Rail transit lighting control method and device based on data analysis

Abstract

The application discloses a rail transit lighting control method and device based on data analysis, and relates to the technical field of intelligent lighting. The method comprises the following steps: collecting multi-dimensional environment and operation data in real time through a sensor network deployed in a rail transit site, and constructing a lighting digital twin that is real-time mapped with the physical site and dynamically evolved; inputting the environment and operation data into a passenger flow prediction model constructed in advance to generate high-precision passenger flow prediction distribution of each area of the rail transit site and corresponding forward-looking lighting demand prediction results; using the lighting digital twin as a virtual evaluation environment, using a swarm intelligence optimization algorithm to solve a multi-objective lighting control problem, generating a globally optimal lighting control scheme, and performing twin calibration and model self-evolution. The problems of response lag, extensive control, lack of global optimization and model solidification in the prior art are solved.

Description

Technical Field

[0001] This invention relates to the field of intelligent lighting technology, and in particular to a method and device for controlling rail transit lighting based on data analysis. Background Technology

[0002] As core hubs of urban public transportation, rail transit stations are characterized by high passenger volume, complex spatial structures, and long operating hours. The station's lighting system not only affects passenger safety and comfort but is also a major component of the station's overall energy consumption. Traditional lighting control methods primarily rely on manual timed switching or simple dimming based on a single light sensor. These methods have the following significant drawbacks: 1) Response lag: Control decisions are usually based on the current or past state and cannot predict future changes in passenger flow, resulting in insufficient lighting preparation before the arrival of passenger flow peaks or energy waste during passenger flow troughs.

[0003] 2) Coarse control: The control strategy is often a "one-size-fits-all" global control, ignoring the differentiated lighting needs of different areas within the station (such as the concourse, platform, and transfer passage) at the same time, making it difficult to achieve refined energy saving.

[0004] 3) Lack of global optimization: Lighting control often only considers brightness or a single energy consumption index, and cannot comprehensively balance multiple objectives such as energy consumption, visual comfort (such as glare and uniformity), passenger flow guidance efficiency, and safety assurance, making it difficult to find the globally optimal control scheme.

[0005] 4) Model rigidity: The system lacks the ability to learn and evolve on its own. It cannot optimize its prediction and control models based on changes in operating models, seasonal changes, or the accumulation of new data, resulting in poor adaptability.

[0006] Therefore, how to provide an intelligent lighting control method that can predict ahead, control precisely, optimize multiple objectives, and has self-evolution capabilities is a technical problem that urgently needs to be solved in the current rail transit field. Summary of the Invention

[0007] This invention provides a data analysis-based method and device for controlling rail transit lighting, which solves the problems of response lag, coarse control, lack of global optimization, and model rigidity in existing technologies.

[0008] In a first aspect, embodiments of the present invention provide a rail transit lighting control method based on data analysis, the method comprising: By deploying sensor networks within rail transit stations, multi-dimensional environmental and operational data are collected in real time, and a lighting digital twin that is mapped and dynamically evolves in real time with the physical station is constructed. By inputting environmental and operational data into a pre-built passenger flow prediction model, a high-precision passenger flow prediction distribution and corresponding forward-looking lighting demand prediction results are generated for each area of ​​the rail transit station. Using a lighting digital twin as a virtual evaluation environment, a swarm intelligence optimization algorithm is used to solve a multi-objective lighting control problem, generate a globally optimal lighting control scheme, and perform twin calibration and model self-evolution.

[0009] The technical solution provided in this application has at least the following beneficial effects: By constructing a passenger flow prediction model based on the ST-GAT algorithm, the distribution of passenger flow in various areas during future periods can be accurately predicted, thereby transforming lighting control from "passive response" to "active prediction" and enabling forward-looking configuration of lighting needs. Utilizing a hybrid multi-objective particle swarm optimization algorithm that integrates chaotic initialization, Levy flight mechanism, and gray wolf cooperation, multi-objective (energy consumption, comfort, guidance, and safety) optimization problems are solved in the lighting digital twin, finding globally optimal or near-optimal lighting control schemes and avoiding the one-sidedness of single-objective optimization. A lighting digital twin integrating a ray tracing engine is constructed, achieving real-time, high-fidelity mapping between physical sites and virtual models. This twin not only reflects real-time status but also serves as a high-precision simulation environment, accurately evaluating the effects of different lighting schemes and providing a reliable basis for optimization decisions. By comparing real data after scheme execution with twin simulation data, Kalman filtering and other techniques are used to calibrate the twin, and new samples are continuously used to train the passenger flow prediction model, enabling the entire scheme to have online learning and self-optimization capabilities, adapting to environmental changes and maintaining high-efficiency operation over the long term.

