-th person; The gradient is calculated through the back-propagation algorithm, the model parameters are updated, and the prediction ability of the color temperature DNN model is optimized.

[0018] Furthermore, the illumination DNN model training process in step S6 includes: When the number of samples in the second sample data set reaches the illumination sample threshold, the preset illumination value data set is combined as the input feature X, and the human factor illumination value data set is combined as the label Y; Assign the input feature X and label Y to the input layer of the illumination DNN model for standardization and normalization; The hidden layer receives the input feature X processed by each input layer, and uses the first weighted input calculation formula to calculate the weighted input of each hidden layer; Performing nonlinear transformation on the weighted input of each hidden layer through the first activation function to obtain the output value of each hidden layer neuron; The output layer receives the output value of each hidden layer neuron and uses the weighted input calculation formula to calculate the weighted input of each output layer; The predicted illumination value data is calculated by the second activation function, and the predicted illumination value data is evaluated by using the loss function until the evaluation is qualified; Evaluate the prediction capability of the optimized illumination DNN model after passing the evaluation; The output layer outputs the qualified predicted illuminance value data as the final illuminance value.

[0019] Furthermore, the first weighted input calculation formula is: ; in, is the preset illumination value data of the i-th neuron in the input layer; The calculation formula of the second activation function is: ; in, is the predicted illumination value; The loss function calculation formula is:

[0020] Where n is the number of samples, is the illuminance value data of the i-th person; The gradient is calculated through the back-propagation algorithm, the model parameters are updated, and the prediction ability of the illumination DNN model is optimized.

[0021] In summary, the present invention has the following beneficial effects compared with the prior art: The preset color temperature value and the preset illuminance value in the present invention are scientifically proven healthy color temperature values ​​and illuminance values. The simulated illumination by the healthy color temperature value and illuminance value has significant positive effects on the human body, including improving mood, improving sleep, enhancing attention, adjusting physiological rhythms, promoting social interaction, and reducing eye fatigue. Appropriate lighting conditions can affect the biological rhythms in the body, helping people to better adapt to daily life, thereby improving overall physical and mental health; the present invention will make personalized adjustments based on different user-defined needs. When the user manually adjusts the color temperature value and illuminance value of the LED lighting system at the current sunrise time every day, and the number of adjustments exceeds a certain threshold, through the training of the neural network model, the color temperature value and illuminance value output by the LED lighting system will be biased towards the user's preferred human color temperature value and human illuminance value, thereby meeting the user's personalized needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flow chart of a human-caused lighting control method with rhythm learning function; Figure 2 Flowchart for training color temperature values ​​for the color temperature DNN model; Figure 3 Flowchart for training illuminance values ​​for the illuminance DNN model; DETAILED DESCRIPTION

[0023] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0024] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0025] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0026] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0027] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may include that the first and second features are in direct contact, or may include that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, a first feature being "above", "above" and "above" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply indicates that the first feature is higher in level than the second feature. A first feature being "below", "below" and "below" a second feature includes that the first feature is directly below and obliquely below the second feature, or simply indicates that the first feature is lower in level than the second feature.

[0028] In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the fact that ordinary technicians in the field can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of the present invention.

[0029] In the following description, suffixes such as "module", "component", "assembly" or "unit" are used only to facilitate the description of the present invention and have no specific meanings. Therefore, they can be used in combination.

[0030] The present invention will be further described in detail below through specific implementation modes in conjunction with the accompanying drawings.

[0031] According to one embodiment of the invention, Figure 1~Figure 3 As shown, a human-caused lighting control method with a rhythm learning function includes: Pre-establish a set of preset light color values ​​corresponding to different time periods within the sunrise and sunset time period, store them in a sunrise and sunset time period table, and construct a neural network model; S1: Use the positioning system to obtain the geographical location data and time data of the user in the lighting space; S2: Obtaining the sunrise and sunset time period corresponding to the geographical location data according to the geographical location data; S3: Loading a preset light color value set in the sunrise and sunset time period table according to the sunrise and sunset time period; S4: acquiring preset light color value data in the sunrise and sunset time period table within the time period of the time data according to the time data; S5: using a data recording system to continuously record the human-induced light color data input by the user at the time data every day, and establishing a mapping between the preset light color value data and the human-induced light color value data as a light color sample data set; S6: when the number of samples of the light color sample data set reaches a sample threshold, assigning the light color sample data set to the neural network model training; S7: The neural network model updates the final predicted light color data after training to the sunrise and sunset time period table within the time period of the time data to form a new preset light color value; S8: Control the LED lighting system to perform lighting regulation according to the new preset light color value.

