Intelligent lighting system for mines

Abstract

The application provides a mine intelligent lighting system, comprising: a perception module for collecting multi-dimensional real-time data of a mine and outputting to an AI module; an AI strategy engine dynamically selects and combines lighting devices and their lighting intensity according to analysis results to form an optimal lighting control strategy; an interactive platform has a lighting network map and is connected with the lighting devices through bus communication, and is used for monitoring the lighting state of the system in real time and updating to the lighting network map. The application dynamically generates a lighting control strategy based on environmental perception data, personnel position and movement trajectory, predicted work route and the like through the AI strategy engine, and realizes global visual management by updating mine personnel distribution, work state and the like information in real time through the map, so as to meet the comprehensive requirements of mine lighting safety, energy saving and visual management and the like.

Description

Technical Field

[0001] This invention relates to the field of mine lighting technology, and specifically to an intelligent lighting system for mines. Background Technology

[0002] Currently, with the increasing demands for safety and energy conservation in the mining industry, underground lighting systems are gradually developing towards intelligence and automation. Traditional mining lighting systems mostly employ manual control or simple timed control methods, resulting in high energy consumption, slow response, and difficult maintenance. This is particularly true in the underground roadway environments of non-coal mines, where the large operating area and high personnel mobility lead to significant temporal and spatial differences in lighting needs. Traditional lighting solutions struggle to meet the requirements of intelligent management and energy conservation. While some existing lighting equipment incorporates sensor control technology, such as infrared or sound-and-light sensors, this often only provides intelligent control of isolated lighting devices, lacking systemic linkage and big data support. Consequently, it fails to meet the comprehensive requirements of modern mines for lighting safety, energy efficiency, visual management, and remote control.

[0003] Current systematic mine lighting solutions, such as patent document CN110418479A "An Intelligent Lighting System for Underground Mines," disclose that "the sensing unit is connected to the intelligent lighting unit via a CAN bus, and the intelligent lighting unit is connected to the background monitoring unit via Ethernet or GPRS; the sensing unit is used to receive sensing signals and send them to the intelligent lighting unit; the background monitoring unit is used to store the received parameter information of the intelligent lighting unit, locate personnel underground in the mine based on the parameter information of the intelligent lighting unit, and control the status of the lighting." This system uses multiple sensing units to monitor personnel positions in real time, and a data bus transmits the data from each sensing unit to the lighting unit and the background to achieve automatic control of the lighting switches. However, the control between the sensing units and the lighting unit in this system is point-to-point or area-based, failing to form a global view of the distribution of personnel and their working status throughout the mine. Furthermore, the lighting control strategy is simplistic and cannot meet the requirements for high efficiency, energy saving, and emission reduction in various complex mine application scenarios.

[0004] Therefore, there is an urgent need to develop a comprehensive intelligent lighting system for mines that can meet the requirements of safety, energy efficiency, and visual management. Summary of the Invention

[0005] This invention aims to overcome the shortcomings of existing technologies and provide a mine intelligent lighting system that enables forward-looking lighting planning, dynamic energy-saving control, and global safety visualization. The invention utilizes an AI strategy engine to dynamically generate corresponding lighting control strategies based on environmental perception data, personnel location and movement trajectories, and predicted operations. The system is linked to an explosion-proof platform to enable timely switching and alarm activation of the emergency lighting system. Furthermore, it achieves global visual management by updating information such as mine personnel distribution and operational status in real time via a map, thus meeting the comprehensive requirements of mine lighting safety, energy efficiency, and visual management.