[0010] In one alternative implementation, a sensor network deployed within rail transit stations is used to collect multi-dimensional environmental and operational data in real time, and to construct a lighting digital twin that is mapped and dynamically evolves in real time to the physical station, including: By deploying a sensor network within rail transit stations, multi-dimensional environmental and operational data are collected in real time, and the data is preprocessed to obtain preprocessed environmental and operational data. Using 3D simulation modeling technology, a 3D geometric model of a rail transit station is created, and preprocessed environmental and operational data is dynamically mapped to the 3D geometric model to obtain an initial lighting digital twin that is mapped in real time to the physical station. Integrate a ray tracing engine into the initial lighting digital twin to obtain a lighting digital twin that is mapped to the physical site in real time and evolves dynamically.

[0011] In one alternative implementation, environmental and operational data includes passenger flow data, ambient light data, train operation data, and time and event data.

[0012] In one alternative implementation, the passenger flow prediction model is built based on the ST-GAT algorithm, and the passenger flow prediction model includes an input layer, a graph attention layer, a gated loop layer, and an output layer.

[0013] In one alternative implementation, environmental and operational data are input into a pre-built passenger flow prediction model to generate high-precision passenger flow forecast distributions for each area of ​​the rail transit station and corresponding forward-looking lighting demand forecasts, including: The preprocessed environmental and operational data is converted into a station spatiotemporal map. The nodes in the station spatiotemporal map are the areas of the rail transit station. The node characteristics include the passenger flow data characteristics, ambient light data characteristics, train operation data characteristics, and time and event data characteristics of the corresponding area. The weight of the edge of the node is set according to the geographical distance. The input layer of the passenger flow prediction model receives the spatiotemporal graph of the station, converts it into a three-dimensional tensor, and inputs it into the first GAT layer within the graph attention layer of the passenger flow prediction model. Using several stacked GAT layers within the graph attention layer, the graph structure features of the site spatiotemporal graph are weighted and aggregated several times based on the attention coefficient of each GAT layer to obtain aggregated graph structure features. The aggregated graph structure feature sequence at several time steps is input into the gated loop unit of the gated loop layer of the passenger flow prediction model in chronological order to obtain the hidden state output sequence. The hidden state output sequence is input into the fully connected network of the output layer of the passenger flow prediction model, and the output function is called to generate a high-precision passenger flow prediction distribution for each area of ​​the rail transit station. Based on the preset passenger flow lighting mapping rule library, the high-precision passenger flow prediction distribution is mapped to the corresponding forward-looking lighting demand prediction results.

[0014] In one alternative implementation, the swarm intelligence optimization algorithm is a hybrid multi-objective particle swarm optimization algorithm that integrates chaotic initialization, the Levy flight mechanism, and the gray wolf cooperative idea.

[0015] In one alternative implementation, a lighting digital twin is used as a virtual evaluation environment. A swarm intelligence optimization algorithm is employed to solve a multi-objective lighting control problem, generating a globally optimal lighting control scheme. Twin calibration and model self-evolution are then performed, including: Define decision variables that match each area of ​​the rail transit station, and encode these decision variables as location vectors for a hybrid multi-objective particle swarm optimization algorithm. The decision variables are... A decision vector of dimension, where, The total number of areas with rail transit stations. This refers to the number of time steps for the prediction. Define at least one objective function for a hybrid multi-objective particle swarm optimization algorithm; Using a lighting digital twin as a virtual evaluation environment and high-precision passenger flow prediction distribution as the motion input of the virtual evaluation environment, a fitness evaluator with a hybrid multi-objective particle swarm optimization algorithm is constructed. Based on the forward-looking lighting demand prediction results and fitness evaluator, a hybrid multi-objective particle swarm optimization algorithm is used to solve the multi-objective lighting control problem and generate the Pareto optimal solution set; Based on the actual scenario of rail transit stations, the weighted TOPSIS method is used to weight each locally optimal lighting control scheme in the Pareto optimal solution set to obtain the relative proximity of each locally optimal lighting control scheme. The local optimal lighting control scheme with the highest relative proximity is output as the global optimal lighting control scheme, and the simulation data of the global optimal lighting control scheme in the virtual evaluation environment is saved. The system executes the globally optimal lighting control scheme, continuously collects real data after execution, and performs twin calibration and model self-evolution based on simulation data, real data, high-precision passenger flow prediction distribution, and forward-looking lighting demand prediction results.

