LED street lamp brightness adaptive adjustment method and system

Abstract

The invention provides an LED street lamp brightness adaptive adjustment method and system, and the method comprises the steps: collecting light intensity time sequence data in a target region, the position-speed associated data of a road surface moving target, and power supply frequency fluctuation data through a Mesh network, so as to create a corresponding fusion data set; transmitting the fusion data set to an edge computing node of the LED street lamp to perform dynamic scene clustering and divide a plurality of target use scenes, and synchronously outputting an initial brightness adjustment parameter matched with each target use scene; historical operation data of the LED street lamp is called, a corresponding brightness attenuation prediction model is constructed, and the brightness attenuation amount in the future preset time is predicted according to the current operation duration and the environment temperature data; and performing pre-compensation correction processing on the initial brightness adjustment parameter according to the brightness attenuation so as to output a corresponding target brightness control parameter, and correspondingly controlling the output luminous flux of the LED street lamp according to the target brightness control parameter. The energy-saving benefit of the street lamp can be effectively improved.

Description

Technical Field

[0001] This invention relates to the field of street light brightness adjustment technology, and in particular to an adaptive brightness adjustment method and system for LED street lights. Background Technology

[0002] The adaptive brightness adjustment method for LED streetlights is a core technology for achieving energy conservation and intelligent control in road lighting. By matching environmental changes with lighting demands, it can reduce public lighting energy consumption and improve traffic safety, demonstrating significant application value. The core of this method is a closed loop of "environmental perception - data processing - brightness control," where multi-source data fusion is the key link connecting perception and control, directly determining the accuracy and reliability of the adjustment method and serving as its core performance indicator.

[0003] Most existing adaptive brightness adjustment methods for LED streetlights employ static data fusion strategies, which involve pre-setting fixed fusion weights and decision rules to integrate the data. While these strategies are simple and low-cost, and can achieve basic adjustment functions under stable and ideal conditions, they fail to adapt to dynamic environments such as day-night cycles and sudden changes in weather. The fixed weights cannot match the real-time reliability differences of the various sensor data, leading to fusion results that deviate from actual lighting requirements and rendering the adjustment method ineffective.

[0004] Furthermore, the shortcomings of static fusion strategies can lead to a series of problems: abnormal brightness output, energy waste, and even traffic safety hazards due to sudden brightness changes. Existing improvement solutions mostly involve manual parameter calibration or adding fixed scene rules, which are not only costly to maintain but also unable to cover complex dynamic environments, failing to fundamentally solve the problem. Therefore, developing a data fusion mechanism adapted to dynamic environments to optimize the performance of adjustment methods has become an urgent need in this field. Summary of the Invention

[0005] Based on this, the purpose of the present invention is to provide an adaptive brightness adjustment method and system for LED streetlights, so as to solve the problem that the static fusion strategy used in the prior art cannot cover dynamic usage scenarios, which easily leads to abnormal brightness output.

[0006] The first aspect of the present invention proposes: An adaptive brightness adjustment method for LED streetlights specifically includes the following steps: A corresponding Mesh network is constructed based on the sensing units inside several LED streetlights within the target area. The Mesh network is used to collect time-series data of light intensity, position-velocity correlation data of moving targets on the road surface, and power supply frequency fluctuation data within the target area to create a corresponding fusion dataset. The fused dataset is transmitted to the edge computing node of the LED street light to perform dynamic scene clustering and divide it into several target usage scenarios, and to synchronously output the initial brightness adjustment parameters adapted to each target usage scenario. The historical operating data of the LED street light is retrieved to construct a corresponding brightness decay prediction model, and the brightness decay amount within a preset time is predicted based on the current running time and ambient temperature data. The initial brightness adjustment parameters are pre-compensated and corrected based on the brightness attenuation amount to output the corresponding target brightness control parameters, and the output luminous flux of the LED street light is controlled accordingly based on the target brightness control parameters.

[0007] The beneficial effects of this invention are as follows: This invention accurately solves the problem of abnormal brightness output caused by the inability of existing static fusion strategies to adapt to dynamic scenarios. It relies on a mesh network constructed from LED street light sensing units to collaboratively collect data on the time sequence of light intensity in the target area, the position-speed correlation of moving targets on the road surface, and power supply frequency fluctuations. Compared to single sensors, it can more comprehensively capture environmental dynamics, providing accurate data support for brightness adjustment. The fused dataset is dynamically clustered through edge computing nodes, enabling real-time matching of initial brightness parameters to adapt to the scene, overcoming the limitations of static strategies. The attenuation prediction model built using historical street light data can predict and compensate for attenuation in advance, avoiding adjustment lag deviations. Ultimately, precise control is achieved, ensuring stable lighting brightness in different scenarios while improving energy-saving benefits.

[0008] Furthermore, the step of transmitting the fused dataset to the edge computing node of the LED street light for dynamic scene clustering and segmentation into several target use scenarios includes: Based on the spatial location coordinates of each LED street light in the Mesh network, a corresponding spatiotemporal correlation matrix is ​​constructed. According to the spatiotemporal correlation matrix, the fused dataset is divided into time slices and spatial grids to extract the spatial propagation features of different time slices within the same spatial grid. The spatial propagation features are processed in a low dimension using the LLE algorithm to generate a corresponding low-dimensional spatiotemporal feature set, and the low-dimensional spatiotemporal feature set is divided into several basic scene clusters according to a preset index. The basic scene clusters are verified by using an adversarial generative network to generate corresponding target scene clusters. The target scene clusters are then iteratively optimized to generate the target usage scenario.