[0032] When this embodiment is implemented, for example, in summer in my country, the earliest city to get dark is Mohe City, Heilongjiang Province, where the sun sets at 5 pm and it starts to get dark, and the latest city to get dark is Kashgar City, Xinjiang, where it starts to get dark at 10:30 pm; if the user in these two areas has not input human-caused color temperature value data or human-caused illuminance value data during this period, or the amount of human-caused color temperature value data input is less than the color temperature sample threshold and the amount of human-caused illuminance value data is less than the illuminance sample threshold, the color temperature value and illuminance value output by the LED lighting system are the preset color temperature value and preset illuminance value data in the sunrise and sunset table for which the neural network model has not participated in the training during this period; When the user has input human-factor color temperature value data or human-factor illuminance value data within the time period, and the data volume of the human-factor color temperature value data reaches the color temperature sample threshold and / or the data volume of the human-factor illuminance value data reaches the illuminance sample threshold, the color temperature value and illuminance value output by the LED lighting system are the preset color temperature value and preset illuminance value data in the sunrise and sunset table updated after the neural network model is trained within the time period.

[0033] For example, the sunrise and sunset table for which the neural network model has not been trained is shown in Table 1: time Preset color temperature / K Preset illumination / Lux Sunrise / Sunset 2000 15 20 minutes after sunrise / before sunset 2100 20 30 minutes after sunrise / before sunset 2400 28 40 minutes after sunrise / before sunset 2900 40 1 hour after sunrise / before sunset 4000 55 late at night 2000 2 Table 1 When the current time in Mohe City, Heilongjiang Province is 4:30 p.m., the positioning system determines that the area is 30 minutes away from sunset, which corresponds to the 30 minutes before sunset time period in the sunrise and sunset table. In the sunrise and sunset time period corresponding to this time, the user has not entered the human-caused color temperature value or the human-caused illuminance value, or the amount of human-caused color temperature value data entered is less than the color temperature sample threshold and the amount of human-caused illuminance value data is less than the illuminance sample threshold. The color temperature value and illuminance value output by the LED lighting system are 2400K and 28Lux 30 minutes before sunset in Table 1. When the time in Mohe City is 5 p.m., the positioning system determines that the area is in sunset, which corresponds to the sunset time period in the sunrise and sunset table. In the sunrise and sunset time period corresponding to this time, the user has not entered the human-caused color temperature value or the human-caused illuminance value, or the amount of human-caused color temperature value data entered is less than the color temperature sample threshold and the human-caused illuminance value The data volume is less than the illuminance sample threshold, and the color temperature and illuminance values ​​output by the LED lighting system are 2000K and 15Lux at sunset in Table 1; if the user modifies the current color temperature to 2000K and the current illuminance to 25Lux at 4:30 p.m. on the same day, and the color temperature and illuminance values ​​input by the user during the sunrise and sunset time period corresponding to that moment do not exceed the corresponding sample threshold; since the color temperature and illuminance values ​​input by the user do not exceed the corresponding sample threshold, the LED lighting system only outputs a color temperature of 2000K and an illuminance of 25lux for lighting during the sunrise and sunset time period corresponding to the current sunrise and sunset moment of the same day. When the user is at 4:30 p.m. the next day, the color temperature output by the LED lighting system during this sunrise and sunset time period is still 2400K, which is 30 minutes before sunset, and the illuminance value is still 28Lux, which is 30 minutes before sunset, in Table 1.

