Solar street lamp intelligent cluster control system based on Internet of Things

Abstract

The invention relates to the technical field of intelligent control, in particular to a solar street lamp intelligent cluster control system based on the Internet of Things, which comprises a data acquisition module, an area monitoring module, an illumination duration prediction module, a sudden change judgment module and an illumination scheduling module. The data acquisition module obtains real-time working state data of the street lamps and environment images and pedestrian flow images of a region, and generates a region state matrix. The area monitoring module extracts the lighting demand state of each area. The illumination duration prediction module predicts average illumination duration through a feedforward neural network in combination with real-time data and night illumination requirements, and generates a first prediction model. The sudden change judgment module calculates an energy-saving value based on the illumination duration output by the first prediction model and a preset threshold value, positions an energy-saving sudden change node through an energy-saving value change slope, and outputs an energy-saving sudden change judgment score. And the illumination scheduling module dynamically adjusts the brightness and the on-off state of the street lamp according to the energy-saving sudden change judgment score, so that accurate illumination control and energy optimization are realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to an Internet of Things-based solar street lamp intelligent cluster control system. Background Art

[0002] As a green and environmentally friendly lighting solution, solar street lights have been widely used in urban areas, parks, and communities in recent years. Traditional solar street light systems typically use fixed brightness or simple adjustments based on ambient light intensity, making it difficult to dynamically adjust lighting strategies based on actual needs. With the development of the Internet of Things (IoT) and artificial intelligence (AI), smart solar street light systems are emerging. These systems use sensors to monitor real-time data such as pedestrian flow and ambient light levels, automatically adjusting brightness based on demand and improving energy efficiency.

[0003] However, in actual applications, when the flow of people in a certain area suddenly increases, the existing smart solar street light system will increase the brightness accordingly to meet the lighting needs. However, this process often lacks refined power consumption control, resulting in excessive power consumption and may even cause battery over-discharge, affecting the normal operation of the street light. Due to the lack of dynamic monitoring and optimization of power consumption, the system is prone to excessive energy consumption during peak traffic periods. Summary of the Invention

[0004] The main purpose of the present invention is to provide an intelligent cluster control system for solar street lights based on the Internet of Things, aiming to solve the technical problem that the existing solar street light system lacks refined power consumption control when the flow of people increases, resulting in excessive power consumption and energy waste.

[0005] To achieve the above-mentioned object, the present invention provides an intelligent cluster control system for solar street lamps based on the Internet of Things, the system comprising: a data acquisition module, an area monitoring module, a lighting duration prediction module, an abrupt change judgment module, and a lighting scheduling module;

[0006] The data acquisition module is connected to the area monitoring module and the lighting duration prediction module respectively, the area monitoring module is connected to the lighting duration prediction module and the lighting scheduling module respectively, the lighting duration prediction module is connected to the sudden change judgment module, and the sudden change judgment module is connected to the lighting scheduling module;

[0007] A data acquisition module is used to obtain real-time operating status data of solar street lights through an Internet of Things sensor network. The real-time operating status data includes ambient light intensity, remaining battery power, lamp switch status, and fault information; obtain environmental images and pedestrian flow images of the area where the street lights are located, and generate a regional status matrix based on the environmental images and pedestrian flow images;

[0008] The regional monitoring module is used to extract the lighting demand status of each area from the regional state matrix based on a convolutional neural network;

[0009] The lighting duration prediction module is used to calculate the lighting duration of the street lights based on the real-time working status data of the solar street lights and the nighttime lighting demand. The lighting demand status of each area is used as the input of the feedforward neural network, and the average lighting duration is used as the output to train and generate a first prediction model;

[0010] a sudden change judgment module, configured to calculate an energy saving value based on the average lighting duration predicted and output by the first prediction model and a preset lighting threshold, locate an energy saving sudden change node by the slope of the energy saving value, use the lighting demand state of each area corresponding to the energy saving sudden change node as input to a BP neural network, and use the energy saving sudden change judgment score as output to train and generate a second prediction model;

[0011] The lighting scheduling module is used to adjust the lighting brightness and switch status of the solar street lamp in real time according to the real-time lighting demand status of each area and the energy-saving sudden change judgment score output by the second prediction model.

[0012] Optionally, acquiring an environment image and a pedestrian flow image of an area where a street lamp is located, and generating an area state matrix based on the environment image and the pedestrian flow image includes:

[0013] Set up several real-time image acquisition devices in the area where the street lights are located to acquire environmental images and pedestrian flow images;

[0014] Splicing the environment images to construct a two-dimensional regional model, and generating a regional matrix according to the two-dimensional regional model;

[0015] The distribution information of pedestrians in the two-dimensional model of the area is obtained according to the pedestrian flow image, and the area matrix is ​​mapped according to the position information in the distribution information. The pedestrian density is preset to the same value and the area matrix parameters are replaced to generate the area state matrix.

[0016] Optionally, the method for extracting the lighting demand state of each area from the area state matrix based on a convolutional neural network includes:

[0017] Normalize the generated regional state matrix and input the normalized regional state matrix into the convolutional neural network;

[0018] Extracting feature information from the regional state matrix, including pedestrian density and light intensity distribution, through convolutional layers and pooling layers in a convolutional neural network;

[0019] The extracted feature information is input into the fully connected layer to output the lighting demand state probability of each area;

[0020] A threshold judgment is performed on the probability of the lighting demand state of each area to determine the lighting demand state of each area.