[0016] In one alternative implementation, based on forward-looking lighting demand predictions and a fitness evaluator, a hybrid multi-objective particle swarm optimization algorithm is used to solve the multi-objective lighting control problem, generating a Pareto optimal solution set, including: Based on the forward-looking lighting demand forecast, the core attraction point of the initial particle swarm in the hybrid multi-objective particle swarm optimization algorithm is generated. Choosing the Logistic map as the chaos generator generates... A two-dimensional chaotic matrix is ​​obtained and linearly mapped to the interval [-1, 1] to obtain a normalized offset direction matrix, where, M For particle swarm scale, For the size of the decision variable; Based on the core attraction point and offset direction matrix, generate including M An initial swarm of particles, where each particle in the swarm corresponds to an alternative lighting control scheme. The alternative lighting control schemes corresponding to each particle in the initial particle swarm are loaded into the lighting digital twin of the fitness evaluator. The high-precision passenger flow prediction distribution is input into the lighting digital twin for simulation. After the simulation, the objective function is used to calculate the corresponding fitness vector based on the key data of the lighting digital twin. The particle with the best fitness vector is stored as the optimal particle in the external archive. Using the fast non-dominated sorting algorithm, the initial particle swarm and the best particles in the external archive are divided into several non-dominated layers, and the crowding distance of each particle in each non-dominated layer is calculated. By introducing the Levy flight mechanism and the gray wolf cooperative idea, the particle state of the initial or previous iteration particle swarm is updated to obtain the updated particle swarm. The fitness evaluator is used to obtain the fitness vector of each particle, and the particle with the best fitness vector is stored as the optimal particle in the external archive. Repeat the iterative update steps until the number of iterations reaches the iteration threshold, and output all the best particles in the external archive as the Pareto optimal solution set.

[0017] In one optional implementation, a globally optimal lighting control scheme is executed, real-world data is continuously collected after execution, and twin calibration and model self-evolution are performed based on simulated data, real-world data, high-precision passenger flow forecasts, and forward-looking lighting demand forecasts. This includes: The system executes the globally optimal lighting control scheme and continuously collects real-time data after execution. This real-time data includes real-time ambient light data and real-time passenger flow data collected by the sensor network. The error value between the real ambient light data and the simulated data is obtained. If the error value exceeds the error threshold, the Kalman filter algorithm is used to automatically adjust the twin parameters in the lighting digital twin to obtain an updated lighting digital twin. The simulated data is the simulated ambient light data obtained by simulating the globally optimal lighting control scheme in the lighting digital twin. Real data, simulated data, and high-precision passenger flow forecast distribution and forward-looking lighting demand forecast results are paired to form new training samples, which are then stored in the training sample library. Periodically select a number of training samples from the training sample library to continuously train the passenger flow prediction model and obtain an updated passenger flow prediction model.

[0018] Secondly, embodiments of the present invention provide a rail transit lighting control device based on data analysis, used to implement a rail transit lighting control method, the device comprising: The data acquisition and twin construction unit is used to collect multi-dimensional environmental and operational data in real time through a sensor network deployed in rail transit stations, and to construct a lighting digital twin that is mapped and dynamically evolves in real time with the physical station. The passenger flow and lighting demand forecasting unit is used to input environmental and operational data into a pre-built passenger flow forecasting model to generate high-precision passenger flow forecasting distribution and corresponding forward-looking lighting demand forecasting results for each area of ​​the rail transit station. The lighting control scheme optimization unit is used to use the lighting digital twin as a virtual evaluation environment, employ swarm intelligence optimization algorithms to solve multi-objective lighting control problems, generate globally optimal lighting control schemes, and perform twin calibration and model self-evolution.