[0034] For another example, when the current time in Kashgar, Xinjiang is 10 o'clock in the evening, the positioning system determines that the area is 30 minutes away from sunset, and the user has not entered the human-caused color temperature value or the human-caused illuminance value in the sunrise and sunset time period corresponding to this time, or the amount of human-caused color temperature value data entered is less than the color temperature sample threshold and the amount of human-caused illuminance value data is less than the illuminance sample threshold, and the color temperature value and illuminance value output by the LED lighting system are 2400K and 28Lux 30 minutes before sunset in Table 1. When the time in Kashgar is 10:30 in the evening, the positioning system determines that the area is at sunset, and the user has not entered the human-caused color temperature value or the human-caused illuminance value in the sunrise and sunset time period corresponding to this time, or the amount of human-caused color temperature value data entered is less than the color temperature sample threshold and the amount of human-caused illuminance value data is less than the illuminance sample threshold, and the color temperature value and illuminance value output by the user in the LED lighting system are 2000K and 15Lux at sunset in Table 1.

[0035] When this embodiment is implemented, the data recording system will continuously record the human-induced color temperature value data and human-induced illuminance value data input by the user every day and store them in the sunrise and sunset mapping table.

[0036] For example, as shown in Table 2, the table reflects the human-caused color temperature value data and human-caused illuminance value data entered by users in Mohe City every day in the 20-minute period before sunset and stored in the sunrise and sunset mapping table: time Preset color temperature / K Preset illumination / Lux Human color temperature / K Human factor illumination / Lux 20 minutes after sunrise / before sunset 2100 20 2200 15 20 minutes after sunrise / before sunset 2100 20 2100 20 20 minutes after sunrise / before sunset 2100 20 2300 15 20 minutes after sunrise / before sunset 2100 20 2600 30 20 minutes after sunrise / before sunset 2100 20 2100 20 20 minutes after sunrise / before sunset 2100 20 2200 15 20 minutes after sunrise / before sunset 2100 20 2200 15 20 minutes after sunrise / before sunset 2100 20 2200 15 20 minutes after sunrise / before sunset 2100 20 2200 15 20 minutes after sunrise / before sunset 2100 20 2200 15 Table 2 When the amount of human-factor color temperature data input by users in Mohe City within the 20 minutes before sunset reaches the color temperature sample threshold or the amount of human-factor illuminance data reaches the illuminance sample threshold; the system will send the preset color temperature data, preset illuminance data, human-factor color temperature data and human-factor illuminance data within the sunrise and sunset time period to the neural network model for training, and the trained predicted color temperature data and predicted illuminance data will update and replace the original predicted color temperature value and predicted illuminance value within the 20 minutes before sunset time period in the sunrise and sunset table, as shown in Table 3 below: time Preset color temperature / K Preset illumination / Lux Sunrise / Sunset 2000 15 20 minutes after sunrise / before sunset 2200 15 30 minutes after sunrise / before sunset 2400 28 40 minutes after sunrise / before sunset 2900 40 1 hour after sunrise / before sunset 4000 55 late at night 2000 2 Table 3 The preset color temperature value in the time period 20 minutes before sunset is updated to 2200K, and the preset illuminance value is updated to 15Lux. At this time, the color temperature value and illuminance value output by the LED lighting system in the sunrise and sunset time period are 2200K and 15Lux in the sunrise and sunset table in Table 3.

[0037] Furthermore, the preset light color value set is a preset color temperature value and a preset illuminance value in different sunrise and sunset time periods.

[0038] Furthermore, the geographic location data in step S1 is the longitude and latitude data of the lighting space where the user is located, and the time data is the current moment data of the lighting space where the user is located.

[0039] When this embodiment is implemented, the longitude and latitude data and time data of the user's area are obtained through GPS or other positioning systems, and the longitude and latitude data and time data are sent to an API interface for calculating the sunrise and sunset times. The API interface contains an astronomical formula for calculating the sunrise and sunset times. This formula is a commonly used formula for calculating astronomy and will not be described here. After the API interface completes the calculation, the sunrise and sunset times in the lighting space where the current user is located are obtained.

[0040] Furthermore, the preset light color value data are the preset color temperature value data and preset illuminance value data of the system in the current sunrise and sunset time period in the sunrise and sunset time period table, and the human-factor light color data are the human-factor color temperature value data and human-factor illuminance value data preferred by the user at the current sunrise and sunset time.