[0021] Optionally, the lighting duration of the street lamps is calculated based on the real-time working status data of the solar street lamps and the nighttime lighting demand, the lighting demand status of each area is used as the input of the feedforward neural network, and the average lighting duration is used as the output to train and generate a first prediction model, including:

[0022] Calculate the theoretical lighting time of each street lamp through the real-time working status data of solar street lamps and night lighting demand data;

[0023] Collect statistics on the lighting duration of each street lamp and calculate the average lighting duration of each area;

[0024] The lighting demand status of each area is used as the input feature, and the average lighting duration of the corresponding time period is used as the output label to build a feedforward neural network model;

[0025] The feedforward neural network model is trained, and the model parameters are adjusted through a back-propagation algorithm to minimize the error between the predicted lighting duration and the actual lighting duration, thereby generating a first prediction model.

[0026] Optionally, the formula for calculating the energy saving value based on the average lighting duration predicted and output by the first prediction model and a preset lighting threshold is as follows:

[0027] Indicates the preset lighting threshold.

[0028] Optionally, locating the energy-saving sudden change node by using the change slope value of the energy-saving value includes:

[0029] Splice the energy-saving values ​​in time series to generate an energy-saving value curve;

[0030] Calculate the slope of the energy saving value curve at each time point;

[0031] A slope threshold is set, and when the slope value of the energy-saving value exceeds the slope threshold, it is determined to be an energy-saving sudden change node.

[0032] Optionally, the formula of the BP neural network is as follows:

[0033]

[0034] Among them, y represents the energy-saving sudden change judgment score, and x represents the lighting demand status of each area. and Respectively represent different weight matrices of BP neural network, and is the bias vector of the BP neural network, and f represents the activation function.

[0035] Optionally, the step of adjusting the lighting brightness and on / off status of the solar street lamp in real time according to the lighting demand status of each area and the energy-saving sudden change judgment score output by the second prediction model includes:

[0036] Input the real-time lighting demand status of each area into the second prediction model and output an energy-saving sudden change judgment score. If the energy-saving sudden change judgment score is greater than or equal to a preset energy-saving threshold, reduce the lighting brightness of the street lights in the area or turn off some street lights;

[0037] If the energy-saving sudden change judgment score is less than the preset energy-saving threshold, the normal lighting brightness of the street lamps in the area will be restored.

[0038] Optionally, the system further comprises: a wireless communication module;

[0039] The wireless communication module is connected to the lighting scheduling module and the terminal device respectively;

[0040] The wireless communication module is configured to receive the control instruction output by the terminal device and transmit the control instruction to the lighting scheduling module;

[0041] The lighting scheduling module is further used to adjust the lighting brightness and switch status of the solar street lamp in real time according to the control instruction when receiving the control instruction.

[0042] Optionally, the lighting scheduling module is further used to send the lighting brightness and switch status information of the solar street lamp to the terminal device through the wireless communication module, so that the terminal device displays the lighting brightness and switch status information of the solar street lamp.

[0043] In the present invention, an intelligent cluster control system of solar street lights based on the Internet of Things is provided, and the system includes: a data acquisition module, an area monitoring module, a lighting duration prediction module, an emergency judgment module and a lighting scheduling module; wherein, the data acquisition module is respectively connected to the area monitoring module and the lighting duration prediction module, the area monitoring module is respectively connected to the lighting duration prediction module and the lighting scheduling module, the lighting duration prediction module is connected to the emergency judgment module, and the emergency judgment module is connected to the lighting scheduling module; the data acquisition module is used to obtain real-time working status data of solar street lights through the Internet of Things sensor network, and the real-time working status data includes ambient light intensity, remaining battery power, lamp switch status and fault information; obtain environmental images and pedestrian flow images of the area where the street lights are located, and generate a regional status matrix based on the environmental images and pedestrian flow images; the regional monitoring module A module is used to extract the lighting demand status of each area from the regional state matrix based on a convolutional neural network; a lighting duration prediction module is used to calculate the lighting duration of the street lights based on the real-time operating status data of the solar street lights and the nighttime lighting demand, using the lighting demand status of each area as the input of the feedforward neural network and the average lighting duration as the output to train and generate a first prediction model; a sudden change judgment module is used to calculate the energy saving value based on the average lighting duration predicted by the first prediction model and a preset lighting threshold, locate the energy saving sudden change node based on the slope of the energy saving value, use the lighting demand status of each area corresponding to the energy saving sudden change node as the input of the BP neural network, and use the energy saving sudden change judgment score as the output to train and generate a second prediction model; and a lighting scheduling module is used to adjust the lighting brightness and on / off status of the solar street lights in real time based on the real-time lighting demand status of each area and the energy saving sudden change judgment score output by the second prediction model. By using the Internet of Things, deep learning, and intelligent scheduling technologies, this invention solves the problem of excessive power consumption and energy waste caused by traditional solar street lights when the flow of people changes, thereby improving the system's intelligence level and energy utilization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a structural diagram of the first embodiment of the solar street light intelligent cluster control system based on the Internet of Things of the present invention;

[0045] Figure 2 This is a structural diagram of the second embodiment of the solar street light intelligent cluster control system based on the Internet of Things of the present invention;

[0046] Figure 3 This is a flow chart for constructing the first prediction model in the solar street light intelligent cluster control system based on the Internet of Things of the present invention.

[0047] Description of Figure Numbers:

[0048] Label name Label name 10 Data acquisition module 20 Regional monitoring module 30 Lighting duration prediction module 40 Sudden change judgment module 50 Lighting scheduling module 60 Wireless communication module

[0049] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0050] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0051] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indications will also change accordingly.