[0009] Furthermore, the step of iteratively optimizing the target scene cluster to generate the corresponding target usage scenario includes: The core feature vectors of each target scene cluster are extracted, and the corresponding feature selection criteria are constructed in combination with the physical constraints of LED street lighting. Redundant and abnormal features with weights lower than a preset weight threshold are removed from the target scene clusters according to the feature selection criteria to generate the corresponding purified feature clusters. The cosine similarity and Euclidean distance between adjacent purified feature clusters are calculated, and the corresponding scene boundary association matrix is ​​constructed. The corresponding boundary features are simulated based on the scene boundary association matrix. The MAML algorithm from meta-learning is introduced to train a scene classifier to generate a corresponding scene classification model. The boundary features are then converted into the target usage scene through the scene classification model.

[0010] Furthermore, the step of predicting the brightness decay amount within a preset time period based on the current running time and ambient temperature data includes: Using the runtime as the time axis, the ambient temperature data is decomposed into instantaneous temperature value, temperature change rate, and temperature sub-characteristics of continuous constant temperature period, and synchronously correlated with the actual operating current of LED street lights to construct the corresponding load correlation factor. A temporal convolutional network is used to extract and enhance the dimensions of the runtime, temperature sub-features and the load correlation factors to generate a corresponding enhanced feature set. Based on the enhanced feature set, the current decay phase of the LED street light is detected by the K-means algorithm. The attenuation phase of the LED street light at different times is collected, and the brightness attenuation is predicted based on the dynamic switching of the attenuation phase.

[0011] Furthermore, the step of predicting the brightness attenuation amount based on the dynamic switching of the attenuation phase includes: Collect full operational data corresponding to each switching moment of the attenuation phase, extract phase steady-state features and phase transient features from the full operational data, and generate the corresponding coupling matrix through principal component analysis; A meta-learning prediction framework is constructed, using the coupling matrix as the meta-training set, and a brightness prediction model adapted to each of the attenuation phases is trained. When a phase attenuation switch is detected, the model parameters of the brightness prediction model are adaptively adjusted through meta-learning, and the corresponding brightness attenuation amount is output.

[0012] Furthermore, the step of performing pre-compensation correction processing on the initial brightness adjustment parameters based on the brightness attenuation amount to output the corresponding target brightness control parameters includes: The trend features of the light intensity time series data are extracted through the edge computing node, and the corresponding light intensity change rate is detected based on the trend features; The brightness attenuation amount and the light intensity change rate are nonlinearly correlated to generate the scene light intensity attenuation value, and the power supply frequency fluctuation data is converted into a voltage stability coefficient. The scene light intensity attenuation value is weighted and calculated with the voltage stability coefficient, and the actual brightness feedback data of the LED street light is collected at the same time to output the target brightness control parameters accordingly.

[0013] Furthermore, the step of weighting the scene light intensity attenuation value with the voltage stability coefficient and simultaneously collecting the actual brightness feedback data of the LED streetlights to output the target brightness control parameters includes: The steady-state value and instantaneous fluctuation value of the actual brightness feedback data are extracted by time-domain filtering, and the strong correlation features between the steady-state value and the instantaneous fluctuation value are enhanced by attention weight allocation strategy in order to remove interference components and generate corresponding brightness reference data. The difference between the brightness reference data and the theoretical brightness value of the LED street light is calculated to obtain the brightness deviation. The deviation is then combined with the scene light intensity attenuation value and the voltage stability coefficient to generate the corresponding initial brightness control parameters. The initial brightness control parameters are embedded into the corresponding hardware delay factor to generate the target brightness control parameters.

[0014] The second aspect of the present invention proposes: An adaptive brightness adjustment system for LED streetlights, wherein the system comprises: The acquisition module is used to construct a corresponding Mesh network based on the sensing units inside several LED streetlights in the target area. The Mesh network is used to acquire light intensity time-series data, position-velocity correlation data of moving targets on the road surface, and power supply frequency fluctuation data in the target area to create a corresponding fusion dataset. The transmission module is used to transmit the fused dataset to the edge computing node of the LED street light to perform dynamic scene clustering and divide it into several target use scenarios, and synchronously output the initial brightness adjustment parameters adapted to each target use scenario. The retrieval module is used to retrieve the historical operating data of the LED street light, construct a corresponding brightness decay prediction model, and predict the brightness decay amount within a preset time in the future based on the current running time and ambient temperature data. The processing module is used to pre-compensate and correct the initial brightness adjustment parameters according to the brightness attenuation amount, so as to output the corresponding target brightness control parameters, and control the output luminous flux of the LED street light according to the target brightness control parameters.

[0015] Furthermore, the transmission module is specifically used for: Based on the spatial location coordinates of each LED street light in the Mesh network, a corresponding spatiotemporal correlation matrix is ​​constructed. According to the spatiotemporal correlation matrix, the fused dataset is divided into time slices and spatial grids to extract the spatial propagation features of different time slices within the same spatial grid. The spatial propagation features are processed in a low dimension using the LLE algorithm to generate a corresponding low-dimensional spatiotemporal feature set, and the low-dimensional spatiotemporal feature set is divided into several basic scene clusters according to a preset index. The basic scene clusters are verified by using an adversarial generative network to generate corresponding target scene clusters. The target scene clusters are then iteratively optimized to generate the target usage scenario.

[0016] Furthermore, the transmission module is specifically used for: The core feature vectors of each target scene cluster are extracted, and the corresponding feature selection criteria are constructed in combination with the physical constraints of LED street lighting. Redundant and abnormal features with weights lower than a preset weight threshold are removed from the target scene clusters according to the feature selection criteria to generate the corresponding purified feature clusters. The cosine similarity and Euclidean distance between adjacent purified feature clusters are calculated, and the corresponding scene boundary association matrix is ​​constructed. The corresponding boundary features are simulated based on the scene boundary association matrix. The MAML algorithm from meta-learning is introduced to train a scene classifier to generate a corresponding scene classification model. The boundary features are then converted into the target usage scene through the scene classification model.