[0019] A third aspect of this invention provides an electronic device, which includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by at least one processor, such that the at least one processor can perform the method proposed in the first aspect of the present invention.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in the first aspect of the present invention. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention; Figure 2 This is a flowchart illustrating the steps of a data analysis-based rail transit lighting control method provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of the functional units of a rail transit lighting control device based on data analysis, provided in an embodiment of the present invention. Detailed Implementation

[0022] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. 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 embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0023] The present invention will be further described below with reference to the accompanying drawings.

[0024] Reference Figure 1 , Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention.

[0025] like Figure 1As shown, the electronic device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0026] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0027] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and an electronic program for a rail transit lighting control device based on data analysis.

[0028] exist Figure 1 In the electronic device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the electronic device of the present invention can be set in the electronic device. The electronic device calls the electronic program of the rail transit lighting control device based on data analysis stored in the memory 1005 through the processor 1001 and executes the rail transit lighting control method based on data analysis provided in the embodiment of the present invention.

[0029] Reference Figure 2 The present invention provides a data analysis-based rail transit lighting control method, the method comprising: S201: Through a sensor network deployed in rail transit stations, multi-dimensional environmental and operational data are collected in real time, and a lighting digital twin that is mapped and dynamically evolves in real time with the physical station is constructed. S202: Input environmental and operational data into a pre-built passenger flow prediction model to generate high-precision passenger flow prediction distribution and corresponding forward-looking lighting demand prediction results for each area of ​​the rail transit station; S203: Using a lighting digital twin as a virtual evaluation environment, a swarm intelligence optimization algorithm is used to solve a multi-objective lighting control problem, generate a globally optimal lighting control scheme, and perform twin calibration and model self-evolution.

[0030] The technical solution provided in this application has at least the following beneficial effects: By constructing a spatio-temporal graph attention network The passenger flow prediction model based on the Network (ST-GAT) algorithm can accurately predict the passenger flow distribution in various areas during future periods, thereby transforming lighting control from "passive response" to "active prediction" and enabling forward-looking configuration of lighting needs. Utilizing a hybrid multi-objective particle swarm optimization algorithm that integrates chaotic initialization, Levy flight mechanism, and gray wolf cooperation, it solves multi-objective (energy consumption, comfort, guidance, and safety) optimization problems within a lighting digital twin, finding globally optimal or near-optimal lighting control schemes and avoiding the one-sidedness of single-objective optimization. A lighting digital twin integrating a ray tracing engine is constructed, achieving real-time, high-fidelity mapping between physical sites and virtual models. This twin not only reflects real-time status but also serves as a high-precision simulation environment, accurately evaluating the effects of different lighting schemes and providing a reliable basis for optimization decisions. By comparing real data after scheme execution with twin simulation data, Kalman filtering and other techniques are used to calibrate the twin, and new samples are continuously used to train the passenger flow prediction model, enabling the entire scheme to have online learning and self-optimization capabilities, adapting to environmental changes and maintaining high efficiency over the long term.