[0006] The present invention achieves the above objectives through the following technical solutions: A mining intelligent lighting system includes several lighting devices deployed within a mine lighting area. It also includes a sensing module, an AI module, a control module, and an interactive platform. The sensing module collects multi-dimensional real-time data from the mine and outputs it to the AI ​​module. This multi-dimensional real-time data includes personnel positioning data, environmental parameters, and equipment operating status parameters. The AI ​​module has a built-in work route prediction model and an AI strategy engine. The work route prediction model performs model calculations on the multi-dimensional real-time data and outputs the route prediction results to the AI ​​strategy engine. The AI ​​strategy engine performs real-time analysis of the current work phase based on the route prediction results. Based on the analysis results, it dynamically selects and combines the lighting devices and their lighting intensity to form an optimal lighting control strategy. It automatically generates a lighting parameter sequence based on the optimal lighting control strategy and outputs it to the control module. The control module issues corresponding lighting control commands to the lighting devices based on the lighting parameter sequence. The interactive platform has a lighting network map and establishes a bus communication connection with the lighting devices to monitor the system lighting status in real time and update the lighting network map accordingly.

[0007] According to the intelligent mining lighting system provided by the present invention, the sensing module includes a personnel positioning unit, an environmental sensing unit, and an equipment status monitoring unit. The personnel positioning unit is used to collect personnel positioning data, which includes personnel location coordinates, movement speed, and direction information. The environmental sensing unit is used to collect environmental parameters, which include gas concentration, temperature, dust concentration, and light intensity. The equipment status monitoring unit is used to monitor the operating status parameters of the lighting device and its supporting equipment.

[0008] According to the intelligent lighting system for mines provided by the present invention, the personnel positioning unit includes a UWB tag worn by personnel and multiple UWB positioning base stations deployed in the mine lighting area. The UWB tag establishes a communication connection with the UWB positioning base stations and periodically sends a UWB pulse signal containing the personnel ID. The UWB positioning base station is used to receive the UWB pulse signal and record the signal arrival time and time difference, and calculate the personnel positioning data based on the arrival time and time difference.

[0009] According to the intelligent lighting system for mining provided by the present invention, the environmental sensing unit adopts a multi-parameter integrated sensor to synchronously collect parameters such as methane, carbon monoxide, oxygen concentration, temperature, and wind speed and output them to the control module.

[0010] The control module includes a safety monitoring unit, which is equipped with a safety threshold to determine whether the environmental parameters exceed the safety threshold. Based on the determination result, the unit automatically generates and sends emergency lighting control commands and safety alarm commands to the interactive platform.

[0011] According to the intelligent lighting system for mining provided by the present invention, the AI ​​module establishes a communication connection with each of the UWB positioning base stations to receive the personnel ID and personnel positioning data, and analyzes the job attributes of the personnel based on the personnel ID.

[0012] The job route prediction model adopts a two-stage prediction model, including a first-stage prediction model and a second-stage prediction model. The first-stage prediction model is trained with multiple LSTM network models according to the job attributes, and the second-stage prediction model adopts a GNN prediction model.

[0013] According to the intelligent lighting system for mining provided by the present invention, the AI ​​module further includes a model invocation unit, which makes an invocation decision on the two-stage prediction model by calculating the operational regularity corresponding to the job type attribute, including: Set a sliding time window, calculate the multi-dimensional regularity index corresponding to the job type attribute within the sliding time window, and output the job regularity after weighted fusion and normalization of the multi-dimensional regularity index.

[0014] An evaluation coefficient is set. When the regularity of the operation is greater than or equal to the evaluation coefficient, it is determined that the person is in a regular operation state, and the first-stage prediction model is invoked; otherwise, it is determined that the person is in an irregular operation state, and the second-stage prediction model is invoked.

[0015] According to the intelligent lighting system for mining provided by the present invention, the multi-dimensional regularity index includes path repetition rate, destination stability, and movement pattern consistency. The path repetition rate is obtained by calculating the similarity between the real-time trajectory of a person and its multiple historical trajectories within the sliding time window; the destination stability is calculated by statistically analyzing the frequency distribution of the person's visits to each of the UWB positioning base stations within the sliding time window; and the movement pattern consistency is calculated by the speed of the movement segment and the duration of the dwell point in the real-time trajectory of the person within the sliding time window.

