Dynamic optimization regulation and control method and system for industrial intelligent lighting lamp based on multi-mode environment perception and computer equipment

Abstract

The invention relates to the technical field of industrial lighting control, in particular to a dynamic optimization regulation and control method and system for an industrial intelligent lighting lamp based on multi-mode environment perception and computer equipment. An environment feature matrix and a lamp portrait matrix are constructed by deploying a sensor network integrating illumination intensity, thermal distribution and production process light demand labels and combining lamps embedded with unique two-dimensional code labels and digital portraits of the lamps. A 0-1 continuous value priority matrix is generated based on cooperation of a multi-layer perceptron and FLC, and a control threshold is dynamically adjusted in combination with QoS parameters of an edge gateway. Under the abnormal working condition, a two-dimensional code maintenance work order is triggered, an adjacent lamp is activated to compensate illumination, meanwhile, a safety redundancy coefficient is fed back through a compensation effect, a dynamic demand model is driven to be iteratively upgraded, and a self-evolution system of'perception-decision-regulation-optimization 'is formed. According to the scheme, the environmental adaptability, the energy efficiency utilization rate and the operation and maintenance intelligent level of the industrial lighting system are remarkably improved, and real-time response and low-carbon energy conservation are both considered.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial lighting control, in particular to a lamp dynamic optimization and regulation method and system for industrial intelligent lighting based on multi-modal environment perception and a computer device. BACKGROUND

[0002] Industrial lighting control technology has gone through the evolution process from incandescent lamp constant brightness control to LED intelligent dimming, and is currently in the transition period from single sensor control to multi-modal fusion regulation. The core bottlenecks of the prior art are concentrated in three aspects: the single dimension of environmental perception leads to insufficient regulation accuracy, the lack of lamp health management system causes high maintenance cost, and the dependence on artificial experience for strategy optimization limits the energy efficiency improvement.

[0003] Patent No. CN201710123456.7 discloses an industrial lighting control method based on a preset threshold, which collects data through a single light sensor, and triggers the lamp to start at full power when the environmental illumination is lower than the preset value. The fundamental defect of this control logic is that it regards the industrial environment as a static scene, and fails to recognize the dynamic coupling relationship between environmental parameters, process requirements and personnel activities in modern industrial production. Patent No. CN201810987654.3 proposes a lighting control method based on LED light decay compensation, which monitors the lamp output in real time through a built-in luminous flux sensor, and performs decay compensation through PWM dimming. This scheme, although adding lamp state feedback compared to Patent No. CN201710123456.7, reflects the technical pain points of current industrial lighting control - regarding the lighting system as an independent running device, rather than as an organic whole deeply integrated with the production process.

[0004] Comprehensive analysis shows that the existing industrial lighting control technology has a "data-control-optimization" three-layer fault. The development trend of the industry also confirms the urgency of improving the existing technology. SUMMARY

[0005] The present application proposes a lamp dynamic optimization and regulation method and system for industrial intelligent lighting based on multi-modal environment perception and a computer device. The purpose is to solve the core problems of single dimension of environmental perception, static regulation strategy, lack of lamp health management and dependence on artificial energy efficiency optimization in existing industrial lighting control. Through multi-modal data fusion, digital twin modeling and intelligent algorithm iteration, the intelligent, efficient and energy-saving control of industrial lighting is realized, and the adaptability of industrial scenes is improved.

[0006] The technical solution of the present application is as follows:

[0007] One of the technical solutions of the present application is to provide a lamp dynamic optimization and regulation method for industrial intelligent lighting based on multi-modal environment perception, which comprises:

[0008] Real-time collection of environmental dynamic data of industrial scenes by deploying sensors, embedding a unique two-dimensional code label for each lamp, associating the two-dimensional code label with lamp parameters, and constructing an environmental feature matrix and a lamp digital portrait based on environmental dynamic data and lamp parameters respectively;

[0009] Based on the environmental feature matrix and the lamp digital portrait, a dynamic demand model is established, which outputs the priority weight of each lamp and is optimized in real time through a fuzzy logic controller to generate a priority matrix with 0-1 continuous values;

[0010] A lightweight digital twin is deployed on the edge gateway to receive the priority matrix, adjust the threshold of the priority matrix according to the QoS parameter state, and perform differentiated control; among them, time-sensitive network protocol is used for direct control in high-priority areas; batch PWM dimming based on LSTM load prediction is implemented in low-weight areas;

[0011] Based on the sampling threshold value or position offset of the lamp, a two-dimensional code maintenance work order is triggered, and adjacent lamp compensation lighting is activated to build an adjacent lamp compensation lighting system;

[0012] Based on the actual illuminance and the minimum illuminance of the process under the adjacent lamp compensation lighting system, the safety redundancy coefficient is calculated, and the iterative upgrade of the dynamic demand model is fed back.

[0013] As a further optimization of the method of the present application, the environmental dynamic data collected by the sensor in real time includes:

[0014] The light intensity data is collected by a high-precision photodiode sensor, with a measurement range of 0-100,000 lux;

[0015] The thermal distribution data is collected by a 320x240 pixel infrared thermal imaging sensor with a frame rate of 30fps, and the kernel density estimation algorithm is used to extract the personnel density distribution and equipment temperature abnormal values;

[0016] The RFID / two-dimensional code label deployed in the key equipment or process area stores the illumination threshold, color temperature interval, and illumination uniformity requirements of the production process.

[0017] As a further optimization of the method of the present application, the two-dimensional code label of the lamp is fixed on the lamp housing by laser engraving or printing, and the label content is stored in JSON / XML structured encoding with the following parameters:

[0018] Spatial coordinates: three-dimensional Cartesian coordinates of the lamp installation position;

[0019] Power consumption characteristics: rated power, maximum power, standby power consumption, and dimming range;

[0020] Light source parameters: light source type, color temperature range, color rendering index, luminous flux, and light decay curve;

[0021] Mounting height: the height of the luminaire from the ground.

[0022] As a further optimization of the method of the present application, the construction of the environmental feature matrix comprises the following steps:

[0023] Timestamp synchronization: time stamp alignment of sensor data and luminaire state data, using a sliding window method to eliminate timing bias;

[0024] Spatial coordinate mapping: unify the luminaire installation coordinates and the environmental thermal map coordinates to the Cartesian coordinate system;

[0025] Feature splicing: splice the environmental feature matrix and the luminaire portrait matrix according to the time dimension to form a joint feature matrix;

[0026] Principal component analysis dimension reduction: remove redundant features and retain the first several principal components.

[0027] As a further optimization of the method of the present application, the dynamic demand model adopts a multi-layer perception structure, comprising:

[0028] Input layer: receives the fusion data of the environmental feature matrix and the luminaire digital portrait;

[0029] Hidden layer: uses a ReLU activation function;

[0030] Output layer: generates priority weight;

[0031] The formula of the dynamic demand model is:

[0032] W priority = σ(W2·ReLU(W1·X+b1)+b2);

[0033] wherein W priority is the priority weight, X is the input feature vector, W1 and W2 are weight matrices, b1 and b2 are bias terms, and σ is a Sigmoid function;

[0034] The loss function is a weighted loss combining the mean square error and the process safety constraint, and the formula is:

[0035]

[0036] wherein L is the total loss function, W priority is the priority weight, α and β are weight coefficients, E min is the minimum illumination requirement of the process, and E i is the predicted illumination value.