[0017] Furthermore, the retrieval module is specifically used for: Using the runtime as the time axis, the ambient temperature data is decomposed into instantaneous temperature value, temperature change rate, and temperature sub-characteristics of continuous constant temperature period, and synchronously correlated with the actual operating current of LED street lights to construct the corresponding load correlation factor. A temporal convolutional network is used to extract and enhance the dimensions of the runtime, temperature sub-features and the load correlation factors to generate a corresponding enhanced feature set. Based on the enhanced feature set, the current decay phase of the LED street light is detected by the K-means algorithm. The attenuation phase of the LED street light at different times is collected, and the brightness attenuation is predicted based on the dynamic switching of the attenuation phase.

[0018] Furthermore, the retrieval module is specifically used for: Collect full operational data corresponding to each switching moment of the attenuation phase, extract phase steady-state features and phase transient features from the full operational data, and generate the corresponding coupling matrix through principal component analysis; A meta-learning prediction framework is constructed, using the coupling matrix as the meta-training set, and a brightness prediction model adapted to each of the attenuation phases is trained. When a phase attenuation switch is detected, the model parameters of the brightness prediction model are adaptively adjusted through meta-learning, and the corresponding brightness attenuation amount is output.

[0019] Furthermore, the processing module is specifically used for: The trend features of the light intensity time series data are extracted through the edge computing node, and the corresponding light intensity change rate is detected based on the trend features; The brightness attenuation amount and the light intensity change rate are nonlinearly correlated to generate the scene light intensity attenuation value, and the power supply frequency fluctuation data is converted into a voltage stability coefficient. The scene light intensity attenuation value is weighted and calculated with the voltage stability coefficient, and the actual brightness feedback data of the LED street light is collected at the same time to output the target brightness control parameters accordingly.

[0020] Furthermore, the processing module is specifically used for: The steady-state value and instantaneous fluctuation value of the actual brightness feedback data are extracted by time-domain filtering, and the strong correlation features between the steady-state value and the instantaneous fluctuation value are enhanced by attention weight allocation strategy in order to remove interference components and generate corresponding brightness reference data. The difference between the brightness reference data and the theoretical brightness value of the LED street light is calculated to obtain the brightness deviation. The deviation is then combined with the scene light intensity attenuation value and the voltage stability coefficient to generate the corresponding initial brightness control parameters. The initial brightness control parameters are embedded into the corresponding hardware delay factor to generate the target brightness control parameters.

[0021] The third aspect of the present invention proposes: A computer includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the LED street light brightness adaptive adjustment method as described above.

[0022] The fourth aspect of the present invention proposes: A readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the LED street light brightness adaptive adjustment method as described above.

[0023] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0024] Figure 1 A flowchart of the LED street light brightness adaptive adjustment method provided in the first embodiment of the present invention; Figure 2 This is a structural block diagram of the LED street light brightness adaptive adjustment system provided in the third embodiment of the present invention.

[0025] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0026] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0027] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0029] Please see Figure 1 The image shows an adaptive brightness adjustment method for LED streetlights provided in the first embodiment of the present invention. This adaptive brightness adjustment method for LED streetlights can dynamically adapt to various usage scenarios of streetlights, thereby improving the energy-saving benefits of streetlights.

[0030] Specifically, this embodiment provides: An adaptive brightness adjustment method for LED streetlights specifically includes the following steps: Step S10: Construct a corresponding Mesh network based on the sensing units inside several LED streetlights in the target area. Collect light intensity time-series data, position-velocity correlation data of moving targets on the road surface, and power supply frequency fluctuation data in the target area through the Mesh network to create a corresponding fusion dataset. It's important to note that, firstly, a mesh network is constructed based on the built-in sensing units of the LED streetlights in the target area. The advantage of this network lies in the self-organizing communication between streetlight nodes, eliminating the need for a central base station. This reduces deployment costs and improves data transmission reliability and coverage. The mesh network synchronously collects three types of core data: light intensity time-series data (reflecting the dynamic changes in ambient natural light, such as gradual changes in light intensity at dusk and sudden drops in light intensity on rainy nights), location-speed correlation data of moving targets on the road surface (reflecting road traffic demand, such as peak traffic flow on main roads and scattered pedestrian traffic on side roads at night), and power supply frequency fluctuation data (characterizing power supply stability; voltage fluctuations directly affect the actual luminance of LEDs). These three types of data are integrated into a fusion dataset, providing comprehensive input for subsequent scene judgment and parameter adjustment, avoiding adjustment deviations caused by a single data dimension.

[0031] Step S20: The fused dataset is transmitted to the edge computing node of the LED street light to perform dynamic scene clustering and divide it into several target usage scenarios, and the initial brightness adjustment parameters adapted to each target usage scenario are output synchronously. It should be noted that, next, the fused dataset is transmitted to the streetlight edge computing node. The localized processing of edge computing reduces the latency of data upload to the cloud, meeting the real-time adjustment requirements of the streetlights. The edge node divides several target usage scenarios through dynamic scene clustering (such as distinguishing between "main roads during peak traffic hours" and "side roads with sparse pedestrians on rainy nights"), and outputs initial brightness adjustment parameters adapted to each scenario (such as adjusting the brightness to 80% of the rated value in peak traffic scenarios and 40% in sparse pedestrian scenarios), achieving a preliminary match between brightness and scenario requirements.