[0016] According to the intelligent lighting system for mines provided by the present invention, the LSTM network model includes an input layer, a hidden layer, and an output layer. The input layer is used to generate a coordinate sequence tensor based on the real-time location coordinates of the personnel and their most recent n time step data. The hidden layer is used to perform model calculations based on the coordinate sequence tensor in time step order and output the final hidden state to the output layer. The output layer is used to perform a nonlinear transformation on the final hidden state through one or more fully connected layers and generate predicted coordinates containing the next m time steps.

[0017] According to the intelligent lighting system for mines provided by the present invention, the GNN prediction model includes an input encoding layer, a graph neural network layer, and a prediction output layer. The input encoding layer is used to input the deployment nodes of the UWB positioning base station and the real-time location coordinates of personnel, and generates node feature codes and context feature codes through a map matching algorithm. The graph neural network layer is used to perform information aggregation and node updates of the graph structure based on the node feature codes, and outputs the updated node matrix. The prediction output layer is used to extract the node where the personnel are currently located, and generate the probability distribution of the target node based on the context feature codes.

[0018] According to the intelligent mining lighting system provided by the present invention, the AI ​​module further includes a built-in rule base. The AI ​​strategy engine extracts the operation stage identification rules stored in the rule base, analyzes the current operation stage based on the predicted coordinates or probability distribution using the operation stage identification rules, calculates the lighting range, dynamically adjusts the lighting device set defined by the lighting range based on the multi-dimensional real-time data, and calculates the optimal brightness value for each lighting device in the lighting device set using a multi-objective optimization algorithm to generate the lighting parameter sequence as follows:

[0019] in, , Let be the decision function. For the identification of the j-th lighting device, The optimal brightness value for the j-th lighting device is... is the current stage of the operation; m is the number of lighting devices in the lighting device set.

[0020] Therefore, the present invention has the following beneficial effects: 1. This invention employs a two-stage prediction model and dynamically adjusts strategies based on job type and regularity. For jobs with high regularity, the LSTM network model accurately predicts personnel coordinates, facilitating follow-up lighting in the lighting system. For jobs with lower regularity, the GNN prediction model calculates and dynamically updates the optimal or most probable path to the target point in real time, allowing the lighting system to plan and activate lighting along the predicted route, ensuring no dark areas during personnel movement. Compared to traditional personnel movement prediction schemes, this invention improves the adaptability of the lighting system and maximizes energy efficiency.

[0021] 2. This invention uses an AI strategy engine to dynamically generate corresponding lighting control strategies based on environmental perception data, personnel location and movement trajectory, and predicted operations. This achieves a high degree of matching between lighting control and the actual required lighting brightness, range, and work process, thus meeting the requirements for energy saving and intelligence in mining lighting.

[0022] 3. This invention enables centralized management and predictive maintenance of lighting equipment by building an interactive platform that integrates real-time monitoring, health assessment and remote configuration, thereby significantly reducing operation and maintenance costs and difficulties.

[0023] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0024] Figure 1 This is a block diagram of a module in an embodiment of an intelligent lighting system for mining according to the present invention.

[0025] Figure 2 This is a schematic diagram of a lighting network map in an embodiment of a mining intelligent lighting system according to the present invention.

[0026] Figure 3 This is a schematic diagram of the alarm statistics interface in an embodiment of a mining intelligent lighting system according to the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0028] An embodiment of a mining intelligent lighting system See Figure 1 The present invention relates to a mine intelligent lighting system, comprising several lighting devices 10 deployed within a mine lighting area, and further comprising a sensing module 20, an AI module 30, a control module 40, and an interactive platform 50. The sensing module 20 is used to collect multi-dimensional real-time data from the mine and output it to the AI ​​module 30. The multi-dimensional real-time data includes personnel positioning data, environmental parameters, and equipment operating status parameters. The AI ​​module 30 has a built-in operation route prediction model and an AI strategy engine 34. The operation route prediction model performs model calculations on the multi-dimensional real-time data and outputs the route prediction results to the AI ​​strategy engine. The AI ​​strategy engine 34 is used to perform real-time analysis of the current operation stage based on the route prediction results, and dynamically select and combine the lighting device 10 and its lighting intensity according to the analysis results to form the best lighting control strategy. It also automatically generates a lighting parameter sequence based on the best lighting control strategy and outputs it to the control module 40. The control module 40 is used to issue corresponding lighting control commands to the lighting device 10 according to the lighting parameter sequence. The interactive platform 50 has a lighting network map and establishes a bus communication connection with the lighting device 10 to monitor the system lighting status in real time and update it in real time.