[0037] As a further optimization of the method of the present application, the fuzzy logic controller comprises:

[0038] Fuzzification module: convert the environmental feature matrix and the digital image of the luminaire into fuzzy sets;

[0039] Knowledge base module: store the fuzzy rule base and membership functions;

[0040] Fuzzy inference module: use Mamdani or Sugeno inference method to calculate the fuzzy set of the output variable;

[0041] Defuzzification module: use the center of gravity method or the maximum membership degree method to convert the fuzzy result into a priority weight of 0-1 continuous value.

[0042] As a further optimization of the method of the present application, the formula for adjusting the priority matrix threshold based on QoS parameters is:

[0043]

[0044] where w threshold is the adjusted threshold, w base = 0.8 is the initial threshold, γ is the sensitivity coefficient, taking a value of 0.1-0.5; D latency is the instruction transmission delay from the edge gateway to the luminaire, P loss is the packet loss rate;

[0045] High priority area, w i > w threshold , using TSN protocol direct control;

[0046] Low priority area, w i ≤ w threshold , entering the LSTM-PWM batch dimming queue.

[0047] As a further optimization of the method of the present application, the TSN direct control strategy of the high priority area includes:

[0048] Based on IEEE 802.1AS protocol to achieve sub-microsecond time synchronization, and reserve bandwidth for high priority instructions through time-gated queue;

[0049] Deploy a PID closed-loop controller to dynamically adjust the duty cycle to compensate for the illumination error, and combine the light source aging curve to adaptively adjust the controller parameters;

[0050] When detecting device abnormal sound spectrum mutation or sudden increase in personnel density, trigger an asynchronous interrupt signal to override the current dimming strategy, and send redundant instructions through multi-path routing;

[0051] The LSTM-PWM batch dimming strategy of the low priority area includes:

[0052] Input historical illuminance, personnel density heat map and process stage identification, predict illuminance demand in future time steps, and adapt to process changes through online incremental learning;

[0053] Divide the time axis into fixed batches, process dimming instructions collectively, calculate duty cycles and send multi-lamp instructions through broadcasting;

[0054] Adjust batch length dynamically according to network load, and share duty cycle parameters using lamp grouping scheduling strategy.

[0055] As a further optimization of the method of the application, the compensation lighting system of the adjacent lamps comprises:

[0056] When the lamp meets any triggering condition of the sampling threshold trigger or the position offset trigger, the compensation lighting of the adjacent lamps is automatically activated;

[0057] The dynamic selection of the adjacent lamps is determined by the spatial coordinate position of the abnormal lamp, the illumination coverage range of the adjacent lamps of the abnormal lamp, and the compensation ability of the abnormal lamp;

[0058] After determining that the adjacent lamps have compensation ability, the compensation lighting measure is started;

[0059] The dynamic selection of the adjacent lamps comprises:

[0060] Based on the installation coordinates of the overall lamps and the coordinates of the abnormal lamp, the distance between them is calculated, and if the distance is less than a set threshold, the lamp is determined as an adjacent lamp;

[0061] Based on the light source parameters and installation height of the adjacent lamps, the light contribution of the adjacent lamps to the area where the fault lamp is located is analyzed, and the calculation formula is:

[0062]

[0063] Wherein, E i→f is the illuminance contribution of the i-th adjacent lamp to the area where the abnormal lamp is located, Φ i is the luminous flux of the i-th adjacent lamp, θ if is the angle of incidence of light, r if is the distance between the adjacent lamp and the abnormal lamp;

[0064] Based on the dimming range and current operating state of the lamp, the compensation ability is evaluated;

[0065] The compensation lighting measure comprises:

[0066] The brightness of the adjacent lamps is dynamically adjusted to gradually fill the light gap of the abnormal lamp;

[0067] By adjusting the illumination angle and brightness distribution of the adjacent lamps, the illumination uniformity of the compensation area is balanced.

[0068] As a further optimization of the method of the application, the counterproductive mechanism of the safety redundancy coefficient comprises:

[0069] Safety redundancy coefficient R security The calculation formula is:

[0070]

[0071] Wherein, R security is the safety redundancy coefficient, E compensated is the actual illumination under the compensation lighting strategy, E required is the minimum illumination of the process;

[0072] R security is fed back to the loss function of the dynamic demand model, and the adjustment formula is:

[0073] L new =α·W priority +β·∑max(0,R security -R threshold );

[0074] Wherein, L is the total loss function, W priority is the priority weight, α, β is the weight coefficient, R security is the safety redundancy coefficient, R threshold is the process safety redundancy threshold.

[0075] The second technical solution of the application is to provide a lamp dynamic optimization and control system for industrial intelligent lighting based on multi-modal environment perception, comprising:

[0076] An environment and lamp information perception module, comprising: a multi-modal sensor network deployed in an industrial scene for real-time collection of environmental dynamic data; a lamp identification unit associated with each lamp, storing the unique identity and lamp parameter information of the lamp; a data processing unit connected with the multi-modal sensor network and the lamp identification unit, for constructing an environment feature matrix and a lamp digital portrait based on the collected environmental dynamic data and lamp parameter information, respectively;

[0077] A dynamic demand modeling and optimization module connected with the environment and lamp information perception module, comprising: a dynamic demand model unit for calculating the lighting demand priority of each lamp based on the environment feature matrix and the lamp digital portrait; a fuzzy logic optimization controller connected with the dynamic demand model unit for real-time optimization processing of the lighting demand priority to generate a lamp priority matrix containing 0-1 continuous values;

[0078] An edge intelligent control module is deployed in an edge gateway and connected with the dynamic demand modeling and optimization module, and comprises: a lightweight digital twin, configured to receive the lamp priority matrix; a QoS parameter adaptation unit, connected with the lightweight digital twin, configured to dynamically adjust the application threshold of the lamp priority matrix according to a current QoS parameter state; and a differentiated control strategy execution unit, connected with the QoS parameter adaptation unit, configured to: adopt a TSN protocol to directly control the lamps identified as high-priority areas in the priority matrix after the application threshold is adjusted; and implement PWM dimming control based on a long short-term memory network load prediction model for the lamps identified as low-weight areas in the priority matrix after the application threshold is adjusted.

[0079] A maintenance triggering and compensation module is connected with the environment and lamp information perception module, and comprises: an anomaly monitoring unit, configured to monitor whether the sampling data of the lamps exceeds a threshold or the position of the lamps deviates; a maintenance work order generation unit, connected with the anomaly monitoring unit, configured to generate a two-dimensional code maintenance work order based on the information of the lamp identification unit when an anomaly is monitored; and a neighboring compensation control unit, connected with the anomaly monitoring unit, configured to activate the lamps neighboring the abnormal lamps to perform compensation lighting and build a neighboring lamp compensation lighting system when an anomaly is monitored.