[0052] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features specified as "first" and "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between the various embodiments can be combined with each other, but this must be based on the fact that ordinary technicians in this field can implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0053] The embodiment of the present invention provides a solar street light intelligent cluster control system based on the Internet of Things. Figure 1 As shown, Figure 1 This is a block diagram of the first embodiment of the solar street light intelligent cluster control system based on the Internet of Things of the present invention. The solar street light intelligent cluster control system based on the Internet of Things of the present invention includes: a data acquisition module 10, an area monitoring module 20, a lighting duration prediction module 30, a sudden change judgment module 40, and a lighting scheduling module 50;

[0054] The data acquisition module 10 is connected to the area monitoring module 20 and the lighting duration prediction module 30 respectively, the area monitoring module 20 is connected to the lighting duration prediction module 30 and the lighting scheduling module 50 respectively, the lighting duration prediction module 30 is connected to the sudden change judgment module 40, and the sudden change judgment module 40 is connected to the lighting scheduling module 50;

[0055] It should be noted that the sensor network deployment includes ambient light sensors, battery status sensors, lamp status sensors, and cameras. The ambient light sensors are installed on streetlights to monitor ambient light intensity in real time. The battery status sensors are installed in solar panels and battery packs to monitor the remaining battery charge. The lamp status sensors are installed on each streetlight to monitor the lamp's on / off status and fault information. Cameras are installed on streetlights or at specific locations to capture images of the environment and pedestrian traffic in the area where the streetlights are located. Specifically, the ambient light sensor collects light intensity data at regular intervals (e.g., every minute) and transmits this data to the data acquisition module 10 via a wireless communication module (e.g., LoRa or NB-IoT). The battery status sensor monitors battery parameters such as voltage, current, and temperature in real time, calculates the remaining battery charge, and transmits this data to the data acquisition module 10. The lamp status sensor monitors the lamp's on / off status and fault information, such as whether the lamp is on or if there is a fault, and transmits this information to the data acquisition module 10. The camera captures images of the environment and pedestrian traffic in the area where the streetlights are located, either periodically or on demand. Image data is transmitted to the data acquisition module 10 via a wired or wireless network. The data acquisition module 10 preprocesses the received raw data, including data cleaning, format conversion, and preliminary analysis. For example, the module can standardize light intensity data and compress and reduce noise on the image data. Based on the collected environmental images and pedestrian flow images, the data acquisition module 10 can use image processing techniques (such as OpenCV) to extract key features and generate a regional state matrix. The regional state matrix contains information such as light intensity and pedestrian flow density in each area.

[0056] The regional monitoring module 20 is used to obtain the regional state matrix from the data acquisition module 10, which contains information such as ambient light intensity and pedestrian flow density. Environmental images and pedestrian flow images are obtained from the data acquisition module 10. The obtained images are preprocessed, including denoising, enhancement, and normalization, to improve the accuracy of subsequent processing. Key features, such as the number of pedestrians and light intensity distribution, are extracted from the images using image processing techniques and deep learning methods (such as convolutional neural networks (CNNs). A convolutional neural network model is constructed to analyze the regional state matrix and image data. The model typically includes multiple convolutional layers, pooling layers, and fully connected layers. A large amount of historical data is collected, including regional state matrices and corresponding lighting demand states at different time points. A corresponding lighting demand state label, such as high demand, medium demand, or low demand, is generated for each sample. The convolutional neural network is trained using a training dataset, and model parameters are optimized to enable it to accurately predict the lighting demand state of each area. The model's performance is evaluated using validation and test sets to ensure its accuracy and robustness in practical applications. The real-time collected regional state matrix and image data are input into a trained convolutional neural network model, which then outputs the lighting demand status for each area. The predicted lighting demand status is then passed to the lighting duration prediction module 30 and the lighting scheduling module 50, providing a basis for subsequent intelligent control. The lighting demand status refers to the degree of demand for street lighting within a specific time and area. It can be comprehensively assessed based on a variety of factors, including but not limited to ambient light intensity, pedestrian density, weather conditions, and seasonal changes.

[0057] The lighting duration prediction module 30 obtains the real-time working status data of the solar street lamp from the data acquisition module 10, including the ambient light intensity, remaining battery power, lamp switch status and fault information, and the lighting demand status of each area from the regional monitoring module 20. The data is preprocessed, including removing outliers and noise data to ensure the accuracy and reliability of the data, extracting and constructing features that are helpful for prediction, such as timestamps, seasons, weather conditions, etc., and standardizing the data so that they are on the same scale to facilitate model training. Among them, the feedforward neural network model includes an input layer, a hidden layer and an output layer. The input layer is used to receive real-time working status data and regional lighting demand status. The hidden layer is a multi-layer fully connected layer, and an activation function (such as ReLU) is used to enhance the nonlinear fitting ability of the model. The output layer is a single node that outputs the predicted average lighting duration. The model training steps include: collecting historical data, including real-time working status data of different time periods, regional lighting demand status and corresponding night lighting duration; generating corresponding night lighting duration labels for each sample; selecting a suitable loss function, such as mean square error (MSE), to measure the difference between the predicted value and the actual value; selecting an optimization algorithm, such as Adam or SGD, to update the model parameters; dividing the data set into a training set, a validation set and a test set. Use the training set to train the model, adjust the hyperparameters through the validation set, and finally evaluate the model performance on the test set; use evaluation indicators (such as MSE, MAE, R 2 etc.) to evaluate the performance of the model on the test set; adjust the model structure, hyperparameters or add more features based on the evaluation results to improve the prediction accuracy; input the real-time collected data into the trained feedforward neural network model; the model outputs the predicted average lighting duration and passes the result to the sudden change judgment module 40 and the lighting scheduling module 50.