[0029] See Figure 2 Specifically, in this embodiment, the interactive platform 50 is provided with a system monitoring interface, which includes a global overview area, a lighting network map, and a device operation status dashboard.

[0030] The global overview area includes system health status, site information, and device access statistics. The health status is a comprehensive system score based on factors such as device online rate, alarm handling, and environmental parameters; the figure shows a health status of 75%. Site information includes site name and number of operating days, visually reflecting the long-term stability of the system; the figure shows the Ekou Iron Mine with 489 operating days. Device access statistics include the quantity and capacity of key equipment such as high-voltage / low-voltage distribution cabinets, used for centralized management of lighting device 10 and its supporting equipment.

[0031] The lighting network map is presented in a purple topology map, which includes tunnels and areas divided into Area 1, Area 2 and Area 3, as well as the current number of people and lighting status in each area, thus intuitively showing the number of people underground and the status of each lighting device 10.

[0032] See Figure 3Specifically, in this embodiment, the intelligent lighting system is linked with the explosion-proof platform to enable timely switching and alarm activation of the emergency lighting system. Alarm analysis and display are generated through the interactive platform 50. The diagram includes a real-time alarm information stream, alarm level statistics, and alarm type statistics. The real-time alarm information stream displays specific alarm events in a scrolling manner, including time, device, and specific information. The alarm levels, from high to low, include important alarms, alarms, warnings, and messages. The alarm type statistics include fault alarms, video alarms, SOE events, displacement alarms, system events, communication information, and limit exceedance alarms, which facilitates alarm information management and subsequent fault analysis.

[0033] In this embodiment, the sensing module 20 includes a personnel positioning unit 21, an environmental sensing unit 22, and an equipment status monitoring unit 23. The personnel positioning unit 21 is used to collect the personnel positioning data, which includes personnel location coordinates, movement speed, and direction information. The environmental sensing unit 22 is used to collect the environmental parameters, which include gas concentration, temperature, dust concentration, and light intensity. The equipment status monitoring unit 23 is used to monitor the operating status parameters of the lighting device 10 and its supporting equipment.

[0034] In this embodiment, the personnel positioning unit 21 includes a UWB tag worn by personnel and multiple UWB positioning base stations deployed in the mine lighting area. The UWB tag establishes a communication connection with the UWB positioning base stations and periodically sends a UWB pulse signal containing the personnel ID. The UWB positioning base station is used to receive the UWB pulse signal and record the signal arrival time and time difference, and calculate the personnel positioning data based on the arrival time and time difference.

[0035] Specifically, this embodiment uses the TDOA (Time Difference of Arrival) positioning method to calculate the location of personnel. This method compares the time difference of the same UWB pulse signal arriving at multiple UWB positioning base stations with known coordinates, and uses hyperboloid intersection to deduce the location of the UWB tag, i.e., the personnel, without measuring the absolute transmission time of the UWB pulse signal.

[0036] Taking four base stations with known coordinates as an example, the first... The coordinates of each base station are represented as follows:

[0037] , Equation (1) And let the coordinates of the UWB tag be: Equation (2) The measurement was taken with base station 0 as a reference. The arrival time difference of the UWB pulse signals of each base station is: , Equation (3) Taking base station 0 as a reference The distance difference between each base station and the UWB tag is: Equation (4) in, The speed of sound.

[0038] Substituting equations (1) and (2) into equation (4), we get: Equation (5) Will Substituting into equation (5) yields three sets of equations, each corresponding to a hyperboloid. The coordinates of the UWB tag are obtained by solving for the intersection of the three hyperboloids.