[0031] In one alternative implementation, a sensor network deployed within rail transit stations is used to collect multi-dimensional environmental and operational data in real time, and to construct a lighting digital twin that is mapped and dynamically evolves in real time to the physical station, including: S2011: Through a sensor network deployed in rail transit stations, multi-dimensional environmental and operational data are collected in real time, and the data is preprocessed to obtain preprocessed environmental and operational data; In this embodiment, the sensor network includes: Depth cameras or lidar with artificial intelligence analysis capabilities are deployed at intervals of 15-20 meters on the ceilings of key areas such as station halls, platforms, and transfer passages to collect passenger flow data. Digital light sensors with spectral analysis are installed near the center of the curtain wall, skylight, and lighting area to collect ambient light data; Secure and reliable data communication is established with the train's automatic control and signaling systems via an application programming interface (API) or a dedicated data gateway to collect train operation data. Connect to an NTP time server to obtain time data; Access the station's broadcasting system, fire alarm system, and equipment monitoring system to obtain event data; Environmental and operational data include passenger flow data, ambient light data, train operation data, and time and event data; It is worth noting that: Passenger flow data: Through edge computing devices or central servers, target detection and tracking algorithms (such as YOLO, DeepSort) are run to output the passenger flow density (people / square meter), average speed (m / second) and main flow vector of each area in real time, with a data collection frequency of no less than 1Hz; Ambient light data: Real-time collection of horizontal illuminance, vertical illuminance, and color temperature data for each area, with a sampling frequency of 0.5Hz; Train operation data includes, but is not limited to: real-time train location (accurate to meters), estimated arrival time (accurate to seconds), stop time, door opening / closing status, and train direction of travel; Time and event data: Precise UTC timestamps, including time data for year, month, day, week, hour, and minute; event data for emergency event trigger signals (such as fire alarms and emergency broadcasts); Data preprocessing: Data cleaning: Moving average filtering or median filtering algorithms are used to remove instantaneous noise and outliers from sensor data. Missing data is filled using linear interpolation or based on predicted values ​​from the same historical period. Data synchronization and alignment: Based on the UTC timestamp, all data streams from all sources are precisely aligned on the timeline to ensure that all data at the same moment constitutes a valid data snapshot; Data normalization: Normalize data of different dimensions (such as passenger flow density and light intensity) to the range [0, 1] to facilitate subsequent model processing; S2012: Using 3D simulation modeling technology, a 3D geometric model of a rail transit station is created, and the pre-processed environmental and operational data is dynamically mapped to the 3D geometric model to obtain an initial lighting digital twin that is mapped to the physical station in real time. In this embodiment, based on Building Information Modeling (BIM) data or 3D laser scanning data, a three-dimensional geometric model of the rail transit station is created using three-dimensional simulation modeling technology, including building structure, equipment layout (such as lighting fixtures, cameras, turnstiles) and furniture; Dynamic mapping: The collected real-time passenger flow data is rendered into the corresponding position of the three-dimensional geometric model in the form of a dynamic three-dimensional virtual human (or heat map); Real-time ambient light data is used as input for global illumination rendering of the 3D geometric model to simulate the impact of natural light on the site environment. Bind the real-time status of the lamps (on / off, brightness, color temperature) to the virtual lamp models in the 3D geometric model one-to-one; S2013: Integrate a ray tracing engine into the initial lighting digital twin to obtain a lighting digital twin that is mapped to the physical site in real time and evolves dynamically; In this embodiment, the ray tracing engine can accurately calculate the illuminance, brightness, and glare value of any point based on the luminaire parameters and the material of the space, thereby achieving physical-level lighting simulation. In this embodiment, to reduce computational pressure, the ray tracing engine uses a Long Short-Term Memory (LSTM) network as a proxy model to dynamically output the illuminance, brightness, and glare values ​​at any point based on the input luminaire parameters, space materials, and preprocessed environmental and operational data.

[0032] In one alternative implementation, the passenger flow prediction model is built based on the ST-GAT algorithm, and the passenger flow prediction model includes an input layer, a graph attention layer, a gated loop layer, and an output layer. In this embodiment, the core function of the input layer is data structuring and tensorization. It is responsible for transforming preprocessed multi-dimensional data from different data sources (passenger flow, ambient light, trains, and time events) into a standardized mathematical structure—a station spatiotemporal graph—that can be understood and processed by the neural network, according to preset rules, and ultimately represented as a three-dimensional tensor. This unifies the originally isolated and heterogeneous data (camera video, sensor readings, and train signals) into a standard mathematical framework, laying the foundation for subsequent deep learning model processing. By constructing the graph, it explicitly tells the model: "These regions are not isolated; they are connected and have influences." This is much more powerful than inputting all the data into a simple fully connected network or time series model because it introduces key spatial prior knowledge. By constructing the three-dimensional tensor, the dynamic evolution sequence of passenger flow over time is fully preserved, preparing for analysis in the time dimension.