[0058] It should be understood that the sudden change determination module 40 is used to evaluate the current energy-saving effect by calculating the energy-saving value and determine whether the preset energy-saving target has been achieved. By analyzing the slope of the energy-saving value, the system can understand the changing trend of the energy-saving effect and promptly detect anomalies. Specifically, the average lighting duration refers to the predicted average lighting duration obtained from the lighting duration prediction module 30; the preset lighting threshold refers to the standard lighting duration set by the system and used to calculate the energy-saving value. The lighting demand status refers to the lighting demand status of each area obtained from the area monitoring module 20. The energy-saving value refers to the difference between the actual predicted average lighting duration and the preset lighting threshold. If the predicted average lighting duration is lower than the preset threshold, the energy-saving value is positive, indicating energy savings; otherwise, it is negative, indicating energy consumption. The slope of the energy-saving value represents the rate of change of the energy-saving value over time. Energy-saving sudden change nodes are points at which the slope of the energy-saving value changes significantly. By analyzing the slope of the energy-saving value, these nodes can be located. For example, using a sliding window or other time series analysis method, sudden change points can be detected to identify sudden change nodes.

[0059] It can be understood that the steps of constructing the BP neural network model can be as follows: the input layer is the lighting demand status of each area corresponding to the energy-saving sudden change node time. The hidden layer is a multi-layer neuron, and the activation function (such as Sigmoid or ReLU) is used to enhance the nonlinear fitting ability of the model. The output layer is a single node, which outputs the energy-saving sudden change judgment score. Specifically, historical data is collected, including the energy-saving sudden change node time, lighting demand status and corresponding energy-saving sudden change judgment score in different time periods; a corresponding energy-saving sudden change judgment score is generated for each sample; a suitable loss function, such as mean square error (MSE), is selected to measure the difference between the predicted value and the actual value; an optimization algorithm, such as the back propagation algorithm, is selected to update the model parameters; the data set is divided into a training set, a validation set and a test set; the model is trained using the training set, the hyperparameters are adjusted through the validation set, and the model performance is finally evaluated on the test set; evaluation indicators (such as MSE, MAE, R 2 etc.) to evaluate the performance of the model on the test set; adjust the model structure, hyperparameters or add more features based on the evaluation results to improve the prediction accuracy; input the real-time collected data into the trained BP neural network model; the model outputs the energy-saving sudden change judgment score and passes the result to the lighting scheduling module 50 for corresponding emergency processing.

[0060] It should be noted that when the lighting demand status in a certain area is "high demand" (such as in commercial areas or during peak nighttime hours), the lighting scheduling module 50 will increase the streetlight brightness if the streetlight power consumption rate threshold is met to ensure pedestrian safety and comfort. When the lighting demand status is "low demand" (such as in residential areas or late at night), the module will reduce the streetlight brightness to reduce energy consumption. If the energy-saving sudden change judgment score indicates that the current energy-saving effect is poor (such as a sudden drop in battery power or a sudden change in light intensity), the module will appropriately reduce the brightness of the solar-powered streetlights in the area or turn off some solar-powered streetlights to save energy while meeting the area's minimum lighting demand. If the energy-saving sudden change judgment score indicates that the energy-saving effect is good, the module will maintain or optimize the current brightness to ensure lighting quality. Specifically, when the lighting demand status is "no demand" (such as during the day or when there is sufficient sunlight), the module will turn off the streetlights to avoid unnecessary energy waste. When the lighting demand status is "demand", the module will turn on the streetlights according to specific needs. If the energy-saving sudden change judgment score indicates low battery power or energy shortage, the module will prioritize turning off streetlights in non-critical areas to ensure lighting in important areas. If the energy-saving sudden change judgment score shows that there is sufficient energy, the module will flexibly adjust the switch status according to demand.

[0061] In this embodiment, an intelligent cluster control system of solar street lamps based on the Internet of Things is provided, and the system includes: a data acquisition module 10, an area monitoring module 20, a lighting duration prediction module 30, an emergency judgment module 40 and a lighting scheduling module 50; wherein, the data acquisition module 10 is connected to the area monitoring module 20 and the lighting duration prediction module 30 respectively, the area monitoring module 20 is connected to the lighting duration prediction module 30 and the lighting scheduling module 50 respectively, the lighting duration prediction module 30 is connected to the emergency judgment module 40, and the emergency judgment module 40 is connected to the lighting scheduling module 50; the data acquisition module 10 is used to obtain real-time working status data of solar street lamps through the Internet of Things sensor network, and the real-time working status data includes ambient light intensity, remaining battery power, lamp switch status and fault information; obtain environmental images and pedestrian flow images of the area where the street lamps are located, and generate pedestrian flow images based on the environmental images and pedestrian flow images. The system comprises a regional state matrix; a regional monitoring module 20 for extracting the lighting demand status of each region from the regional state matrix based on a convolutional neural network; a lighting duration prediction module 30 for calculating the lighting duration of the solar street lights based on the real-time operating status data of the solar street lights and the nighttime lighting demand, using the lighting demand status of each region as the input of a feedforward neural network and the average lighting duration as the output to train and generate a first prediction model; a sudden change judgment module 40 for calculating the energy saving value based on the average lighting duration predicted and output by the first prediction model and a preset lighting threshold, locating the energy saving sudden change node based on the slope of the energy saving value, using the lighting demand status of each region corresponding to the energy saving sudden change node as the input of a BP neural network and the energy saving sudden change judgment score as the output to train and generate a second prediction model; and a lighting scheduling module 50 for adjusting the lighting brightness and on / off status of the solar street lights in real time based on the real-time lighting demand status of each region and the energy saving sudden change judgment score output by the second prediction model. This embodiment solves the problem of excessive power consumption and energy waste caused by traditional solar street lights when the flow of people changes by using the Internet of Things, deep learning, and intelligent scheduling technologies, thereby improving the system's intelligence level and energy utilization efficiency.