[0039] In this embodiment, the environmental sensing unit 22 uses a multi-parameter integrated sensor to simultaneously collect parameters such as methane, carbon monoxide, oxygen concentration, temperature, and wind speed and output them to the control module 40.

[0040] The control module 40 includes a safety monitoring unit, which is equipped with a safety threshold to determine whether the environmental parameters exceed the safety threshold. Based on the determination result, the control module 40 automatically generates and sends emergency lighting control commands and safety alarm commands to the interactive platform 50.

[0041] Specifically, the multi-parameter integrated sensor in this embodiment includes a detection chip, an MCU, and a communication unit. The detection chip integrates multiple dedicated sensitive elements, including an electrochemical gas sensor, a thermistor, and an infrared absorption unit, for simultaneously detecting multiple environmental parameters. The MCU performs temperature compensation, correction, and cross-interference correction on the raw detection signal to improve measurement accuracy. The communication unit sends environmental parameter data to the AI ​​module 30 and control module 40 of the lighting system via a digital interface. The safety threshold is set according to the alarm threshold for each environmental parameter during mine operation.

[0042] In this embodiment, the AI ​​module 30 establishes a communication connection with each of the UWB positioning base stations to receive the personnel ID and personnel positioning data, and analyzes the job attributes of the personnel based on the personnel ID.

[0043] The job route prediction model adopts a two-stage prediction model, including a first-stage prediction model and a second-stage prediction model. The first-stage prediction model is trained with multiple LSTM network models 32 according to the job attributes, and the second-stage prediction model adopts a GNN prediction model 33.

[0044] In this embodiment, the AI ​​module 30 further includes a model invocation unit 31. The model invocation unit 31 makes invocation decisions for the two-stage prediction model by calculating the job regularity corresponding to the job attribute, including: Set a sliding time window, calculate the multi-dimensional regularity index corresponding to the job type attribute within the sliding time window, and output the job regularity after weighted fusion and normalization of the multi-dimensional regularity index.

[0045] An evaluation coefficient is set. When the regularity of the operation is greater than or equal to the evaluation coefficient, it is determined that the person is in a regular operation state, and the first-stage prediction model is invoked; otherwise, it is determined that the person is in an irregular operation state, and the second-stage prediction model is invoked.

[0046] Specifically, in this embodiment, for fixed-route inspection workers who perform reciprocating, periodic inspections along fixed conveyor lines or tracks, the LSTM network model 32 can accurately predict personnel coordinates, facilitating the lighting system to provide follow-up illumination. For task-oriented technicians or maintenance personnel, whose task locations are random and paths are not fixed, the GNN prediction model 33 can calculate and dynamically update the optimal or most probable path to the target point in real time. This allows the lighting system to plan and activate lighting along the predicted route, ensuring no dark areas during personnel movement. This embodiment, by employing a two-stage prediction model and dynamically adjusting strategies based on job type and regularity, provides accurate lighting guidance for complex scenarios such as regular inspections and unexpected tasks, improving the adaptability of the lighting system and maximizing energy efficiency.

[0047] Specifically, the multi-dimensional regularity indicators described in this embodiment include path repetition rate, destination stability, and mobility pattern consistency.

[0048] Specifically, the path repetition rate S_path is obtained by calculating the similarity between the real-time trajectory of a person and its multiple historical trajectories within the sliding time window, taking the average value and performing normalization processing, where 0≤S_path≤1.

[0049] Specifically, the probability distribution of the access frequency of each UWB positioning base station is obtained by statistically analyzing the frequency of personnel accessing each base station within the sliding time window, and then normalizing the data. The distribution information entropy is then calculated.

[0050] in, For the first The probability of each base station access frequency. The destination stability S_dest is obtained by normalizing and inverting the distribution information entropy, where 0 ≤ S_dest ≤ 1.