[0033] The graph attention layer (GAT) is the core of the model's spatial intelligence. Its main function is to dynamically and adaptively learn the importance of each node and its neighbors, and aggregate neighbor information accordingly to update the feature representation of each node. The specific working mechanism includes: Attention coefficient calculation: For each node in the graph, the GAT layer calculates the "attention score" of its neighbors. This score is not fixed but dynamically calculated based on the features of the two nodes. For example, during peak hours, the passenger flow to the platform stairs has a greater impact on the concourse area; the GAT can automatically learn this relationship and assign it higher weight. Weighted aggregation: The attention score is normalized using the Softmax function, and then the features of the neighboring nodes are weighted and summed to obtain a new feature of the node that incorporates spatial context information. Multi-layer stacking: By stacking multiple GAT layers, the model can capture higher-order spatial intelligence. The first layer aggregates direct neighbors, while the second layer aggregates "neighbors of neighbors," thus perceiving a wider range of spatial influences. Capturing dynamic spatial dependencies is GAT's greatest strength, breaking the limitation of fixed edge weights in traditional graph neural networks. In rail transit, the influence between regions is highly dynamic (e.g., train arrivals instantly change passenger flow), and GAT can perfectly adapt to this change. After processing by GAT layers, the feature vector of each node no longer merely represents its own state but incorporates high-level semantic features that integrate information from its surrounding key regions. For example, the features of a passageway region now "know" whether two platforms it connects to will soon have trains arriving. Even if the graph's topology is not perfectly defined, GAT can automatically ignore unimportant connections and strengthen important ones by learning attention weights, thus having a better tolerance for small errors in the model definition. The gated recurrent layer is the core of the model's temporal intelligence. Its function is to learn and extract the long-term dependencies and dynamic evolution patterns of passenger flow data over time. It receives feature sequences from the GAT layer that have already incorporated spatial information (each time step is an updated graph), and then processes them chronologically. The gated recurrent unit, through its "gating mechanism" (update gate and reset gate), intelligently determines at each time step which historical information needs to be "forgotten," which new information needs to be "remembered," and which long-term memories need to be retained and passed on to the next time point. At each time step, the gated recurrent unit updates its "hidden state." This hidden state can be viewed as... It encodes all historical information up to the current moment; rail transit passenger flow exhibits obvious periodicity (daily, weekly) and trends (such as morning and evening peak hours); GRU can effectively capture these patterns spanning long time steps, for example, it can "remember" that it is currently the morning peak period and passenger flow will continue to increase; passenger flow changes are complex nonlinear processes, and the gated recurrent unit, as a type of recurrent neural network, is very good at modeling such dynamic systems, and can learn the transition rules of passenger flow from one state to another; the hidden state sequence output by the gated recurrent unit is the essence of the entire model after analysis in the spatiotemporal dimensions; providing a highly condensed and information-rich input for the final prediction task; The output layer's function is to decode and generate the final prediction result, mapping the hidden state output by the GRU layer, which contains all spatiotemporal information, to specific future passenger flow values ​​of interest; Specific implementation: Fully connected networks: Typically composed of one or more fully connected layers, responsible for learning the complex mapping relationship from hidden states to the prediction target; Predicting the future: The output layer is designed to predict passenger flow at multiple future time steps (e.g., 5 minutes in the future). For each future time step, it outputs a vector containing the passenger flow characteristics (e.g., passenger density, speed) that need to be predicted for each area at that moment; transforming the model's abstract "understanding" into concrete, quantifiable numbers. For example, it is no longer "platform passenger flow will increase," but "in 3 minutes, the passenger flow density in the Line 1 platform area will reach 2.1 people / square meter"; This high-precision, distributed passenger flow prediction result is the direct and core basis for subsequent lighting control decisions; By predicting the future, the entire lighting control scheme is given the ability to "foresee the future" and act in advance, which is fundamental to achieving "proactive lighting" and is crucial for improving passenger experience and operational efficiency.