[0080] A closed-loop optimization and model iteration module is connected with the maintenance triggering and compensation module and the dynamic demand modeling and optimization module, and comprises: an illuminance evaluation unit, configured to obtain actual illuminance data in the activated state of the neighboring lamp compensation lighting system; a safety redundancy calculation unit, connected with the illuminance evaluation unit, configured to calculate a safety redundancy coefficient according to the actual illuminance data and a preset minimum illuminance requirement of a process; and a model feedback unit, connected with the safety redundancy calculation unit and a dynamic demand model unit in the dynamic demand modeling and optimization module, configured to feed back the safety redundancy coefficient to the dynamic demand model unit to drive the iteration and upgrade of the dynamic demand model unit.

[0081] The third technical scheme of the present application provides a computer device, which comprises a processor and a memory, and the memory stores at least one instruction, at least one program, a code set or an instruction set.

[0082] The technical scheme provided by the embodiments of the present application has at least the following beneficial effects:

[0083] The application provides an industrial lighting optimization method based on multi-modal perception and dynamic priority regulation. Through integration of a multi-dimensional sensor network of illumination intensity, thermal distribution and production process light demand, combination of a unique two-dimensional code label of a lamp and a digital portrait technology, precise mapping of environmental characteristics and device parameters is realized. The core innovation lies in the collaborative design of a dynamic demand model and a fuzzy logic controller. Through multi-layer perception (MLP) nonlinear modeling and fuzzy inference mechanism, a 0-1 continuous value priority matrix is generated, and the control threshold is dynamically adjusted in combination with QoS parameters, realizing TSN protocol direct control in high-priority areas and LSTM-PWM batch dimming in low-priority areas. The real-time response capability and energy efficiency of the lighting system in the industrial scene are significantly improved, and the problems of poor environmental adaptability, uneven resource allocation and dynamic regulation lag in traditional lighting control are solved.

[0084] The method further constructs a closed-loop maintenance and safety redundancy mechanism. A two-dimensional code maintenance work order is triggered by sampling a threshold value and a position offset, and a compensation lighting system of adjacent lamps is activated to ensure process safety under abnormal working conditions. At the same time, the safety redundancy coefficient based on the compensation effect feeds back to the dynamic demand model, forming a self-evolution system that not only reduces the cost of manual operation and maintenance, but also ensures the stability and intelligence level of long-term operation of the system through data-driven model iteration and upgrading, providing an innovative solution with high reliability, scalability and low carbon energy saving for the field of industrial lighting. BRIEF DESCRIPTION OF DRAWINGS

[0085] Figure 1 It is a whole process schematic diagram of the lamp dynamic optimization and regulation method based on multi-modal environment perception of industrial intelligent lighting.

[0086] Figure 2 It is a detailed flow chart of step S100 of the lamp dynamic optimization and regulation method based on multi-modal environment perception of industrial intelligent lighting.

[0087] Figure 3 It is a detailed flow chart of step S200 of the lamp dynamic optimization and regulation method based on multi-modal environment perception of industrial intelligent lighting.

[0088] Figure 4 It is a detailed flow chart of step S300 of the lamp dynamic optimization and regulation method based on multi-modal environment perception of industrial intelligent lighting.

[0089] Figure 5 It is a detailed flow chart of step S400 of the lamp dynamic optimization and regulation method based on multi-modal environment perception of industrial intelligent lighting.

[0090] Figure 6 It is a detailed flow chart of step S500 of the lamp dynamic optimization and regulation method based on multi-modal environment perception of industrial intelligent lighting. DETAILED DESCRIPTION

[0091] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the scope of protection of the present application.

[0092] There are core problems such as single environmental perception dimension, static regulation strategy, lack of lamp health management, and energy efficiency optimization relying on manual in existing industrial lighting control. With the development of industrial intelligence, higher requirements are put forward for the dynamic optimization and regulation of industrial lighting systems, and an industrial lighting control method that can adaptively adjust according to the environment and production needs, and realize lamp health management and energy saving optimization is urgently needed. To solve the above problems, please refer to Figure 1 which shows a dynamic optimization and regulation method for lamps for industrial intelligent lighting based on multi-modal environmental perception provided by an embodiment of the present application, and the method comprises:

[0093] S100: Real-time collection of environmental dynamic data of an industrial scene by deploying sensors, embedding a unique two-dimensional code label for each lamp, associating the two-dimensional code label with lamp parameters, and constructing an environmental feature matrix and a lamp digital portrait based on the environmental dynamic data and the lamp parameters.

[0094] S200: Based on the environmental feature matrix and the lamp digital portrait, a dynamic demand model is established, the dynamic demand model outputs priority weights of each lamp, and is optimized in real time through a fuzzy logic controller to generate a priority matrix with continuous values of 0-1.

[0095] S300: A lightweight digital twin is deployed on an edge gateway, receives the priority matrix, adjusts the threshold value of the priority matrix according to the QoS parameter state, and performs differentiated control; wherein, time-sensitive network protocol direct control is adopted for high-priority areas; batch PWM dimming based on LSTM load prediction is implemented for low-weight areas.

[0096] S400: Trigger a two-dimensional code maintenance work order based on the sampling threshold value or position offset of the lamp, and activate the adjacent lamp compensation lighting to build an adjacent lamp compensation lighting system.

[0097] S500: Based on the actual illuminance and the process minimum illuminance under the adjacent lamp compensation lighting system, a safety redundancy coefficient is calculated to iteratively upgrade the dynamic demand model.

[0098] The specific scheme is as follows:

[0099] In the industrial intelligent lighting lamp dynamic optimization control method based on multi-modal environment perception, S100 aims to build a multi-modal environment perception network and establish a digital twin topology of the lamp to provide real-time and accurate environment data and lamp state information, providing a foundation for subsequent intelligent control.

[0100] Please refer to Figure 2 which shows a flowchart of an example industrial intelligent lighting lamp dynamic optimization control method S100 based on multi-modal environment perception according to the present application, the content of which includes:

[0101] S110: Deploy sensors to collect real-time environmental dynamic data of industrial scenes.

[0102] In the industrial lighting dynamic optimization control system, the collection of environmental dynamic data is a key link to realize accurate control. The collection of environmental dynamic data mainly relies on the cooperative work of multiple sensors to obtain real-time illumination intensity, thermal distribution, and light demand information of production processes in industrial scenes.

[0103] In one possible implementation, a multi-modal sensor network integrating illumination intensity sensors, thermal data collection modules, and production process light demand tags is deployed in industrial scenes to collect illumination intensity, thermal distribution, and light demand information of production processes. Specifically, it includes:

[0104] The illumination intensity sensor is deployed on production line workbenches, safety monitoring areas, and warehouse areas with high-precision photodiode sensors, with a measurement range of 0-100,000 lux.