[0051] Specifically, the standard deviation of the speed during the movement phase in the real-time trajectory of the person within the sliding time window is calculated and normalized to obtain speed consistency S_speed, 0≤S_speed≤1; the location and duration of dwell points in the real-time trajectory of the person within the sliding time window are analyzed, and the dwell regularity S_stop is obtained through a similar calculation process to the destination stability above, 0≤S_stop≤1; the movement pattern consistency S_move is obtained by weighting speed consistency S_speed and dwell regularity S_stop, 0≤S_move≤1.

[0052] Specifically, the calculation formula for the multi-dimensional regularity index described in this embodiment is as follows:

[0053] in, These are the weights for path repetition rate, destination stability, and movement pattern consistency, respectively. These weights can be set according to job attributes. For example, the path repetition rate S_path has a larger weight for inspectors, the destination stability S_dest has a larger weight for operators, and the movement pattern consistency S_move has a larger weight for maintenance workers.

[0054] In this embodiment, the LSTM network model 32 includes an input layer, a hidden layer, and an output layer. The input layer is used to generate a coordinate sequence tensor based on the real-time location coordinates of the person and their most recent n time step data. The hidden layer is used to perform model calculations based on the coordinate sequence tensor in time step order and output the final hidden state to the output layer. The output layer is used to perform a nonlinear transformation on the final hidden state through one or more fully connected layers and generate predicted coordinates containing the next m time steps.

[0055] Specifically, the construction of the LSTM network model 32 in this embodiment includes: extracting historical trajectory data of workers in jobs with a regularity greater than or equal to the evaluation coefficient, and processing them in batches; setting the sliding time window to T=20 time steps, taking the three-dimensional coordinates of consecutive T=20 time steps in the Nth batch to form a training sample, and generating a coordinate sequence tensor (N,20, 3).

[0056] The hidden layer employs a multi-layer structure, including at least a first hidden layer and a second hidden layer. The first hidden layer has 128 units, and the second hidden layer has 64 units. The output of the first hidden layer serves as the input to the second hidden layer for model calculation.

[0057] In this embodiment, the GNN prediction model 33 includes an input encoding layer, a graph neural network layer, and a prediction output layer. The input encoding layer is used to input the deployment nodes of the UWB positioning base station and the real-time location coordinates of the personnel, and generates node feature codes and context feature codes through a map matching algorithm. The graph neural network layer is used to perform information aggregation and node updates of the graph structure based on the node feature codes, and outputs the updated node matrix. The prediction output layer is used to extract the node where the personnel are currently located, and generate the probability distribution of the target node based on the context feature codes.

[0058] Specifically, the context features in this embodiment include target personnel context, task context, and time context. The target personnel context is the previous node and node sequence of the target personnel, the task context is the current task work order and task target node, and the time context is the shift time and shift handover time, etc.

[0059] Specifically, the graph structure of the GNN prediction model 33 in this embodiment includes nodes and node features, edges and edge weights. The nodes are the deployment nodes of the UWB positioning base station. The node features include the environmental alarm status, equipment operation status and area lighting status of the corresponding node. The edges are the paths between two nodes, and the edge weights are the distance or travel time between the two nodes.

[0060] Specifically, in this embodiment, the GNN model makes predictions by learning the spatial relationships and rules of the entire mine, as well as the interaction patterns between personnel behavior and the environment, and is used to handle scenarios such as sudden tasks, environmental changes, and travel to unfamiliar areas.

[0061] In this embodiment, the AI ​​module 30 also has a built-in rule base. The AI ​​strategy engine 34 extracts the operation stage identification rules stored in the rule base, and analyzes the current operation stage and calculates the lighting range based on the predicted coordinates or probability distribution using the operation stage identification rules. Based on the multi-dimensional real-time data, it dynamically adjusts the set of lighting devices defined by the lighting range, and calculates the optimal brightness value for each lighting device in the set of lighting devices using a multi-objective optimization algorithm to generate the lighting parameter sequence as follows:

[0062] in, , Let be the decision function. For the identification of the j-th lighting device, The optimal brightness value for the j-th lighting device is... is the current stage of the operation; m is the number of lighting devices in the lighting device set.