[0062] Furthermore, the acquiring of an environmental image and a pedestrian flow image of an area where a street lamp is located, and generating a regional state matrix based on the environmental image and the pedestrian flow image, includes:

[0063] Set up several real-time image acquisition devices in the area where the street lights are located to acquire environmental images and pedestrian flow images;

[0064] Splicing the environment images to construct a two-dimensional regional model, and generating a regional matrix according to the two-dimensional regional model;

[0065] The distribution information of pedestrians in the two-dimensional model of the area is obtained according to the pedestrian flow image, and the area matrix is ​​mapped according to the position information in the distribution information. The pedestrian density is preset to the same value and the area matrix parameters are replaced to generate the area state matrix.

[0066] It should be noted that several real-time image acquisition devices (such as cameras and infrared sensors) are installed at key locations in the streetlight area (such as intersections, squares, and sidewalks) to ensure full coverage and avoid blind spots. Environmental images refer to static images of the area where the streetlights are located, including buildings, roads, and vegetation. Pedestrian flow images refer to dynamic images of pedestrian flow, capturing the location, density, and movement of pedestrians. Specifically, the acquired environmental images are de-noised, corrected, and enhanced to ensure image quality. Image stitching algorithms (such as feature point matching and image fusion) are used to stitch multiple environmental images into a complete two-dimensional model of the area. The stitched two-dimensional model of the area is converted into a digital matrix, where each element represents a pixel or grid cell in the area. The matrix values ​​are quantitative representations of environmental characteristics (such as light intensity and obstacle locations). Pedestrian detection algorithms (such as YOLO and OpenCV) are used to identify pedestrian locations from the pedestrian flow images. Pedestrian distribution information within the two-dimensional model of the area is extracted, including the number of pedestrians, their location coordinates, and density. Map the location information of pedestrians into the regional matrix to determine the matrix unit where the pedestrians are located. Preset a pedestrian density value (such as a value between 0 and 1) for each matrix unit to represent the pedestrian density of the unit. Replace the parameters of the corresponding unit in the regional matrix with the preset pedestrian density value, and finally generate a regional status matrix. Each element in the matrix contains environmental characteristics and pedestrian density information. The regional status matrix integrates the information of the environmental image and the pedestrian flow image to provide comprehensive data support for subsequent lighting scheduling and energy-saving control. Through the regional status matrix, the system can accurately understand the environmental characteristics and pedestrian density of each area, and realize zoning management and precise control.

[0067] Furthermore, the method for extracting the lighting demand state of each area from the area state matrix based on a convolutional neural network includes:

[0068] Normalize the generated regional state matrix and input the normalized regional state matrix into the convolutional neural network;

[0069] Extracting feature information from the regional state matrix, including pedestrian density and light intensity distribution, through convolutional layers and pooling layers in a convolutional neural network;

[0070] The extracted feature information is input into the fully connected layer to output the lighting demand state probability of each area;

[0071] A threshold judgment is performed on the probability of the lighting demand state of each area to determine the lighting demand state of each area.

[0072] It should be noted that the values ​​in the generated regional state matrix are normalized to a certain range (e.g., between 0 and 1). This step helps improve the model's convergence speed and stability. The normalized regional state matrix is ​​used as input data to the convolutional neural network. At this point, the regional state matrix can be viewed as a two-dimensional image, with each element representing a pixel or grid cell. The convolutional layer in the convolutional neural network convolves the input matrix with a convolution kernel (filter) to extract local features. This convolution operation can capture features such as pedestrian density and light intensity distribution. Each convolution kernel generates a feature map, and multiple convolution kernels can extract different types of features. After the convolution operation, an activation function (such as ReLU) is typically applied to introduce nonlinearity and enhance the model's expressiveness. The pooling layer uses downsampling operations (such as max pooling or average pooling) to reduce the size of the feature map, reducing computational complexity while retaining key features. The pooling layer also enhances the model's translation invariance and robustness. The feature information extracted by the convolutional and pooling layers is flattened into a one-dimensional vector and then input into the fully connected layer. The fully connected layer performs a linear transformation through a series of weight matrices and bias terms, and outputs the lighting demand state probability of each area through an activation function (such as Softmax). The Softmax function can convert the output into a probability distribution, so that the sum of the probabilities of the lighting demand state of each area is 1. The output layer outputs the lighting demand state probability of each area, such as high demand state, medium demand state, and low demand state; different thresholds are set to determine the lighting demand state of each area. For example, if the probability of a high demand state in a certain area is greater than 0.7, the area is judged to be in a high demand state. If the probability of a medium demand state in a certain area is greater than 0.5 and less than 0.7, the area is judged to be in a medium demand state. If the probability of a low demand state in a certain area is greater than 0.5, the area is judged to be in a low demand state. Based on the results of the threshold judgment, the lighting demand state of each area is ultimately determined.