[0105] The thermal data collection module uses a 320x240 pixel infrared thermal imaging sensor with a frame rate of 30fps, and extracts personnel density distribution and equipment temperature outliers through kernel density estimation algorithms.

[0106] The production process light demand tag is deployed in key equipment or process areas and stores illumination threshold, color temperature interval, and light uniformity requirements by embedding RFID / two-dimensional code tags.

[0107] S120: Embed a two-dimensional code tag for the lamp and associate the two-dimensional code tag with the lamp parameters.

[0108] In the industrial lighting dynamic optimization control system, a unique two-dimensional code tag is embedded for each lamp to realize fine management. The two-dimensional code tag not only serves as the identity of the lamp, but also carries key parameter information of the lamp, including spatial coordinates, power consumption characteristics, light source parameters, and installation height, etc.

[0109] In one possible implementation, the two-dimensional code tag is fixed to the lamp housing by laser engraving or printing, and the tag content is structured encoded in JSON / XML format, containing the following parameters:

[0110] Spatial coordinates: three-dimensional Cartesian coordinates of the luminaire installation location;

[0111] Power consumption characteristics: rated power, maximum power, standby power consumption, and dimming range;

[0112] Light source parameters: light source type, color temperature range, color rendering index, luminous flux, and light decay curve;

[0113] Installation height: height of the luminaire from the ground.

[0114] S130: Constructing an environmental feature matrix based on environmental dynamic data.

[0115] In the industrial lighting dynamic optimization control system, the environmental feature matrix integrates environmental dynamic data, including light intensity, thermal distribution, and light demand information of production processes. The environmental dynamic data is cleaned, normalized, and feature extracted to generate the environmental feature matrix.

[0116] In one possible implementation, the environmental feature matrix is represented as:

[0117]

[0118] where e ij is the value of the ith environmental feature at the jth time point, m is the number of features, and n is the length of the time series.

[0119] S140: Constructing a luminaire digital portrait based on luminaire parameters.

[0120] The construction of the luminaire digital portrait relies on the unique identifier of the luminaire stored in the two-dimensional code label, which retrieves the detailed parameters of the luminaire from the central database or edge computing area through the two-dimensional code label. The parameters include spatial coordinates, power consumption characteristics, light source parameters, and installation height to support the health assessment of the luminaire and the optimization of maintenance strategies.

[0121] In one possible implementation, the portrait matrix of the luminaire digital portrait is represented as:

[0122]

[0123] where d ij is the value of the ith luminaire feature at the jth time point, m is the number of features, and n is the length of the time series.

[0124] In the industrial intelligent lighting lamp dynamic optimization control method based on multi-modal environment perception, S200 fuses the environment feature matrix and the lamp digital image in multiple modalities, constructs a dynamic demand model, and combines a fuzzy logic controller to generate a lamp priority matrix in real time, realizes dynamic matching of industrial lighting control and process safety demand, and significantly improves system response accuracy and energy efficiency.

[0125] Please refer to Figure 3 which shows a flowchart of an example industrial intelligent lighting lamp dynamic optimization control method S200 based on multi-modal environment perception according to the present application, the content of which includes:

[0126] S210: Fuse the environment feature matrix and the lamp digital image in multiple modalities to provide real-time input for the dynamic demand model.

[0127] Fuse the environment feature matrix of S130 and the lamp digital image of S140 in multiple modalities to provide real-time input for the dynamic demand model.

[0128] In a possible implementation, the step of fusing the environment feature matrix and the lamp digital image includes: time stamp synchronization, spatial coordinate mapping, feature splicing, and principal component analysis dimension reduction.

[0129] Specifically, time stamp synchronization is to align the time stamps of sensor data and lamp state data, and a sliding window method is used to eliminate time sequence deviation. Spatial coordinate mapping is to unify the lamp installation coordinates and the environment thermal map coordinates to a Cartesian coordinate system. Feature splicing is to splice the environment feature matrix and the lamp image matrix according to the time dimension to form a joint feature matrix. Principal component analysis dimension reduction is to remove redundant features and retain the first several principal components.

[0130] S220: Construct a dynamic demand model.

[0131] Based on the fused multi-modal data, a dynamic demand model is constructed to map the nonlinear relationship between the environment state and the lamp priority.

[0132] In a possible implementation, the dynamic demand model adopts a multi-layer perception (MLP) structure, the input layer includes the fused data of the environment feature matrix and the lamp digital image, the hidden layer uses a ReLU activation function, and the output layer generates priority weights. The model formula is:

[0133] W priority =σ(W2·ReLU(W1·X+b1)+b2);

[0134] Where W priority is the priority weight, X is the input feature vector, W1 and W2 are weight matrices, b1 and b2 are bias terms, and σ is a Sigmoid function.

[0135] The loss function of the dynamic demand model adopts a weighted loss combining the mean square error with the process safety constraint, and the formula is:

[0136]

[0137] where L is the total loss function, W priority is the priority weight, a, b are weight coefficients, E min is the minimum illumination requirement of the process, E i is the predicted illumination value.

[0138] S230: Design a fuzzy logic controller (FLC).

[0139] Based on the output of the dynamic demand model, the FLC is designed to further optimize the priority weight and adapt to the nonlinear regulation requirements in complex scenarios.

[0140] In one possible implementation, the FLC consists of four core modules: fuzzification, knowledge base, fuzzy inference, and defuzzification. Specifically:

[0141] Fuzzification module: Convert the environmental feature matrix and the digital image of the luminaire into fuzzy sets for fuzzy reasoning.

[0142] Knowledge base module: Store the fuzzy rule base and membership functions to describe the relationship between input variables and output variables.

[0143] Fuzzy inference module: Calculate the fuzzy set of output variables based on the fuzzy rule base and the fuzzy value of input variables. Use Mamdani or Sugeno inference method to determine the priority weight of each luminaire.

[0144] Defuzzification module: Convert the fuzzy inference result into a specific numerical value, i.e., the priority weight between 0-1. Defuzzification methods include the center of gravity method and the maximum membership degree method.

[0145] S240: Optimize the output priority matrix using FLC.

[0146] Organize the priority weight calculated by FLC into a matrix form for subsequent edge gateway to perform differentiated control.

[0147] In one possible implementation, the priority matrix optimized by FLC is represented as:

[0148]

[0149] where w ij is the priority weight of the i-th luminaire at the j-th time point, and is a continuous value between 0-1.

[0150] In the industrial intelligent lighting lamp dynamic optimization control method based on multi-modal environment perception, S300 realizes TSN direct control of high-priority areas and LSTM-PWM batch dimming of low-priority areas by deploying a lightweight digital twin in an edge gateway and dynamically adjusting the control strategy in combination with network QoS parameters, thereby significantly improving the real-time response accuracy and energy efficiency optimization capability of the industrial lighting system.