[0063] Specifically, the work phase identification rule described in this embodiment is as follows:

[0064] in, This represents the probability distribution of the predicted coordinate sequence or target node for personnel over the next m time steps. The current coordinates of the personnel; This represents the current movement speed of the personnel. Threshold recognition rules include:

[0065] in, , , These are the personnel speed threshold, the proximity threshold to the target node, and the rest / stay time determination; The shortest distance from the current location of the personnel to the predicted coordinates or the predicted target node; the set of critical operation nodes is the set of fixed work points that require fine lighting.

[0066] Specifically, in this embodiment, when a person is moving, the lighting range is a fan-shaped or rectangular area extending forward along the direction of movement, starting from the person's current position; when the person is working or stationary, the lighting range is a small circular or square area centered on the work node / rest node, to provide local high-brightness lighting.

[0067] Specifically, the dynamic adjustment of the lighting device set in this embodiment includes: determining the status of the lighting devices in the set based on standby operation status parameters; if a device is faulty or offline, it is automatically removed and the nearest lighting device is added to the set. The environmental conditions of the personnel are determined based on environmental parameters; if the environment is characterized by high dust or low visibility, the number of lighting devices is increased with the center of the lighting device set as the center; if the environment is characterized by increased methane concentration, emergency lighting devices are added to the set to forcibly illuminate the risk area and escape routes.

[0068] Specifically, the multi-objective optimization algorithm in this embodiment includes safety, energy-saving, and comfort objectives. The objective values ​​are calculated for each objective and dynamically weighted according to operational needs to generate the optimal brightness value. The safety objective ensures that the illuminance on the work surface meets the standard; the energy-saving objective minimizes total energy consumption; and the comfort objective smooths out the brightness difference between adjacent luminaires and avoids glare.

[0069] The above embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of protection of the present invention. Any non-substantial changes and substitutions made by those skilled in the art based on the present invention shall fall within the scope of protection claimed by the present invention.

Claims

1. A mine intelligent lighting system, comprising a plurality of lighting devices deployed within a mine lighting area, characterized in that, Also includes: The system comprises a perception module, an AI module, a control module, and an interaction platform. The perception module collects multi-dimensional real-time data from the mine and outputs it to the AI ​​module. This multi-dimensional real-time data includes personnel positioning data, environmental parameters, and equipment operating status parameters. The AI ​​module has a built-in model invocation unit, a work route prediction model, and an AI strategy engine. The model invocation unit makes invocation decisions based on the current work regularity of the personnel to the work route prediction model. The work route prediction model includes a first-stage prediction model and a second-stage prediction model. Based on the invocation decision, the first-stage prediction model or the second-stage prediction model is invoked to perform model calculations on the multi-dimensional real-time data. The system outputs route prediction results to the AI ​​strategy engine. The AI ​​strategy engine performs real-time analysis of the current operation's implementation phase based on the route prediction results, dynamically selects and combines the lighting devices and their lighting intensity according to the analysis results to form the optimal lighting control strategy, and automatically generates a lighting parameter sequence based on the optimal lighting control strategy and outputs it to the control module. The control module issues corresponding lighting control commands to the lighting devices based on the lighting parameter sequence. The interactive platform has a lighting network map and establishes a bus communication connection with the lighting devices to monitor the system lighting status in real time and update it to the lighting network map in real time. The AI ​​module establishes a communication connection with each UWB positioning base station to receive personnel IDs and personnel positioning data, and analyzes their job attributes based on the personnel IDs. The work route prediction model adopts a two-stage prediction model. The first-stage prediction model is trained with multiple LSTM network models according to the job attributes, and the second-stage prediction model adopts a GNN prediction model. The model invocation unit makes invocation decisions for the two-stage prediction model by calculating the regularity of the work type attribute, including: Set a sliding time window, calculate the multi-dimensional regularity index corresponding to the job type attribute within the sliding time window, and output the job regularity after weighted fusion and normalization of the multi-dimensional regularity index. An evaluation coefficient is set. When the regularity of the operation is greater than or equal to the evaluation coefficient, it is determined that the person is in a regular operation state, and the first-stage prediction model is invoked; otherwise, it is determined that the person is in an irregular operation state, and the second-stage prediction model is invoked.