[0032] Step S30: Retrieve the historical operating data of the LED street light, construct the corresponding brightness decay prediction model, and predict the brightness decay amount within a preset time in the future based on the current running time and ambient temperature data. It should be noted that LED streetlights experience brightness decay as operating time increases and ambient temperature changes (e.g., LED chips age faster in high-temperature environments, and brightness may decrease by 15% after 10,000 hours of operation). Adjusting only the initial parameters will result in actual brightness being lower than the required value. Therefore, it is necessary to retrieve historical operating data of the streetlights (such as brightness decay records at different durations and temperatures) to construct a brightness decay prediction model. By combining the current operating time and ambient temperature data, the brightness decay amount within a preset time period can be accurately predicted, providing a quantitative basis for parameter compensation.

[0033] Step S40: Perform pre-compensation correction processing on the initial brightness adjustment parameters according to the brightness attenuation amount to output the corresponding target brightness control parameters, and control the output luminous flux of the LED street light according to the target brightness control parameters.

[0034] It should be noted that, finally, the initial brightness adjustment parameters are pre-compensated and corrected based on the brightness attenuation (e.g., if the predicted attenuation is 15%, the initial parameters are increased by 15% to offset the attenuation), and the target brightness control parameters are output and sent to the street light driver module to complete the brightness adjustment. The entire process realizes a closed loop of "environmental demand perception - equipment status adaptation - precise brightness output," which ensures both the safety of road lighting and energy-saving requirements and equipment lifespan.

[0035] Second Embodiment Furthermore, the step of transmitting the fused dataset to the edge computing node of the LED street light for dynamic scene clustering and segmentation into several target use scenarios includes: Based on the spatial location coordinates of each LED street light in the Mesh network, a corresponding spatiotemporal correlation matrix is ​​constructed. According to the spatiotemporal correlation matrix, the fused dataset is divided into time slices and spatial grids to extract the spatial propagation features of different time slices within the same spatial grid. The spatial propagation features are processed in a low dimension using the LLE algorithm to generate a corresponding low-dimensional spatiotemporal feature set, and the low-dimensional spatiotemporal feature set is divided into several basic scene clusters according to a preset index. The basic scene clusters are verified by using an adversarial generative network to generate corresponding target scene clusters. The target scene clusters are then iteratively optimized to generate the target usage scenario.

[0036] It should be noted that, firstly, a spatiotemporal correlation matrix is ​​constructed based on the spatial coordinates (such as latitude and longitude, road segment number) of each street light in the Mesh network. Each element in the matrix corresponds to the multi-source data correlation relationship of "a certain spatial grid and a certain time slice". The fused dataset is doubly divided into time slices (such as 5 minutes as a time unit) and spatial grids (such as 100 meters × 100 meters as a grid). Spatial propagation features of different time slices within the same spatial grid are extracted (such as the speed gradient of traffic flow moving from east to west within a certain grid, and the attenuation trend of light intensity from the edge to the center). This feature can accurately characterize the spatiotemporal dynamics of the scene and avoid confusion between scenes of "different road segments at the same time" and "different time segments at the same road segment".

[0037] Because spatial propagation features have high dimensionality (including multi-dimensional data such as light intensity, traffic flow, and power supply), direct clustering would result in computational redundancy. Therefore, the Local Linear Embedding (LLE) algorithm is used for low-dimensional processing. The LLE algorithm can reduce dimensionality while preserving the local geometric structure of the data, generating a low-dimensional spatiotemporal feature set. This reduces the computational burden on edge nodes while retaining the core features of the scene. Based on preset indicators (such as the magnitude of light intensity change and traffic flow density threshold), the low-dimensional feature set is divided into several basic scene clusters, completing the initial classification of the scenes.

[0038] To ensure that the feature distribution of the basic scene clusters conforms to the actual lighting scene patterns (avoiding "meaningless abnormal scenes" in algorithm clustering), a Generative Adversarial Network (GAN) is used to verify the feature distribution of the basic scene clusters. Specifically, the generator and discriminator of the GAN can iteratively optimize the feature distribution, making the data within the cluster more closely match the feature patterns of the real scene, generating target scene clusters. Then, the target scene clusters are iteratively optimized (such as merging clusters with extremely high feature similarity and splitting clusters with excessive feature differences), ultimately generating target usage scenes with clear boundaries and strong adaptability, providing accurate scene basis for the output of initial brightness parameters.

[0039] Furthermore, the step of iteratively optimizing the target scene cluster to generate the corresponding target usage scenario includes: The core feature vectors of each target scene cluster are extracted, and the corresponding feature selection criteria are constructed in combination with the physical constraints of LED street lighting. Redundant and abnormal features with weights lower than a preset weight threshold are removed from the target scene clusters according to the feature selection criteria to generate the corresponding purified feature clusters. The cosine similarity and Euclidean distance between adjacent purified feature clusters are calculated, and the corresponding scene boundary association matrix is ​​constructed. The corresponding boundary features are simulated based on the scene boundary association matrix. The MAML algorithm from meta-learning is introduced to train a scene classifier to generate a corresponding scene classification model. The boundary features are then converted into the target usage scene through the scene classification model.

[0040] It should be noted that, firstly, the core feature vectors of each target scene cluster are extracted (e.g., the core features of the "rainy night main road" scene are "low light intensity, high traffic density, and stable power supply"). Combined with the physical constraints of LED street lighting (e.g., the minimum brightness of main roads in urban road lighting standards is not less than 2 cd / m², and that of branch roads is not less than 0.75 cd / m², and the upper limit of brightness to avoid glare, etc.), feature selection criteria are constructed. Based on the criteria, redundant features (e.g., slight power supply fluctuations unrelated to lighting needs) and abnormal features (e.g., instantaneous false alarms of extremely high light intensity data from sensors) with weights below a preset threshold within the cluster are removed, generating purified feature clusters to ensure that the features of each cluster are strongly correlated with actual lighting needs.