[0073] Further, refer to Figure 3 The method calculates the lighting duration of the street lamps by using the real-time working status data of the solar street lamps and the nighttime lighting demand, takes the lighting demand status of each area as the input of the feedforward neural network, and takes the average lighting duration as the output to train and generate a first prediction model, including:

[0074] Calculate the theoretical lighting time of each street lamp through the real-time working status data of solar street lamps and night lighting demand data;

[0075] Collect statistics on the lighting duration of each street lamp and calculate the average lighting duration of each area;

[0076] The lighting demand status of each area is used as the input feature, and the average lighting duration of the corresponding time period is used as the output label to build a feedforward neural network model;

[0077] The feedforward neural network model is trained, and the model parameters are adjusted through a back-propagation algorithm to minimize the error between the predicted lighting duration and the actual lighting duration, thereby generating a first prediction model.

[0078] It should be noted that the lighting demand status of each area (e.g., high demand, medium demand, low demand) is used as an input feature. The input features can be further expanded to include environmental characteristics (e.g., light intensity, temperature, etc.) and pedestrian flow. The output label is the average lighting duration for the corresponding time period. A feedforward neural network is constructed, which can include multiple fully connected layers. The input layer inputs the feature dimension, the hidden layers are used for multiple fully connected layers, each layer uses an activation function, and the output layer is a single neuron with a linear activation function. The feedforward neural network model is trained, and the model parameters are adjusted using a backpropagation algorithm to minimize the error between the predicted lighting duration and the actual lighting duration. To generate the first prediction model, the dataset is divided into a training set, a validation set, and a test set. An appropriate loss function, such as mean squared error, is selected, along with an appropriate optimizer, such as stochastic gradient descent (SGD) or Adam. The model parameters are adjusted using a backpropagation algorithm to minimize the loss function. During training, performance on the validation set is monitored to prevent overfitting. The model performance is evaluated on the test set to ensure good performance on unseen data. Save the trained model for later use.

[0079] The theoretical lighting duration of each streetlight is calculated using real-time operating status data from solar streetlights and nighttime lighting demand data. The average lighting duration for each area is then statistically calculated. A feedforward neural network model is then constructed and trained, using the lighting demand status of each area as input features and the average lighting duration for the corresponding time period as output labels. The model parameters are adjusted using a backpropagation algorithm to minimize the error between the predicted and actual lighting durations, ultimately generating the first prediction model. This process enables accurate lighting duration predictions, providing reliable data support for intelligent lighting scheduling.

[0080] Furthermore, the formula for calculating the energy saving value based on the average lighting duration predicted and output by the first prediction model and the preset lighting threshold is as follows:

[0081]

[0082] Among them, E represents the energy saving value, T a represents the average lighting duration predicted by the first prediction model, Tp Indicates the preset lighting threshold.

[0083] It should be noted that when T p =T a When T p <T a When the energy saving value E>0, it means that the actual lighting time is lower than the preset threshold, and the energy saving effect is achieved. p >T a When the energy-saving value E is less than 0, it means that the actual lighting duration exceeds the preset threshold and the energy-saving effect is not achieved. The energy-saving value E can be used to evaluate the energy-saving effect of solar street lights and provide a basis for optimizing lighting strategies. If the energy-saving value E is low (such as close to 0 or a negative value), the lighting strategy can be adjusted, such as reducing the brightness or shortening the lighting duration. The energy-saving value E is monitored in real time and the results are fed back to the intelligent control system for dynamic adjustment and optimization.

[0084] Furthermore, locating the energy-saving sudden change node by using the change slope value of the energy-saving value includes:

[0085] Splice the energy-saving values ​​in time series to generate an energy-saving value curve;

[0086] Calculate the slope of the energy saving value curve at each time point;

[0087] A slope threshold is set, and when the slope value of the energy-saving value exceeds the slope threshold, it is determined to be an energy-saving sudden change node.

[0088] It should be noted that energy-saving value data is collected over a period of time, and each time point corresponds to an energy-saving value. These energy-saving values ​​are arranged in chronological order to generate a time series. The time series data is plotted into a curve graph, with the horizontal axis being time and the vertical axis being the energy-saving value. The slope value of the energy-saving value curve at each time point is calculated, where the slope represents the instantaneous rate of change of the curve at a certain point. For discrete time series data, the differential method can be used to approximate the slope. A slope threshold is set. When the slope value of the energy-saving value exceeds the slope threshold, it is determined to be an energy-saving sudden change node. The slope threshold is set according to actual needs and historical data analysis, and a reasonable slope threshold is set. Traverse the slope value of each time point. If the slope value at a certain time point exceeds the set threshold, the time is determined to be an energy-saving sudden change node.

[0089] Furthermore, the formula of the BP neural network is as follows:

[0090]

[0091] Among them, y represents the energy-saving sudden change judgment score, and x represents the lighting demand status of each area. and Respectively represent different weight matrices of BP neural network, and is the bias vector of the BP neural network, and f represents the activation function.