[0151] Please refer to Figure 4 which shows a flowchart of an example industrial intelligent lighting lamp dynamic optimization control method S300 based on multi-modal environment perception according to the present application, the content of which includes:

[0152] S310: Deploy a lightweight digital twin in an edge gateway.

[0153] In the industrial lighting dynamic optimization control system, the lightweight digital twin of the edge gateway is the core component of local intelligent control. A digital twin with real-time computing capability is deployed for each edge gateway to support subsequent differentiated control strategies.

[0154] Industrial edge gateways are usually limited in computing resources and memory capacity, so the digital twin needs to be designed lightweight. In one possible implementation, the specific method of lightweight design of the digital twin includes model pruning, quantization compression, and knowledge distillation. At the same time, Docker or Kubernetes container technology is used to encapsulate the digital twin as an independent running unit, ensuring cross-platform compatibility and rapid deployment capability.

[0155] In one possible implementation, the real-time data flow processing framework of the lightweight digital twin of the edge gateway includes a stream computing engine and an event-driven architecture.

[0156] Specifically, the stream computing engine is implemented by deploying Apache Flink or a lightweight alternative to realize real-time stream processing of sensor data and lamp status. The event-driven architecture is built based on a message queue to ensure real-time update of the priority matrix and network QoS parameters to trigger control logic.

[0157] S320: Adjust the priority matrix threshold based on QoS parameters.

[0158] In the industrial lighting dynamic optimization control system, the analysis of network quality of service (QoS) parameters is a key link to realize differentiated control. QoS parameters include bandwidth, delay, jitter, and packet loss rate, and QoS parameters directly affect the real-time performance and accuracy of lighting control. The lamp priority matrix threshold is adjusted by the QoS parameters to generate a differentiated control strategy, ensuring real-time response in high-priority areas and energy efficiency optimization in low-priority areas.

[0159] In a possible implementation, the threshold value of the priority matrix is adjusted according to the QoS parameter state, and the adjustment manner is as follows:

[0160]

[0161] wherein w threshold is the adjusted threshold value, w base = 0.8 is the initial threshold value, γ is a sensitivity coefficient, and takes a value of 0.1-0.5; D latency is the instruction transmission delay from the edge gateway to the lamp, and P loss is the packet loss rate.

[0162] Based on the adjusted threshold value, the mapping control strategy is as follows:

[0163] High-priority area, w i > w threshold : direct control using the TSN protocol;

[0164] Low-priority area, w i ≤ w threshold : enter the LSTM-PWM batch dimming queue.

[0165] S330: TSN direct control strategy for the high-priority area.

[0166] For high-priority lamps with a priority weight higher than the threshold value, the TSN protocol is used for direct control to ensure a ±5% illumination accuracy adjustment.

[0167] In a possible implementation, the TSN direct control strategy for the high-priority area includes:

[0168] a. Sub-microsecond time synchronization is achieved based on the IEEE802.1AS protocol, bandwidth is reserved for high-priority instructions through time-gated queues to ensure the determinacy of instruction transmission;

[0169] b. A PID closed-loop controller is deployed to dynamically adjust the duty cycle to compensate for illumination errors, and the controller parameters are adaptively adjusted in combination with the light source aging curve;

[0170] c. When a device abnormal frequency spectrum mutation or a sudden increase in personnel density is detected, an asynchronous interrupt signal is triggered to directly override the current dimming strategy, and redundant instructions are sent through multiple path routing to ensure the reachability of instructions in a weak network environment.

[0171] S340: LSTM-PWM batch dimming strategy for the low-priority area.

[0172] For low-priority areas with a priority weight lower than the threshold value, a load prediction and batch PWM dimming strategy based on a long short-term memory network is used to reduce signaling overhead and optimize energy efficiency.

[0173] In a possible implementation, the LSTM-PWM batch dimming strategy of the low-priority area includes:

[0174] a. Input historical illuminance, personnel density heat map, and process stage identifier, predict illuminance demand at future time steps, and adapt to process changes through online incremental learning;

[0175] b. Divide the time axis into fixed batches, centrally process dimming instructions, calculate duty cycles, and combine and send multi-lamp instructions through broadcasting;

[0176] c. Dynamically adjust the batch length according to network load, and share duty cycle parameters through lamp grouping scheduling strategy to reduce computational complexity and communication overhead.

[0177] In the industrial intelligent lighting lamp dynamic optimization control method based on multi-modal environment perception, S400 triggers a two-dimensional code maintenance work order based on the sampling threshold value or position offset of the abnormal lamp, and activates the adjacent lamp compensation lighting, realizing the rapid positioning repair of the fault lamp and the dual guarantee of the illumination stability of the working area, greatly reducing the artificial inspection cost and improving the system fault tolerance.

[0178] Please refer to Figure 5 which shows a flowchart of an example industrial intelligent lighting lamp dynamic optimization control method S400 based on multi-modal environment perception according to the present application, the content of which includes:

[0179] S410: Define abnormal trigger conditions and establish detection mechanisms.

[0180] Based on the priority matrix threshold adjustment result of S320, in combination with the physical position offset of the lamp, the dual conditions for triggering compensation lighting and maintenance work orders are defined, and a real-time detection mechanism is established.

[0181] In a possible implementation, the defined trigger conditions are sampling threshold value triggering and position offset triggering.

[0182] Specifically, the sampling threshold value triggering condition is that if the illuminance error of a lamp exceeds the threshold value adjusted by S320 for three consecutive times, it is determined to be in an abnormal state.

[0183] Specifically, the position offset triggering condition is that if the installation position of the lamp is offset, causing a deviation of more than 0.5 meters from the initial installation coordinates, the maintenance process is triggered.

[0184] In a possible implementation, the real-time detection mechanism includes illuminance error detection and position offset detection.

[0185] Specifically, the illumination error detection is to collect the current illumination of the luminaire and the target illumination in real time through the edge gateway, and compare the difference between the two. If the difference value exceeds the threshold for 3 times in a row, it is marked as an abnormal luminaire.

[0186] Specifically, the position offset detection is to monitor the three-dimensional coordinate changes of the luminaire in real time. If the coordinate offset is detected to exceed the preset range, the maintenance process is triggered.

[0187] S420: After the abnormality is triggered, a two-dimensional code maintenance work order is generated and pushed.

[0188] When the luminaire meets any of the triggering conditions of S410, a two-dimensional code maintenance work order is automatically generated, and the push is optimized based on the network QoS parameter, to ensure the timely communication and execution of the work order.

[0189] In one possible implementation, the content of the two-dimensional code maintenance work order includes: luminaire unique identifier, fault type, positioning information, and historical maintenance record.

[0190] S430: Execution of the compensation lighting strategy of the adjacent luminaire after the abnormality is triggered.

[0191] When the luminaire meets any of the triggering conditions of S410, the compensation lighting of the adjacent luminaire is automatically activated to ensure that the illumination conditions of the work area meet the process safety requirements and dynamically optimize the compensation effect.

[0192] In one possible implementation, the dynamic selection of the adjacent luminaire is determined by the spatial coordinate position of the abnormal luminaire, the illumination coverage range of the adjacent luminaire of the abnormal luminaire, and the compensation ability of the abnormal luminaire.