[0041] Further, a data recording system is used to continuously record the human-caused color temperature value data and human-caused illuminance value data input by the user at the current sunrise and sunset time every day; Matching the human-induced color temperature value data at the current sunrise and sunset time with the preset color temperature value data in the corresponding current sunrise and sunset time period and establishing a mapping relationship; Matching the human factor illumination value data at the current sunrise and sunset time with the preset illumination value data in the corresponding current sunrise and sunset time period and establishing a mapping relationship; The matched human color temperature value data and the preset color temperature value data are used as a first sample data set; The matched human factor illumination value data and the preset illumination value data are used as the second sample data set.

[0042] When this embodiment is implemented, the human-factor color temperature value data input by the user, the preset color temperature value data for the sunrise and sunset time periods corresponding to the human-factor color temperature value data, the human-factor illuminance value data, and the preset illuminance value data for the sunrise and sunset time periods corresponding to the human-factor illuminance value data are all stored in the sunrise and sunset mapping table.

[0043] Furthermore, the sample threshold in step S6 includes a color temperature sample threshold and an illuminance sample threshold, and the neural network model includes a color temperature DNN model and an illuminance DNN model.

[0044] When this embodiment is implemented, the color temperature sample threshold and the illuminance sample threshold are both pre-stored in the system. In this embodiment, the color temperature sample threshold and the illuminance sample threshold are both set to 10.

[0045] Furthermore, the color temperature DNN model training process in step S6 includes: When the number of samples in the first sample data set reaches the color temperature sample threshold, the preset color temperature value data set is combined as the input feature X, and the human color temperature value data set is combined as the label Y; Assigning the input feature X and label Y to the input layer of the color temperature DNN model for standardization and normalization; The hidden layer receives the input feature X processed by each input layer, and uses the first weighted input calculation formula to calculate the weighted input of each hidden layer; Performing nonlinear transformation on the weighted input of each hidden layer through the first activation function to obtain the output value of each hidden layer neuron; The output layer receives the output value of each hidden layer neuron and uses the weighted input calculation formula to calculate the weighted input of each output layer; Calculate predicted color temperature value data by using a second activation function, and evaluate the predicted color temperature value data using a loss function until the evaluation is qualified; After the evaluation is qualified, the prediction capability of the color temperature DNN model is optimized; The output layer outputs the qualified predicted color temperature value data as the final color temperature value.

[0046] When this embodiment is implemented, the color temperature DNN model training process may also be: Assigning a preset color temperature value dataset as an input feature X and a human-induced color temperature value dataset as a label Y to an input layer of the color temperature DNN model; The input layer standardizes and normalizes the input feature X and outputs it to the convolution layer; The convolution layer uses multiple convolution kernels to slide on the input data, perform convolution operations, and generate feature maps to output to the activation layer; The activation layer uses the activation function to activate the feature map, and then transfers the activated feature map to the pooling layer; The pooling layer performs a pooling operation on the activated feature map and transmits the pooled feature map to the fully connected layer; The fully connected layer converts the pooled high-dimensional feature map into a one-dimensional vector, performs high-level feature combination, and outputs the output vector to the output layer; The output layer converts the output vector into predicted color temperature value data through the activation function, and compares and evaluates it with the label Y until it is qualified; After the evaluation is qualified, the prediction capability of the color temperature DNN model is optimized; The output layer outputs the predicted color temperature value that meets the evaluation criteria as the final color temperature value.

[0047] Furthermore, the first weighted input calculation formula is: ; in, is the weighted input of the jth neuron; is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer, is the preset color temperature value data of the i-th neuron in the input layer, is the bias of the jth neuron in the hidden layer; The formula of the first activation function is: ; in, is the output value of the jth neuron in the hidden layer; The first weighted input calculation formula is: ; in, is the weighted input to each output layer, is the weight from the ith neuron to each output layer neuron, is the output value of the i-th neuron in the hidden layer, is the bias of each output layer neuron; The calculation formula of the second activation function is: ; in, To predict the color temperature value; The loss function calculation formula is:

[0048] Where n is the number of samples, is the color temperature value data of the i-th person; The gradient is calculated through the back propagation algorithm, the model parameters are updated, and the prediction ability of the color temperature DNN model is optimized.