[0092] It should be noted that y is the final output of the BP neural network, representing the energy-saving abrupt change judgment score calculated based on the input data x. This score can be used to determine whether a specific area has an energy-saving abrupt change node. x is the input vector of the BP neural network, containing information about the lighting demand status of each area. For example, this can include high demand, medium demand, and low demand. W1 and W2 are the weight matrices from the input layer to the hidden layer. They determine how the input feature x is mapped to the neurons in the hidden layer. The weight matrix controls the connection strength between the layers in the neural network and is a key parameter in model learning. and is the bias vector of the hidden layer, used to adjust the activation values ​​of the hidden layer neurons. The bias vector allows the model to produce non-zero outputs even when there is no input, increasing the model's flexibility and expressiveness. f represents the activation function. The activation function introduces nonlinearity, enabling the neural network to learn and represent complex nonlinear relationships. The activation function enables the neural network to handle nonlinear problems and facilitates gradient propagation during training.

[0093] Furthermore, the method of adjusting the lighting brightness and on / off status of the solar street lamp in real time according to the lighting demand status of each area and the energy-saving sudden change judgment score output by the second prediction model includes:

[0094] Input the real-time lighting demand status of each area into the second prediction model and output an energy-saving sudden change judgment score. If the energy-saving sudden change judgment score is greater than or equal to a preset energy-saving threshold, reduce the lighting brightness of the street lights in the area or turn off some street lights;

[0095] If the energy-saving sudden change judgment score is less than the preset energy-saving threshold, the normal lighting brightness of the street lamps in the area will be restored.

[0096] It should be noted that a reasonable energy-saving threshold value T is set to judge the sudden change of energy saving. For example, T = 0.5 can be set, and the specific value is adjusted according to the actual situation. The energy-saving sudden change judgment score y output by the second prediction model and the preset energy-saving threshold value T. If y ≥ T, it is considered that there is an energy-saving sudden change node in this area. If y < T, it is considered that there is no energy-saving sudden change node in this area. If y ≥ T, the lighting brightness of the street lamps in this area is reduced under the condition of meeting the minimum lighting demand of this area. If y ≥ T and further energy saving is required, some street lamps can be selected to be turned off. For example, every other street lamp is turned off under the condition of meeting the minimum lighting demand of this area, or the lighting street lamps in the area with lower lighting demand are turned off. If y < T, the normal lighting brightness of the street lamps in this area, that is, the preset lighting brightness, is restored.

[0097] Refer to Figure 2 , Figure 2 FIG. is a schematic structural diagram of the second embodiment of the intelligent cluster control system for solar street lamps based on the Internet of Things according to the present invention; based on the above first embodiment, the second embodiment of the intelligent cluster control system for solar street lamps based on the Internet of Things according to the present invention is proposed.

[0098] Furthermore, in this embodiment, the system further includes: a wireless communication module 60;

[0099] The wireless communication module 60 is respectively connected to the lighting scheduling module 50 and the terminal device;

[0100] The wireless communication module 60 is used to receive the control instruction output by the terminal device and transmit the control instruction to the lighting scheduling module 50;

[0101] The lighting scheduling module 50 is further used to, when receiving the control instruction, adjust the lighting brightness and switch state of the solar street lamps in real time according to the control instruction.

[0102] It should be noted that, to achieve more flexible and intelligent solar street light scheduling, the system also includes a wireless communication module 60. This module serves as a key communication bridge in the system, connecting terminal devices and the lighting scheduling module 50. The wireless communication module 60 is responsible for receiving control commands from terminal devices, which can be smartphones, tablets, monitoring centers, etc., and transmitting the received control commands to the lighting scheduling module 50 for corresponding operations. The wireless communication module 60 can use common wireless communication protocols such as Wi-Fi, Bluetooth, LoRaWAN, and NB-IoT. The specific choice depends on the system's coverage and communication requirements. In addition, it ensures reliable and real-time data transmission and supports two-way communication, that is, the terminal device can send commands and receive feedback information. The lighting scheduling module 50 receives control commands from the terminal device from the wireless communication module 60 and adjusts the lighting brightness and on / off status of the solar street light in real time based on the received control commands. Users can send control commands through the terminal device, such as adjusting the lighting brightness of a specific area or turning specific street lights on or off, thus enabling manual control and increasing the flexibility of the system.

[0103] Furthermore, the lighting scheduling module 50 is also used to send the lighting brightness and switch status information of the solar street lamp to the terminal device through the wireless communication module 60, so that the terminal device displays the lighting brightness and switch status information of the solar street lamp.

[0104] It should be noted that in order to achieve a more intelligent and user-friendly solar street light management system, the lighting scheduling module 50 not only adjusts the lighting brightness and switch status of the street light according to the energy-saving sudden change judgment score and control instructions, but also sends this information to the terminal device through the wireless communication module 60 so that the user can view and monitor the status of the street light in real time.

[0105] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A solar street light intelligent cluster control system based on the Internet of Things, characterized in that: The system includes: a data acquisition module, an area monitoring module, a lighting duration prediction module, an emergency change judgment module and a lighting scheduling module; The data acquisition module is connected to the area monitoring module and the lighting duration prediction module respectively, the area monitoring module is connected to the lighting duration prediction module and the lighting scheduling module respectively, the lighting duration prediction module is connected to the sudden change judgment module, and the sudden change judgment module is connected to the lighting scheduling module; A data acquisition module is used to obtain real-time operating status data of solar street lights through an Internet of Things sensor network. The real-time operating status data includes ambient light intensity, remaining battery power, lamp switch status, and fault information; obtain environmental images and pedestrian flow images of the area where the street lights are located, and generate a regional status matrix based on the environmental images and pedestrian flow images; The regional monitoring module is used to extract the lighting demand status of each area from the regional state matrix based on a convolutional neural network; The lighting duration prediction module is used to calculate the lighting duration of the street lights based on the real-time working status data of the solar street lights and the nighttime lighting demand. The lighting demand status of each area is used as the input of the feedforward neural network, and the average lighting duration is used as the output to train and generate a first prediction model; a sudden change judgment module, configured to calculate an energy saving value based on the average lighting duration predicted and output by the first prediction model and a preset lighting threshold, locate an energy saving sudden change node by the slope of the energy saving value, use the lighting demand state of each area corresponding to the energy saving sudden change node as input to a BP neural network, and use the energy saving sudden change judgment score as output to train and generate a second prediction model; The lighting scheduling module is used to adjust the lighting brightness and switch status of the solar street lamp in real time according to the real-time lighting demand status of each area and the energy-saving sudden change judgment score output by the second prediction model.