2. The intelligent lighting system for mining according to claim 1, characterized in that, include: The sensing module includes a personnel positioning unit, an environmental sensing unit, and an equipment status monitoring unit. The personnel positioning unit is used to collect personnel positioning data, which includes personnel location coordinates, movement speed, and direction information. The environmental sensing unit is used to collect environmental parameters, which include gas concentration, temperature, dust concentration, and light intensity. The equipment status monitoring unit is used to monitor the operating status parameters of the lighting device and its supporting equipment.

3. The intelligent lighting system for mining according to claim 2, characterized in that: The personnel positioning unit includes a UWB tag worn by personnel and multiple UWB positioning base stations deployed in the mine lighting area. The UWB tag establishes a communication connection with the UWB positioning base station and periodically sends a UWB pulse signal containing the personnel ID. The UWB positioning base station is used to receive the UWB pulse signal and record the signal arrival time and time difference, and calculate the personnel positioning data based on the arrival time and time difference.

4. The intelligent lighting system for mining according to claim 2, characterized in that: The environmental sensing unit uses a multi-parameter integrated sensor to simultaneously collect parameters such as methane, carbon monoxide, oxygen concentration, temperature, and wind speed, and output them to the control module. The control module includes a safety monitoring unit, which is equipped with a safety threshold to determine whether the environmental parameters exceed the safety threshold. Based on the determination result, the unit automatically generates and sends emergency lighting control commands and safety alarm commands to the interactive platform.

5. The intelligent lighting system for mining according to claim 1, characterized in that: The multi-dimensional regularity indicators include path repetition rate, destination stability, and movement pattern consistency. The path repetition rate is obtained by calculating the similarity between the real-time trajectory of a person within the sliding time window and its multiple historical trajectories. The destination stability is calculated by statistically analyzing the frequency distribution of the person's visits to each of the UWB positioning base stations within the sliding time window. The movement pattern consistency is calculated by the speed of the movement segment and the duration of the dwell point in the real-time trajectory of the person within the sliding time window.

6. The intelligent lighting system for mining according to claim 1, characterized in that: The LSTM network model includes an input layer, a hidden layer, and an output layer. The input layer generates a coordinate sequence tensor based on the real-time location coordinates of the person and their most recent n time steps. The hidden layer performs model calculations based on the coordinate sequence tensor in time step order and outputs the final hidden state to the output layer. The output layer performs a nonlinear transformation on the final hidden state through one or more fully connected layers and generates predicted coordinates containing the next m time steps.

7. The intelligent lighting system for mining according to claim 1, characterized in that: The GNN prediction model includes an input encoding layer, a graph neural network layer, and a prediction output layer. The input encoding layer is used to input the deployment nodes of the UWB positioning base station and the real-time location coordinates of personnel, and generates node feature codes and context feature codes through a map matching algorithm. The graph neural network layer is used to aggregate information and update nodes of the graph structure based on the node feature encoding, and output the updated node matrix; the prediction output layer is used to extract the node where the person is currently located, and generate the probability distribution of the target node based on the context feature encoding.

8. The intelligent lighting system for mining according to claim 6 or 7, characterized in that: The AI ​​module also has a built-in rule base. The AI ​​strategy engine extracts the operation stage identification rules stored in the rule base, and analyzes the current operation stage and calculates the lighting range based on the predicted coordinates or probability distribution using the operation stage identification rules. Based on the multi-dimensional real-time data, it dynamically adjusts the set of lighting devices defined by the lighting range, and calculates the optimal brightness value for each lighting device in the set of lighting devices using a multi-objective optimization algorithm to generate the lighting parameter sequence as follows: in, , Let be the decision function. For the identification of the j-th lighting device, The optimal brightness value for the j-th lighting device is... is the current stage of the operation; m is the number of lighting devices in the lighting device set.

Images