[0049] Furthermore, the illumination DNN model training process in step S6 includes: When the number of samples in the second sample data set reaches the illumination sample threshold, the preset illumination value data set is combined as the input feature X, and the human factor illumination value data set is combined as the label Y; Assigning the input feature X and label Y to the input layer of the illumination DNN model for standardization and normalization; The hidden layer receives the input feature X processed by each input layer, and uses the first weighted input calculation formula to calculate the weighted input of each hidden layer; Performing nonlinear transformation on the weighted input of each hidden layer through the first activation function to obtain the output value of each hidden layer neuron; The output layer receives the output value of each hidden layer neuron and uses the weighted input calculation formula to calculate the weighted input of each output layer; Calculate predicted illuminance value data by using a second activation function, and evaluate the predicted illuminance value data using a loss function until the evaluation is qualified; After the evaluation is qualified, the prediction capability of the illumination DNN model is optimized; The output layer outputs the qualified predicted illuminance value data as the final illuminance value.

[0050] When this embodiment is implemented, the training process of the illumination DNN model may also be: Assigning a preset illuminance value dataset as an input feature X and a human factor illuminance value dataset as a label Y to an input layer of the illuminance DNN model; The input layer standardizes and normalizes the input feature X and outputs it to the convolution layer; The convolution layer uses multiple convolution kernels to slide on the input data, perform convolution operations, and generate feature maps to output to the activation layer; The activation layer uses the activation function to activate the feature map, and then transfers the activated feature map to the pooling layer; The pooling layer performs a pooling operation on the activated feature map and transmits the pooled feature map to the fully connected layer; The fully connected layer converts the pooled high-dimensional feature map into a one-dimensional vector, performs high-level feature combination, and outputs the output vector to the output layer; The output layer converts the output vector into predicted illumination value data through the activation function, and compares and evaluates it with the label Y until it is qualified; After the evaluation is qualified, the prediction capability of the illumination DNN model is optimized; The output layer outputs the qualified predicted illuminance value as the final illuminance value.

[0051] Furthermore, the first weighted input calculation formula is: ; in, is the preset illumination value data of the i-th neuron in the input layer; The calculation formula of the second activation function is: ; in, is the predicted illumination value; The loss function calculation formula is:

[0052] Where n is the number of samples, is the illuminance value data of the i-th person; The gradient is calculated through the back-propagation algorithm, the model parameters are updated, and the prediction ability of the illumination DNN model is optimized.

[0053] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A human-induced lighting control method with rhythm learning function, characterized in that: The method comprises: Pre-establish a set of preset light color values ​​corresponding to different time periods within the sunrise and sunset time period, store them in a sunrise and sunset time period table, and construct a neural network model; S1: Use the positioning system to obtain the geographical location data and time data of the user in the lighting space; S2: Obtaining the sunrise and sunset time period corresponding to the geographical location data according to the geographical location data; S3: Loading a preset light color value set in the sunrise and sunset time period table according to the sunrise and sunset time period; S4: acquiring preset light color value data in the sunrise and sunset time period table within the time period of the time data according to the time data; S5: using a data recording system to continuously record the human-induced light color data input by the user at the time data every day, and establishing a mapping between the preset light color value data and the human-induced light color value data as a light color sample data set; S6: when the number of samples of the light color sample data set reaches a sample threshold, assigning the light color sample data set to the neural network model training; S7: The neural network model updates the final predicted light color data after training to the sunrise and sunset time period table within the time period of the time data to form a new preset light color value; S8: Control the LED lighting system to perform lighting regulation according to the new preset light color value.

2. According to claim 1, a human-induced lighting control method with rhythm learning function is characterized in that: The preset light color value set includes preset color temperature values ​​and preset illuminance values ​​in different sunrise and sunset time periods.

3. According to claim 1, a human-induced lighting control method with rhythm learning function is characterized in that: The geographic location data in step S1 is the longitude and latitude data of the lighting space where the user is located, and the time data is the current moment data of the lighting space where the user is located.

4. According to claim 1, a human-induced lighting control method with rhythm learning function is characterized in that: In step S5, the preset light color value data are the preset color temperature value data and preset illuminance value data of the system in the current sunrise and sunset time period in the sunrise and sunset time period table, and the human-factor light color data are the human-factor color temperature value data and human-factor illuminance value data preferred by the user at the current sunrise and sunset time.