2. The solar street light intelligent cluster control system based on the Internet of Things according to claim 1, characterized in that: The step of acquiring an environment image and a pedestrian flow image of an area where a street lamp is located, and generating a regional state matrix based on the environment image and the pedestrian flow image, includes: Set up several real-time image acquisition devices in the area where the street lights are located to acquire environmental images and pedestrian flow images; Splicing the environment images to construct a two-dimensional regional model, and generating a regional matrix according to the two-dimensional regional model; The distribution information of pedestrians in the two-dimensional model of the area is obtained according to the pedestrian flow image, and the area matrix is ​​mapped according to the position information in the distribution information. The pedestrian density is preset to the same value and the area matrix parameters are replaced to generate the area state matrix.

3. The solar street light intelligent cluster control system based on the Internet of Things as claimed in claim 1, characterized in that: The method for extracting the lighting demand state of each area from the regional state matrix based on a convolutional neural network includes: Normalize the generated regional state matrix and input the normalized regional state matrix into the convolutional neural network; Extracting feature information from the regional state matrix, including pedestrian density and light intensity distribution, through convolutional layers and pooling layers in a convolutional neural network; The extracted feature information is input into the fully connected layer to output the lighting demand state probability of each area; A threshold judgment is performed on the probability of the lighting demand state of each area to determine the lighting demand state of each area.

4. The solar street light intelligent cluster control system based on the Internet of Things according to claim 1, characterized in that: The method calculates the lighting duration of the street lamps by using the real-time working status data of the solar street lamps and the nighttime lighting demand, takes the lighting demand status of each area as the input of the feedforward neural network, and takes the average lighting duration as the output to train and generate a first prediction model, including: Calculate the theoretical lighting time of each street lamp through the real-time working status data of solar street lamps and night lighting demand data; Collect statistics on the lighting duration of each street lamp and calculate the average lighting duration of each area; The lighting demand status of each area is used as the input feature, and the average lighting duration of the corresponding time period is used as the output label to build a feedforward neural network model; The feedforward neural network model is trained, and the model parameters are adjusted through a back-propagation algorithm to minimize the error between the predicted lighting duration and the actual lighting duration, thereby generating a first prediction model.

5. The solar street light intelligent cluster control system based on the Internet of Things as claimed in claim 1, characterized in that: The formula for calculating the energy saving value based on the average lighting duration predicted by the first prediction model and the preset lighting threshold is as follows: Among them, E represents the energy saving value, T a represents the average lighting duration predicted by the first prediction model, T p Indicates the preset lighting threshold.

6. The solar street light intelligent cluster control system based on the Internet of Things according to claim 1, characterized in that: The method of locating the energy-saving sudden change node according to the change slope of the energy-saving value includes: Splice the energy-saving values ​​in time series to generate an energy-saving value curve; Calculate the slope of the energy saving value curve at each time point; A slope threshold is set, and when the slope value of the energy-saving value exceeds the slope threshold, it is determined to be an energy-saving sudden change node.

7. The solar street light intelligent cluster control system based on the Internet of Things according to claim 1, characterized in that: The formula of the BP neural network is as follows: Among them, y represents the energy-saving sudden change judgment score, and x represents the lighting demand status of each area. and Respectively represent different weight matrices of BP neural network, and is the bias vector of the BP neural network, and f represents the activation function.

8. The solar street light intelligent cluster control system based on the Internet of Things as claimed in claim 1, characterized in that: The method of adjusting the lighting brightness and on / off status of the solar street lamp in real time according to the lighting demand status of each area and the energy-saving sudden change judgment score output by the second prediction model includes: Input the real-time lighting demand status of each area into the second prediction model and output an energy-saving sudden change judgment score. If the energy-saving sudden change judgment score is greater than or equal to a preset energy-saving threshold, reduce the lighting brightness of the street lights in the area or turn off some street lights; If the energy-saving sudden change judgment score is less than the preset energy-saving threshold, the normal lighting brightness of the street lamps in the area will be restored.

9. The solar street light intelligent cluster control system based on the Internet of Things according to claim 1, characterized in that: The system further includes: a wireless communication module; The wireless communication module is connected to the lighting scheduling module and the terminal device respectively; The wireless communication module is configured to receive the control instruction output by the terminal device and transmit the control instruction to the lighting scheduling module; The lighting scheduling module is further used to adjust the lighting brightness and switch status of the solar street lamp in real time according to the control instruction when receiving the control instruction.

10. The solar street light intelligent cluster control system based on Internet of Things according to claim 9, characterized in that: The lighting scheduling module is also used to send the lighting brightness and switch status information of the solar street lamp to the terminal device through the wireless communication module, so that the terminal device displays the lighting brightness and switch status information of the solar street lamp.

Images