[0193] Specifically, first, based on the installation coordinates of the overall luminaire and the coordinates of the abnormal luminaire, the distance between the two is calculated. If the distance is less than a set threshold, the luminaire is determined to be an adjacent luminaire.

[0194] Then, based on the light source parameters and installation height of the adjacent luminaire, the illumination contribution of the adjacent luminaire to the area where the faulty luminaire is located is analyzed. In one possible implementation, based on the Lambert cosine law and the inverse square law, the illumination contribution of the adjacent luminaire to the area where the abnormal luminaire is located is calculated:

[0195]

[0196] where E i→f is the illumination contribution of the i-th adjacent luminaire to the area where the abnormal luminaire is located, Φ i is the luminous flux of the i-th adjacent luminaire, θ if is the angle of incidence of light, and r if is the distance between the adjacent luminaire and the abnormal luminaire.

[0197] Finally, based on the dimming range and the current operating state of the luminaire, its compensation ability is evaluated. For example, if the remaining dimming ability of a luminaire exceeds the target illumination requirement, it is determined that the luminaire has compensation ability.

[0198] After determining that the adjacent luminaire has compensation ability, the compensation lighting measure is started, specifically:

[0199] The brightness of the adjacent luminaire is dynamically adjusted to gradually fill the illumination gap of the abnormal luminaire. For example, if the illumination gap of the abnormal luminaire is 20%, the brightness of the adjacent luminaire is proportionally increased to ensure that the overall illumination of the compensation area meets the standard.

[0200] By adjusting the illumination angle and brightness distribution of the adjacent luminaire, the illumination uniformity of the compensation area is balanced to avoid local over-brightness or over-darkness.

[0201] In another possible implementation, if the area illumination after compensation is still lower than the minimum illumination threshold required by the process, a secondary emergency response is triggered. The preset redundant luminaire, such as the channel emergency lighting, is automatically enabled, and the TSN protocol is directly controlled to ensure ±5% accuracy, ensuring the lighting needs of the key area.

[0202] In the method for dynamically optimizing and regulating the industrial intelligent lighting luminaire based on multi-modal environmental perception, S500 calculates the safety redundancy coefficient of the adjacent luminaire under the compensation lighting system, and feeds it back to the loss function of the dynamic demand model, driving the model to iterate and upgrade, thereby realizing the self-evolution closed-loop control of the industrial lighting system under abnormal working conditions, continuously improving the safety redundancy ability and process adaptability of the dynamic optimization and regulation strategy.

[0203] Please refer to Figure 6 which shows a flowchart of an example method S500 for dynamically optimizing and regulating an industrial intelligent lighting luminaire based on multi-modal environmental perception, the content of which includes:

[0204] S510: Calculation of safety redundancy coefficient.

[0205] Safety redundancy coefficient R security represents the illumination guarantee ability under extreme working conditions, which is calculated based on the actual illumination under the compensation lighting system of the adjacent luminaire and the process minimum illumination, and the calculation formula is:

[0206]

[0207] where R security is the safety redundancy coefficient, E compensated is the actual illumination under the compensation lighting strategy, and E required is the process minimum illumination.

[0208] S520: Safety redundancy coefficient feedback mechanism.

[0209] Through the counterproductive mechanism of the safety redundancy coefficient, the iteration upgrade of the dynamic demand model S200 is driven, and the continuous optimization of the global regulation strategy is realized.

[0210] In a possible implementation, the counterproductive mechanism of the safety redundancy coefficient is to feed back the safety redundancy coefficient R security to the dynamic demand model S200, add an input channel of the safety redundancy coefficient to the loss function of the dynamic demand model, and adjust the loss function:

[0211] L new = a W priority + b åmax(0, R security - R threshold );

[0212] Wherein, L is the total loss function, W priority is the priority weight, a, b are weight coefficients, R security is the safety redundancy coefficient, R threshold is the process safety redundancy threshold.

[0213] According to the counterproductive of R security , the priority matrix after counterproductive is output by FLC.

[0214] S100-S500 proposes an industrial lighting optimization method based on multi-modal perception and dynamic priority regulation. By deploying a sensor network integrating illumination intensity, thermal distribution and production process light demand tags, combined with lamps embedded with unique two-dimensional code tags and their digital portraits, an environment feature matrix and a lamp portrait matrix are constructed. Based on the collaborative generation of multi-layer perception and FLC, a 0-1 continuous value priority matrix is generated, and the control threshold is dynamically adjusted combined with the QoS parameter of the edge gateway: high priority area uses time sensitive network (TSN) protocol direct control ± 5% illumination accuracy, low priority area realizes batch PWM dimming through LSTM load prediction. Under abnormal conditions, trigger two-dimensional code maintenance work order and activate adjacent lamp compensation lighting, at the same time, through the compensation effect counterproductive safety redundancy coefficient, drive dynamic demand model iteration upgrade, form a "perception-decision-regulation-optimization" self-evolution system. This scheme significantly improves the environmental adaptability, energy efficiency and operation intelligent level of the industrial lighting system, and takes into account real-time response and low-carbon energy saving.

[0215] The application also provides a multi-modal environment perception based industrial intelligent lighting lamp dynamic optimization and regulation system, comprising:

[0216] An environment and luminaire information perception module, comprising: a multi-modal sensor network deployed in an industrial scene for real-time collection of environmental dynamic data; a luminaire identification unit associated with each luminaire, storing the unique identity and luminaire parameter information of the luminaire; a data processing unit connected to the multi-modal sensor network and the luminaire identification unit, for constructing an environment feature matrix and a luminaire digital portrait based on the collected environmental dynamic data and luminaire parameter information, respectively;

[0217] A dynamic demand modeling and optimization module connected to the environment and luminaire information perception module, comprising: a dynamic demand model unit for calculating the lighting demand priority of each luminaire based on the environment feature matrix and the luminaire digital portrait; a fuzzy logic optimization controller connected to the dynamic demand model unit for real-time optimization processing of the lighting demand priority to generate a luminaire priority matrix containing 0-1 continuous values;

[0218] An edge intelligent control module deployed in an edge gateway and connected to the dynamic demand modeling and optimization module, comprising: a lightweight digital twin for receiving the luminaire priority matrix; a QoS parameter adaptation unit connected to the lightweight digital twin for dynamically adjusting the application threshold of the luminaire priority matrix according to the current QoS parameter state; a differentiated control strategy execution unit connected to the QoS parameter adaptation unit for: using TSN protocol to directly control the luminaire identified as a high-priority area in the priority matrix after application threshold adjustment; implementing PWM dimming control based on a long short-term memory network load prediction model for the luminaire identified as a low-weight area in the priority matrix after application threshold adjustment;