5. According to claim 4, a human-induced lighting control method with rhythm learning function is characterized in that: The step S5 comprises: Use the data recording system to continuously record the human-factor color temperature value data and human-factor illuminance value data input by the user at the current sunrise and sunset time every day; Matching the human-induced color temperature value data at the current sunrise and sunset time with the preset color temperature value data in the corresponding current sunrise and sunset time period and establishing a mapping relationship; Matching the human factor illumination value data at the current sunrise and sunset time with the preset illumination value data in the corresponding current sunrise and sunset time period and establishing a mapping relationship; The matched human color temperature value data and the preset color temperature value data are used as a first sample data set; The matched human factor illumination value data and the preset illumination value data are used as the second sample data set.

6. According to claim 5, a human-induced lighting control method with rhythm learning function is characterized in that: The sample threshold in step S6 includes a color temperature sample threshold and an illuminance sample threshold, and the neural network model includes a color temperature DNN model and an illuminance DNN model.

7. The human-caused lighting control method with rhythm learning function according to claim 6, characterized in that: The color temperature DNN model training process in step S6 includes: When the number of samples in the first sample data set reaches the color temperature sample threshold, the preset color temperature value data set is combined as the input feature X, and the human color temperature value data set is combined as the label Y; Assigning the input feature X and label Y to the input layer of the color temperature DNN model for standardization and normalization; The hidden layer receives the input feature X processed by each input layer, and uses the first weighted input calculation formula to calculate the weighted input of each hidden layer; Performing nonlinear transformation on the weighted input of each hidden layer through the first activation function to obtain the output value of each hidden layer neuron; The output layer receives the output value of each hidden layer neuron and uses the weighted input calculation formula to calculate the weighted input of each output layer; The predicted color temperature value data is calculated by the second activation function, and the predicted color temperature value data is evaluated using the loss function until the evaluation is qualified; After the evaluation is qualified, the prediction capability of the color temperature DNN model is optimized; The output layer outputs the qualified predicted color temperature value data as the final color temperature value.

8. The human-caused lighting control method with rhythm learning function according to claim 6, characterized in that: The first weighted input calculation formula is: ； in, is the weighted input of the jth neuron; is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer, is the preset color temperature value data of the i-th neuron in the input layer, is the bias of the jth neuron in the hidden layer; The formula of the first activation function is: ； in, is the output value of the jth neuron in the hidden layer; The first weighted input calculation formula is: ； in, is the weighted input to each output layer, is the weight from the ith neuron to each output layer neuron, is the output value of the i-th neuron in the hidden layer, is the bias of each output layer neuron; The calculation formula of the second activation function is: ； in, To predict the color temperature value; The loss function calculation formula is: Where n is the number of samples, is the color temperature value data of the i-th person; The gradient is calculated through the back propagation algorithm, the model parameters are updated, and the prediction ability of the color temperature DNN model is optimized.

9. The human-caused lighting control method with rhythm learning function according to claim 6, characterized in that: The illumination DNN model training process in step S6 includes: When the number of samples in the second sample data set reaches the illumination sample threshold, the preset illumination value data set is combined as the input feature X, and the human factor illumination value data set is combined as the label Y; Assigning the input feature X and label Y to the input layer of the illumination DNN model for standardization and normalization; The hidden layer receives the input feature X processed by each input layer, and uses the first weighted input calculation formula to calculate the weighted input of each hidden layer; Performing nonlinear transformation on the weighted input of each hidden layer through the first activation function to obtain the output value of each hidden layer neuron; The output layer receives the output value of each hidden layer neuron and uses the weighted input calculation formula to calculate the weighted input of each output layer; The predicted illumination value data is calculated by the second activation function, and the predicted illumination value data is evaluated by using the loss function until the evaluation is qualified; After the evaluation is qualified, the prediction capability of the illumination DNN model is optimized; The output layer outputs the qualified predicted illuminance value data as the final illuminance value.

10. The human-caused lighting control method with rhythm learning function according to claim 9, characterized in that: The first weighted input calculation formula is: ； in, is the preset illumination value data of the i-th neuron in the input layer; The calculation formula of the second activation function is: ； in, is the predicted illumination value; The loss function calculation formula is: Where n is the number of samples, is the illuminance value data of the i-th person; The gradient is calculated through the back-propagation algorithm, the model parameters are updated, and the prediction ability of the illumination DNN model is optimized.

Images