[0041] To address the issue of blurred boundaries between adjacent scene clusters (such as the transition period between "peak traffic" and "off-peak traffic"), we calculate the cosine similarity (measuring the consistency of feature orientation) and Euclidean distance (measuring the difference in feature values) between adjacent purified feature clusters, constructing a scene boundary association matrix. The matrix elements quantify the association strength and degree of difference between clusters. Based on the matrix, we simulate boundary features (such as the mixed features of "traffic transition from peak to off-peak") to fill the gaps in scene classification.

[0042] Considering the diversity and dynamic changes in urban lighting scenarios (such as sudden road construction and temporary traffic control), traditional classification models struggle to quickly adapt to new scenarios. Therefore, the MAML algorithm from meta-learning is introduced to train a scene classifier. The advantage of the MAML algorithm lies in its ability to quickly adapt to new tasks with a small number of samples, generating a scene classification model with strong generalization capabilities. This model can uniformly classify boundary features and refined cluster features, transforming them into well-defined and comprehensive target usage scenarios, ensuring that each scenario accurately corresponds to the actual lighting needs of the road.

[0043] Furthermore, the step of predicting the brightness decay amount within a preset time period based on the current running time and ambient temperature data includes: Using the runtime as the time axis, the ambient temperature data is decomposed into instantaneous temperature value, temperature change rate, and temperature sub-characteristics of continuous constant temperature period, and synchronously correlated with the actual operating current of LED street lights to construct the corresponding load correlation factor. A temporal convolutional network is used to extract and enhance the dimensions of the runtime, temperature sub-features and the load correlation factors to generate a corresponding enhanced feature set. Based on the enhanced feature set, the current decay phase of the LED street light is detected by the K-means algorithm. The attenuation phase of the LED street light at different times is collected, and the brightness attenuation is predicted based on the dynamic switching of the attenuation phase.

[0044] It's important to note that, firstly, using runtime as the time axis, the ambient temperature data is broken down into three key sub-features: instantaneous temperature value (e.g., 35℃ at noon, 10℃ at night), temperature change rate (e.g., a 10℃ drop in temperature within 2 hours due to a cold wave), and continuous constant temperature period (e.g., 3 consecutive days above 30℃ in summer). Different temperature characteristics have varying impacts on LED degradation (e.g., sustained high temperatures accelerate phosphor aging, while sudden temperature changes cause stress damage to the chip encapsulation layer). Simultaneously, the actual operating current of the LED streetlights is correlated (current fluctuations exacerbate chip heating), constructing a load correlation factor. This factor quantifies the synergistic effect of "temperature-current-runtime," avoiding the bias of predictions based on a single factor.

[0045] A Temporal Convolutional Network (TCN) is employed to extract features and enhance the dimensionality of runtime, temperature sub-features, and load-related factors. The causal and dilated convolutional properties of the TCN effectively capture long-term dependencies in time-series data (e.g., the correlation between attenuation after 10,000 hours of runtime and the cumulative temperature effect over the previous 5,000 hours), generating an enhanced feature set. Based on this enhanced feature set, the K-means algorithm is used to detect the current attenuation phase of the LED streetlights (e.g., "initial stable phase," "mid-term slow attenuation phase," and "late-term rapid attenuation phase"). The attenuation patterns of different phases differ significantly (initial phase attenuation rate <5%, late-term phase attenuation rate can reach 10%-20%). Phase segmentation is a crucial prerequisite for accurate prediction.

[0046] Finally, the attenuation phase of the streetlights at different times is collected, and the dynamic switching pattern of the phase is analyzed (e.g., when the running time exceeds 8000 hours and the continuous high temperature exceeds 7 days, it will switch from the mid-term phase to the late-term phase). Combined with the attenuation rate of each phase, the brightness attenuation amount within a preset time in the future is accurately predicted, providing a quantitative basis for equipment status for subsequent parameter compensation.

[0047] Furthermore, the step of predicting the brightness attenuation amount based on the dynamic switching of the attenuation phase includes: Collect full operational data corresponding to each switching moment of the attenuation phase, extract phase steady-state features and phase transient features from the full operational data, and generate the corresponding coupling matrix through principal component analysis; A meta-learning prediction framework is constructed, using the coupling matrix as the meta-training set, and a brightness prediction model adapted to each of the attenuation phases is trained. When a phase attenuation switch is detected, the model parameters of the brightness prediction model are adaptively adjusted through meta-learning, and the corresponding brightness attenuation amount is output.

[0048] It should be noted that, firstly, full operational data (such as runtime, temperature peak, current fluctuation amplitude, and actual brightness value at each switching moment of the decay phase) is collected. Two types of core features are extracted from this data: steady-state phase features (such as the average decay rate and temperature adaptation range within a certain phase) and transient phase features (such as the temperature jump value and current fluctuation peak at the moment of switching). Principal component analysis (PCA) is used to reduce the dimensionality and couple these two types of features to generate a coupling matrix. This matrix can condense the key influencing factors of phase switching and reduce the dimensionality of the model input.

[0049] A meta-learning prediction framework is constructed, using the coupling matrix as the meta-training set. The core of meta-learning is to enable the model to learn "learning ability," rather than simply fitting the decay pattern of a single phase. Based on the meta-training set, a brightness prediction model adapted to each decay phase is trained. The model for each phase can quickly adapt to the decay characteristics of that phase (e.g., the later phase model strengthens the weights of high temperature and duration).

[0050] When a phase shift in attenuation is detected (e.g., from a mid-phase to a late-phase phase), the meta-learning framework can adaptively adjust the parameters of the prediction model (without retraining the entire model), quickly adapting to the attenuation pattern of the new phase and outputting accurate brightness attenuation. This adaptive adjustment mechanism ensures real-time prediction during phase shifts while improving the overall accuracy of attenuation prediction, avoiding undercompensation or overcompensation due to phase changes.