[0219] A maintenance trigger and compensation module connected to the environment and luminaire information perception module, comprising: an anomaly monitoring unit for monitoring whether the sampling data of the luminaire is over-threshold or the position of the luminaire has deviated; a maintenance work order generation unit connected to the anomaly monitoring unit for generating a two-dimensional code maintenance work order based on the information of the luminaire identification unit when an anomaly is detected; a neighboring compensation control unit connected to the anomaly monitoring unit for activating the luminaire adjacent to the abnormal luminaire for compensation lighting to build a neighboring luminaire compensation lighting system when an anomaly is detected;

[0220] The closed loop optimization and model iteration module is connected with the maintenance trigger and compensation module and the dynamic demand modeling and optimization module, and comprises: an illumination evaluation unit, configured to acquire actual illumination data in the activation state of the adjacent luminaire compensation lighting system; a safety redundancy calculation unit connected with the illumination evaluation unit, configured to calculate a safety redundancy coefficient according to the actual illumination data and a preset process minimum illumination requirement; and a model feedback unit connected with the safety redundancy calculation unit and a dynamic demand model unit in the dynamic demand modeling and optimization module, configured to feed back the safety redundancy coefficient to the dynamic demand model unit to drive iterative upgrading of the dynamic demand model unit.

[0221] The application also provides a computer device, comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to realize the method for dynamically optimizing and regulating the luminaire based on the multi-modal environment perception of the industrial intelligent lighting according to any one of the technical solutions.

[0222] The basic principles of the application are described above in combination with specific embodiments, but it should be noted that the advantages, advantages, effects and the like mentioned in the application are only examples and are not limiting, and these advantages, advantages, effects and the like cannot be considered as the must-have of each embodiment of the application. In addition, the above specific details are only for the purpose of example and for the purpose of understanding, and are not limited to the above specific details. The application must be implemented using the above specific details.

[0223] The block diagrams of the devices, apparatuses and equipment involved in the application are only illustrative examples and are not intended to require or imply that the connection, arrangement and configuration shown in the block diagrams must be used. As those skilled in the art will recognize, these devices, apparatuses and equipment can be connected, arranged and configured in any way. Words such as "include", "contain", "have" and the like are open-ended words, which mean "include but not limited to", and can be used interchangeably. The words "or" and "and" used herein mean the word "and / or", and can be used interchangeably unless the context clearly indicates otherwise. The word "such as" used herein means the phrase "such as but not limited to", and can be used interchangeably.

[0224] It should also be noted that in the devices, apparatuses and methods of the application, each component or each step can be decomposed and / or recombined. These decompositions and / or recombination should be considered as equivalent solutions of the application.

[0225] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0226] The preferred embodiments of the application are described above in detail for the purposes of clarity and understanding. Such embodiments are provided by way of example and are not intended to limit the scope of the application, which is limited only by the scope of the appended claims.

Claims

1. A method for dynamic optimization and regulation of industrial intelligent lighting luminaires based on multi-modal environment perception, characterized in that, Comprise: Real-time collection of environmental dynamic data of industrial scene by deploying sensors, embedding unique QR code label for each lamp, the QR code label is associated with lamp parameters, and constructing environmental feature matrix and lamp digital portrait based on environmental dynamic data and lamp parameters respectively; Based on the environmental feature matrix and the lamp digital portrait, a dynamic demand model is established, the dynamic demand model outputs the priority weight of each lamp, and the priority matrix of 0-1 continuous value is generated through real-time optimization of fuzzy logic controller; Lightweight digital twin is deployed in edge gateway, receives priority matrix, adjusts threshold value of priority matrix according to QoS parameter state, and performs differentiated control; wherein, time sensitive network protocol is directly controlled for high priority area; batch PWM dimming based on LSTM load prediction is implemented for low weight area; Based on the sampling threshold value or position offset of the lamp, a two-dimensional code maintenance work order is triggered, and the adjacent lamp compensation lighting system is activated to build the adjacent lamp compensation lighting system; Based on the actual illuminance and the minimum illuminance of the process under the adjacent lamp compensation lighting system, the safety redundancy coefficient is calculated, and the iterative upgrade of the dynamic demand model is returned. 2.The method of claim 1, wherein, The environmental dynamic data collected by the sensor in real time comprises: The light intensity data is collected by high-precision photodiode sensor, the measurement range is 0-100,000 lux; The thermal distribution data is collected by 320*240 pixel infrared thermal imaging sensor, the frame rate is 30fps, and the kernel density estimation algorithm is used to extract the personnel density distribution and equipment temperature abnormal value; The RFID / two-dimensional code label deployed in the key equipment or process area stores the illumination threshold, color temperature interval and illumination uniformity requirement of the production process. 3.The method of claim 1, wherein, The QR code label of the lamp is fixed on the lamp shell by laser engraving or printing method, and the label content is stored in JSON / XML structured coding with the following parameters: Spatial coordinates: three-dimensional Cartesian coordinates of lamp installation position; Power consumption characteristics: rated power, maximum power, standby power consumption and dimming range; Light source parameters: light source type, color temperature range, color rendering index, luminous flux and light decay curve; Installation height: the height of the lamp from the ground. 4.The method of claim 1, wherein, The construction of the environmental feature matrix comprises the following steps: Timestamp synchronization: align the timestamps of sensor data and lamp state data, and use sliding window method to eliminate time sequence deviation; Spatial coordinate mapping: unify the lamp installation coordinates and environmental thermal map coordinates to Cartesian coordinate system; Feature splicing: splice the environmental feature matrix and the lamp portrait matrix according to the time dimension to form the joint feature matrix; Principal component analysis dimension reduction: remove redundant features and retain the first several principal components. 5.The method for dynamic optimization and regulation of industrial intelligent lighting lamps based on multi-modal environment perception according to claim 1, wherein, The dynamic demand model adopts a multi-layer perception structure, comprising: Input layer: receiving the fusion data of environmental feature matrix and lamp digital portrait; Hidden layer: using ReLU activation function; Output layer: generating priority weight; The formula of dynamic demand model is: W priority = σ(W2·ReLU(W1·X + b1) + b2); wherein W priority is a priority weight, X is an input feature vector, W1, W2 are weight matrices, b1, b2 are bias terms, and σ is a sigmoid function. The loss function is the weighted loss of mean square error combined with process safety constraint, and the formula is: Wherein, L is the total loss function, W priority is the priority weight, a, b is the weight coefficient, E min is the process minimum illumination requirement, E i is the predicted illumination value. 6.The method for dynamic optimization and regulation of industrial intelligent lighting luminaries based on multi-modal environment perception according to claim 1, wherein, The fuzzy logic controller comprises: Fuzzification module: convert environmental feature matrix and lamp digital portrait into fuzzy set; Knowledge base module: store fuzzy rule base and membership function; Fuzzy inference module: calculate the fuzzy set of output variables using Mamdani or Sugeno inference method; Defuzzification module: convert the fuzzy result into 0-1 continuous value priority weight using the barycenter method or maximum membership degree method. 7.The method for dynamic optimization and regulation of the luminaire based on multi-modal environment perception and industrial intelligent lighting according to claim 1, wherein, The formula for adjusting the priority matrix threshold based on QoS parameters is: wherein, w threshold is the adjusted threshold, w base = 0.8 is the initial threshold, γ is a sensitivity coefficient, and takes a value of 0.1-0.5; D latency is the instruction transmission delay from the edge gateway to the lamp, and P loss is the packet loss rate. High priority area, w i > w threshold , direct control with TSN protocol; Low priority region, w i ≤ w threshold , enter the LSTM-PWM batch dimming queue. 8.The method for dynamic optimization and regulation of the luminaire based on multi-modal environment perception and industrial intelligent lighting according to claim 7, wherein, The TSN direct control strategy of the high priority area includes: Based on the IEEE 802.1AS protocol, sub-microsecond level time synchronization is achieved, and bandwidth is reserved for high priority instructions through time-gated queues; Deploy a PID closed-loop controller to dynamically adjust the duty cycle to compensate for the illumination error, and adaptively adjust the controller parameters based on the light source aging curve; When detecting device abnormal spectrum mutation or sudden increase in personnel density, trigger an asynchronous interrupt signal to override the current dimming strategy, and send redundant instructions through multi-path routing; The LSTM-PWM batch dimming strategy of the low priority area includes: Input historical illumination, personnel density heat map and process stage identifier, predict future illumination demand at several time steps, and adapt to process changes through online incremental learning; Divide the time axis into fixed batches, process dimming instructions in batches, calculate the duty cycle and send multi-lamp instructions through broadcast; According to the network load, dynamically adjust the batch length, and use lamp grouping scheduling strategy to share the duty cycle parameters. 9.The method of claim 1, wherein, The adjacent lamp compensation lighting system includes: When the lamp meets any trigger condition of the sampling threshold trigger or the position offset trigger, the compensation lighting of the adjacent lamp is automatically activated; The dynamic selection of adjacent lamps is determined by the spatial coordinate position of the abnormal lamp, the illumination coverage range of the adjacent lamp of the abnormal lamp, and the compensation ability of the abnormal lamp; After determining that the adjacent lamp has compensation ability, the compensation lighting measure is started; The dynamic selection of adjacent lamps includes: Based on the installation coordinates of the overall lamp and the coordinates of the abnormal lamp, the distance between them is calculated, and if the distance is less than the set threshold, the lamp is determined as the adjacent lamp; Based on the light source parameters and installation height of the adjacent lamp, the light contribution of the adjacent lamp to the area where the faulty lamp is located is analyzed, and the calculation formula is: wherein E i→f is the illuminance contribution of the i-th neighboring luminaire to the area where the abnormal luminaire is located, Φ i is the luminous flux of the i-th neighboring luminaire, θ if is the angle of incidence of the light rays, r if is the distance between the neighboring luminaire and the abnormal luminaire; Based on the dimming range and current operating state of the lamp, the compensation ability of the lamp is evaluated; The compensation lighting measure includes: Dynamically adjust the brightness of the adjacent lamp to gradually fill the light gap of the abnormal lamp; By adjusting the irradiation angle and brightness distribution of the adjacent lamp, the uniformity of the light in the compensation area is balanced. 10.The method of claim 1, wherein the method further comprises: determining a current location of the mobile device; and determining a current location of the mobile device based on the current location of the mobile device and the location information of the at least one object. The safety redundancy coefficient back-pumping mechanism includes: Safety redundancy factor R security The formula for calculating R is: wherein R security is a safety redundancy factor, E compensated is a compensation for the actual illuminance under the lighting strategy, E required is the process minimum illuminance; R security The loss function fed back to the dynamic demand model is adjusted by the formula: L new = a - W priority + β - ∑max(0, R security - R threshold ); Wherein, L is the total loss function, W priority is the priority weight, a, b is the weight coefficient, R security is the safety redundancy coefficient, R threshold is the process safety redundancy threshold.

11. A dynamic optimization and regulation system for industrial intelligent lighting luminaires based on multi-modal environment perception, utilizing the method of any one of claims 1-10, characterized in that, It includes: An environment and lamp information perception module, including: a multi-modal sensor network deployed in an industrial scene for real-time collection of environmental dynamic data; a lamp identification unit associated with each lamp, storing the unique identity and lamp parameter information of the lamp; a data processing unit connected to the multi-modal sensor network and lamp identification unit, for constructing an environmental feature matrix and a lamp digital portrait based on the collected environmental dynamic data and lamp parameter information, respectively; The dynamic demand modeling and optimization module, connected with the environment and lamp information perception module, comprises: a dynamic demand model unit for calculating lighting demand priorities of each lamp based on the environment feature matrix and the lamp digital portrait; and a fuzzy logic optimization controller connected with the dynamic demand model unit for performing real-time optimization processing on the lighting demand priorities to generate a lamp priority matrix containing 0-1 continuous values. The edge intelligent control module, deployed at an edge gateway and connected with the dynamic demand modeling and optimization module, comprises: a lightweight digital twin for receiving the lamp priority matrix; a QoS parameter adaptation unit connected with the lightweight digital twin for dynamically adjusting application thresholds of the lamp priority matrix according to a current QoS parameter state; and a differentiated control strategy execution unit connected with the QoS parameter adaptation unit for: directly controlling lamps identified as high-priority areas in the priority matrix after application threshold adjustment by using a TSN protocol; and implementing PWM dimming control based on a long short-term memory network load prediction model for lamps identified as low-weight areas in the priority matrix after application threshold adjustment. The maintenance triggering and compensation module, connected with the environment and lamp information perception module, comprises: an abnormality monitoring unit for monitoring whether the sampling data of the lamp is over the threshold or the position of the lamp is deviated; a maintenance work order generation unit connected with the abnormality monitoring unit for generating a two-dimensional code maintenance work order based on the information of the lamp identification unit when an abnormality is monitored; and a neighboring compensation control unit connected with the abnormality monitoring unit for activating lamps neighboring the abnormal lamp to perform compensation lighting and constructing a neighboring lamp compensation lighting system when an abnormality is monitored. The closed-loop optimization and model iteration module, connected with the maintenance triggering and compensation module and the dynamic demand modeling and optimization module, comprises: an illuminance evaluation unit for acquiring actual illuminance data in the activated state of the neighboring lamp compensation lighting system; a safety redundancy calculation unit connected with the illuminance evaluation unit for calculating a safety redundancy coefficient according to the actual illuminance data and a preset process minimum illuminance requirement; and a model feedback unit connected with the safety redundancy calculation unit and the dynamic demand model unit in the dynamic demand modeling and optimization module for feeding back the safety redundancy coefficient to the dynamic demand model unit to drive iterative upgrading of the dynamic demand model unit.

12. A computer device, comprising: The computer device comprises a processor and a memory, and the memory stores at least one instruction, at least one program, a code set or an instruction set, which are loaded and executed by the processor to implement the industrial intelligent lighting lamp dynamic optimization and control method based on multi-modal environment perception according to any one of claims 1 to 10.

Images