[0051] Furthermore, the step of performing pre-compensation correction processing on the initial brightness adjustment parameters based on the brightness attenuation amount to output the corresponding target brightness control parameters includes: The trend features of the light intensity time series data are extracted through the edge computing node, and the corresponding light intensity change rate is detected based on the trend features; The brightness attenuation amount and the light intensity change rate are nonlinearly correlated to generate the scene light intensity attenuation value, and the power supply frequency fluctuation data is converted into a voltage stability coefficient. The scene light intensity attenuation value is weighted and calculated with the voltage stability coefficient, and the actual brightness feedback data of the LED street light is collected at the same time to output the target brightness control parameters accordingly.

[0052] It should be noted that, firstly, trend features of light intensity time-series data (such as the slope of light intensity decrease and fluctuation frequency) are extracted through edge computing nodes, and based on this, the rate of change of light intensity is detected (e.g., the rate of change of light intensity during dusk is -5 cd / m²). min, rainy night -10cd / ㎡ The rate of change of light intensity directly reflects the urgency of the environment's demand for street light brightness. The greater the rate of change, the higher the response speed required for brightness adjustment.

[0053] The predicted brightness attenuation is nonlinearly correlated with the rate of change of light intensity (e.g., the greater the attenuation and the steeper the rate of change of light intensity, the higher the scene light intensity attenuation value) to generate a scene light intensity attenuation value. This value takes into account both the device's own attenuation and the changes in ambient light. At the same time, the power supply frequency fluctuation data is converted into a voltage stability coefficient (the closer the coefficient is to 1, the more stable the power supply). Voltage instability will cause the actual brightness of the LED to be lower than the set value (e.g., if the voltage decreases by 10%, the brightness may decrease by 8%). Therefore, this coefficient needs to be included in the parameter correction system.

[0054] The scene light intensity attenuation value and voltage stability coefficient are weighted and calculated (e.g., the scene light intensity attenuation value has a weight of 0.6 and the voltage stability coefficient has a weight of 0.4, which can be adjusted according to the actual scene). At the same time, the actual brightness feedback data of the LED street light is collected (e.g., the actual luminous brightness of the street light is detected in real time through a photosensitive sensor). The feedback data is used as the basis for closed-loop verification to correct the weighted calculation result. Finally, the target brightness control parameters are output to ensure that the parameters not only offset the equipment attenuation, but also adapt to changes in ambient light and power supply status.

[0055] Furthermore, the step of weighting the scene light intensity attenuation value with the voltage stability coefficient and simultaneously collecting the actual brightness feedback data of the LED streetlights to output the target brightness control parameters includes: The steady-state value and instantaneous fluctuation value of the actual brightness feedback data are extracted by time-domain filtering, and the strong correlation features between the steady-state value and the instantaneous fluctuation value are enhanced by attention weight allocation strategy in order to remove interference components and generate corresponding brightness reference data. The difference between the brightness reference data and the theoretical brightness value of the LED street light is calculated to obtain the brightness deviation. The deviation is then combined with the scene light intensity attenuation value and the voltage stability coefficient to generate the corresponding initial brightness control parameters. The initial brightness control parameters are embedded into the corresponding hardware delay factor to generate the target brightness control parameters.

[0056] It should be noted that, firstly, the actual brightness feedback data is subjected to time-domain filtering to separate steady-state values ​​(such as the stable brightness value of a street light emitting light normally) from instantaneous fluctuation values ​​(such as brightness fluctuations caused by instantaneous power supply fluctuations). An attention weight allocation strategy is used to strengthen the strong correlation features between steady-state values ​​(weight 0.8) and instantaneous fluctuation values ​​(such as brightness changes caused by voltage fluctuations lasting more than 10 seconds). Invalid components such as sensor noise and instantaneous electromagnetic interference are removed to generate brightness benchmark data, which can truly reflect the actual emitting state of the street light.

[0057] The brightness baseline data is compared with the theoretical brightness value of the LED street light to obtain the brightness deviation (e.g., if the baseline data is 75% of the theoretical value, the deviation is -25%). This deviation reflects the difference between the actual brightness and the theoretical value under the current parameters. Combined with the scene light intensity attenuation value and voltage stability coefficient, a composite calculation is performed (e.g., target parameter = theoretical adjustment value + scene attenuation compensation value + voltage fluctuation correction value - brightness deviation) to generate the initial brightness control parameters.

[0058] Because LED street light driver modules have hardware response delays (e.g., the driver circuit needs 50ms to complete brightness switching after an adjustment command is issued), directly using initial parameters would lead to adjustment lag. Therefore, the initial brightness control parameters need to be embedded with a hardware response delay factor (e.g., issuing adjustment commands earlier based on the delay time, or correcting the adjustment range of the parameters) to ultimately generate target brightness control parameters. These parameters take into account both multi-dimensional adaptation needs and hardware physical characteristics, ensuring the accuracy and real-time performance of street light brightness adjustment and completing the entire adaptive adjustment process.

[0059] Please see Figure 2 The third embodiment of the present invention provides: An adaptive brightness adjustment system for LED streetlights, wherein the system comprises: The acquisition module is used to construct a corresponding Mesh network based on the sensing units inside several LED streetlights in the target area. The Mesh network is used to acquire light intensity time-series data, position-velocity correlation data of moving targets on the road surface, and power supply frequency fluctuation data in the target area to create a corresponding fusion dataset. The transmission module is used to transmit the fused dataset to the edge computing node of the LED street light to perform dynamic scene clustering and divide it into several target use scenarios, and synchronously output the initial brightness adjustment parameters adapted to each target use scenario. The retrieval module is used to retrieve the historical operating data of the LED street light, construct a corresponding brightness decay prediction model, and predict the brightness decay amount within a preset time in the future based on the current running time and ambient temperature data. The processing module is used to pre-compensate and correct the initial brightness adjustment parameters according to the brightness attenuation amount, so as to output the corresponding target brightness control parameters, and control the output luminous flux of the LED street light according to the target brightness control parameters.

[0060] Furthermore, the transmission module is specifically used for: Based on the spatial location coordinates of each LED street light in the Mesh network, a corresponding spatiotemporal correlation matrix is ​​constructed. According to the spatiotemporal correlation matrix, the fused dataset is divided into time slices and spatial grids to extract the spatial propagation features of different time slices within the same spatial grid. The spatial propagation features are processed in a low dimension using the LLE algorithm to generate a corresponding low-dimensional spatiotemporal feature set, and the low-dimensional spatiotemporal feature set is divided into several basic scene clusters according to a preset index. The basic scene clusters are verified by using an adversarial generative network to generate corresponding target scene clusters. The target scene clusters are then iteratively optimized to generate the target usage scenario.

[0061] Furthermore, the transmission module is specifically used for: The core feature vectors of each target scene cluster are extracted, and the corresponding feature selection criteria are constructed in combination with the physical constraints of LED street lighting. Redundant and abnormal features with weights lower than a preset weight threshold are removed from the target scene clusters according to the feature selection criteria to generate the corresponding purified feature clusters. The cosine similarity and Euclidean distance between adjacent purified feature clusters are calculated, and the corresponding scene boundary association matrix is ​​constructed. The corresponding boundary features are simulated based on the scene boundary association matrix. The MAML algorithm from meta-learning is introduced to train a scene classifier to generate a corresponding scene classification model. The boundary features are then converted into the target usage scene through the scene classification model.

[0062] Furthermore, the retrieval module is specifically used for: Using the runtime as the time axis, the ambient temperature data is decomposed into instantaneous temperature value, temperature change rate, and temperature sub-characteristics of continuous constant temperature period, and synchronously correlated with the actual operating current of LED street lights to construct the corresponding load correlation factor. A temporal convolutional network is used to extract and enhance the dimensions of the runtime, temperature sub-features and the load correlation factors to generate a corresponding enhanced feature set. Based on the enhanced feature set, the current decay phase of the LED street light is detected by the K-means algorithm. The attenuation phase of the LED street light at different times is collected, and the brightness attenuation is predicted based on the dynamic switching of the attenuation phase.

[0063] Furthermore, the retrieval module is specifically used for: Collect full operational data corresponding to each switching moment of the attenuation phase, extract phase steady-state features and phase transient features from the full operational data, and generate the corresponding coupling matrix through principal component analysis; A meta-learning prediction framework is constructed, using the coupling matrix as the meta-training set, and a brightness prediction model adapted to each of the attenuation phases is trained. When a phase attenuation switch is detected, the model parameters of the brightness prediction model are adaptively adjusted through meta-learning, and the corresponding brightness attenuation amount is output.

[0064] Furthermore, the processing module is specifically used for: The trend features of the light intensity time series data are extracted through the edge computing node, and the corresponding light intensity change rate is detected based on the trend features; The brightness attenuation amount and the light intensity change rate are nonlinearly correlated to generate the scene light intensity attenuation value, and the power supply frequency fluctuation data is converted into a voltage stability coefficient. The scene light intensity attenuation value is weighted and calculated with the voltage stability coefficient, and the actual brightness feedback data of the LED street light is collected at the same time to output the target brightness control parameters accordingly.

[0065] Furthermore, the processing module is specifically used for: The steady-state value and instantaneous fluctuation value of the actual brightness feedback data are extracted by time-domain filtering, and the strong correlation features between the steady-state value and the instantaneous fluctuation value are enhanced by attention weight allocation strategy in order to remove interference components and generate corresponding brightness reference data. The difference between the brightness reference data and the theoretical brightness value of the LED street light is calculated to obtain the brightness deviation. The deviation is then combined with the scene light intensity attenuation value and the voltage stability coefficient to generate the corresponding initial brightness control parameters. The initial brightness control parameters are embedded into the corresponding hardware delay factor to generate the target brightness control parameters.

[0066] The fourth embodiment of the present invention provides a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the LED street light brightness adaptive adjustment method as described above.

[0067] The fifth embodiment of the present invention provides a readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the LED street light brightness adaptive adjustment method as described above.

[0068] In summary, the LED street light brightness adaptive adjustment method and system provided in the above embodiments of the present invention can dynamically control the street light to adapt to various usage scenarios, thereby improving the energy-saving benefits of the street light.

[0069] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0070] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0071] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0072] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0073] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0074] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A method for adaptive brightness adjustment of LED streetlights, characterized in that, The method includes: A corresponding Mesh network is constructed based on the sensing units inside several LED streetlights within the target area. The Mesh network is used to collect time-series data of light intensity, position-velocity correlation data of moving targets on the road surface, and power supply frequency fluctuation data within the target area to create a corresponding fusion dataset. The fused dataset is transmitted to the edge computing node of the LED street light to perform dynamic scene clustering and divide it into several target usage scenarios, and to synchronously output the initial brightness adjustment parameters adapted to each target usage scenario. The historical operating data of the LED street light is retrieved to construct a corresponding brightness decay prediction model, and the brightness decay amount within a preset time is predicted based on the current running time and ambient temperature data. The initial brightness adjustment parameters are pre-compensated and corrected based on the brightness attenuation amount to output the corresponding target brightness control parameters, and the output luminous flux of the LED street light is controlled accordingly based on the target brightness control parameters.

2. The LED street light brightness adaptive adjustment method according to claim 1, characterized in that, The step of transmitting the fused dataset to the edge computing node of the LED street light for dynamic scene clustering and segmentation into several target use scenarios includes: Based on the spatial location coordinates of each LED street light in the Mesh network, a corresponding spatiotemporal correlation matrix is ​​constructed. According to the spatiotemporal correlation matrix, the fused dataset is divided into time slices and spatial grids to extract the spatial propagation features of different time slices within the same spatial grid. The spatial propagation features are processed in a low dimension using the LLE algorithm to generate a corresponding low-dimensional spatiotemporal feature set, and the low-dimensional spatiotemporal feature set is divided into several basic scene clusters according to a preset index. The basic scene clusters are verified by using an adversarial generative network to generate corresponding target scene clusters. The target scene clusters are then iteratively optimized to generate the target usage scenario.

3. The LED street light brightness adaptive adjustment method according to claim 2, characterized in that, The step of iteratively optimizing the target scene cluster to generate the target use scenario includes: The core feature vectors of each target scene cluster are extracted, and the corresponding feature selection criteria are constructed in combination with the physical constraints of LED street lighting. Redundant and abnormal features with weights lower than a preset weight threshold are removed from the target scene clusters according to the feature selection criteria to generate the corresponding purified feature clusters. The cosine similarity and Euclidean distance between adjacent purified feature clusters are calculated, and the corresponding scene boundary association matrix is ​​constructed. The corresponding boundary features are simulated based on the scene boundary association matrix. The MAML algorithm from meta-learning is introduced to train a scene classifier to generate a corresponding scene classification model. The boundary features are then converted into the target usage scene through the scene classification model.

4. The LED street light brightness adaptive adjustment method according to claim 1, characterized in that, The step of predicting the brightness decay within a preset time based on the current running time and ambient temperature data includes: Using the runtime as the time axis, the ambient temperature data is decomposed into instantaneous temperature value, temperature change rate, and temperature sub-characteristics of continuous constant temperature period, and synchronously correlated with the actual operating current of LED street lights to construct the corresponding load correlation factor. A temporal convolutional network is used to extract and enhance the dimensions of the runtime, temperature sub-features and the load correlation factors to generate a corresponding enhanced feature set. Based on the enhanced feature set, the current decay phase of the LED street light is detected by the K-means algorithm. The attenuation phase of the LED street light at different times is collected, and the brightness attenuation is predicted based on the dynamic switching of the attenuation phase.

5. The LED street light brightness adaptive adjustment method according to claim 4, characterized in that, The step of predicting the brightness attenuation amount based on the dynamic switching of the attenuation phase includes: Collect full operational data corresponding to each switching moment of the attenuation phase, extract phase steady-state features and phase transient features from the full operational data, and generate the corresponding coupling matrix through principal component analysis; A meta-learning prediction framework is constructed, using the coupling matrix as the meta-training set, and a brightness prediction model adapted to each of the attenuation phases is trained. When a phase attenuation switch is detected, the model parameters of the brightness prediction model are adaptively adjusted through meta-learning, and the corresponding brightness attenuation amount is output.

6. The LED street light brightness adaptive adjustment method according to claim 1, characterized in that, The step of performing pre-compensation correction processing on the initial brightness adjustment parameters based on the brightness attenuation amount to output the corresponding target brightness control parameters includes: The trend features of the light intensity time series data are extracted through the edge computing node, and the corresponding light intensity change rate is detected based on the trend features; The brightness attenuation amount and the light intensity change rate are nonlinearly correlated to generate the scene light intensity attenuation value, and the power supply frequency fluctuation data is converted into a voltage stability coefficient. The scene light intensity attenuation value is weighted and calculated with the voltage stability coefficient, and the actual brightness feedback data of the LED street light is collected at the same time to output the target brightness control parameters accordingly.

7. The LED street light brightness adaptive adjustment method according to claim 6, characterized in that, The step of weighting the scene light intensity attenuation value with the voltage stability coefficient and simultaneously collecting the actual brightness feedback data of the LED streetlights to output the target brightness control parameters includes: The steady-state value and instantaneous fluctuation value of the actual brightness feedback data are extracted by time-domain filtering, and the strong correlation features between the steady-state value and the instantaneous fluctuation value are enhanced by attention weight allocation strategy in order to remove interference components and generate corresponding brightness reference data. The difference between the brightness reference data and the theoretical brightness value of the LED street light is calculated to obtain the brightness deviation. The deviation is then combined with the scene light intensity attenuation value and the voltage stability coefficient to generate the corresponding initial brightness control parameters. The initial brightness control parameters are embedded into the corresponding hardware delay factor to generate the target brightness control parameters.

8. An adaptive brightness adjustment system for LED streetlights, characterized in that, The system includes: The acquisition module is used to construct a corresponding Mesh network based on the sensing units inside several LED streetlights in the target area. The Mesh network is used to acquire light intensity time-series data, position-velocity correlation data of moving targets on the road surface, and power supply frequency fluctuation data in the target area to create a corresponding fusion dataset. The transmission module is used to transmit the fused dataset to the edge computing node of the LED street light to perform dynamic scene clustering and divide it into several target use scenarios, and synchronously output the initial brightness adjustment parameters adapted to each target use scenario. The retrieval module is used to retrieve the historical operating data of the LED street light, construct a corresponding brightness decay prediction model, and predict the brightness decay amount within a preset time in the future based on the current running time and ambient temperature data. The processing module is used to pre-compensate and correct the initial brightness adjustment parameters according to the brightness attenuation amount, so as to output the corresponding target brightness control parameters, and control the output luminous flux of the LED street light according to the target brightness control parameters.

9. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the LED street light brightness adaptive adjustment method as described in any one of claims 1 to 7.

10. A readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the LED street light brightness adaptive adjustment method as described in any one of claims 1 to 7.

Images