[0034] In one alternative implementation, environmental and operational data are input into a pre-built passenger flow prediction model to generate high-precision passenger flow forecast distributions for each area of ​​the rail transit station and corresponding forward-looking lighting demand forecasts, including: S2021: Convert the preprocessed environmental and operational data into a station spatiotemporal map. The nodes in the station spatiotemporal map are the areas of the rail transit station. The node characteristics include the passenger flow data characteristics, ambient light data characteristics, train operation data characteristics, and time and event data characteristics of the corresponding area. The weight of the edge of the node is set according to the geographical distance. In this embodiment, nodes are defined as follows: rail transit stations (such as station halls, platforms, transfer passages, and stairwells) are divided into... N Each region is logically or physically independent, and each region is abstracted as a node in the graph; Edges: Based on the station topology and passenger flow logic, define the connection relationships between nodes. If region 1 and region 2 are physically adjacent, or there is a significant passenger flow direction (such as direct access from the stairwell to the platform), then at the node... v 1 and v Establish an edge between 2 e 12 Edge weight w 12 Initialize to: The formula for distance weight is:

[0035] In the formula, For nodes v 1 and v 2. Physical distance between the center points; S2022: The input layer of the passenger flow prediction model receives the spatiotemporal graph of the station, converts it into a three-dimensional tensor, and inputs it into the first GAT layer in the graph attention layer of the passenger flow prediction model. In this embodiment, the continuous data stream of the site's spatiotemporal graph is divided into time windows of fixed length, and a historical observation window is defined. (e.g., a snapshot every minute over the past 15 minutes) and a prediction window (e.g., the next 5 minutes); Input to the model: a three-dimensional tensor ,in, F The dimension of node features; representing the past A dynamic snapshot sequence of the entire site map within each time step; S2023: Using several stacked GAT layers within the graph attention layer, the graph structure features of the site spatiotemporal graph are weighted and aggregated several times according to the attention coefficient of each GAT layer to obtain aggregated graph structure features. In this embodiment, for each time step of the graph snapshot in the input sequence, the graph attention layer processes it. For each node, it calculates the importance of its neighboring nodes (attention coefficient) and uses the Softmax function to normalize the attention score to obtain the final attention coefficient. By stacking multiple GAT layers, higher-order spatial dependencies can be captured; S2024: Input the aggregated graph structure feature sequence of several time steps into the gated loop unit of the gated loop layer of the passenger flow prediction model in chronological order to obtain the hidden state output sequence. In this embodiment, the gated loop unit can learn the long-term dependencies and dynamic evolution patterns in the time series. At each time step, the gated loop unit updates its hidden state, which encodes all historical information up to the current time. S2025: Input the hidden state output sequence into the fully connected network of the output layer of the passenger flow prediction model, call the output function, and generate a high-precision passenger flow prediction distribution for each area of ​​the rail transit station. In this embodiment, the output layer is designed to predict the future. The output of the passenger flow at each time step is also a three-dimensional tensor. ,in, C The dimensions of passenger flow characteristics to be predicted (typically including at least passenger flow density and speed). The prediction formula is:

[0036] In the formula, For the future k At a time step, in the region j The predicted passenger flow density is used to construct a high-precision passenger flow prediction distribution for each area of ​​the rail transit station. This is a fully connected network function in the output layer specifically responsible for predicting density; As a window for historical observation The hidden state output sequence; l For time step indication; S2026: Based on the preset passenger flow lighting mapping rule library, map the high-precision passenger flow prediction distribution to the corresponding forward-looking lighting demand prediction results; It is worth noting that the passenger flow lighting mapping rule base is a rule base driven by expert experience, lighting design standards, and passenger behavior analysis. This rule base employs fuzzy logic, decision trees, or simple threshold functions. For example, IF >1.5 people / m² THEN = "HIGH"; IF 0.5< <= 1.5 people / m² THEN = "MEDIUM"; IF <= 0.5 people / m² THEN = "LOW"; Qualitative demand levels ("HIGH", "MEDIUM", "LOW") are mapped to quantitative target illuminance values ​​or luminance percentages. For example, "HIGH" → target illuminance. E target = 300 Lux (or brightness L_target = 90%); "MEDIUM" → Target Illuminance E target = 200 Lux (or brightness L_target = 60%); "LOW" → Target Illuminance E target = 100 Lux (or brightness L_target = 30%); "PRE-EMPTIVE_HIGH" → Adds a leading coefficient to "HIGH", such as L target = 95%; Traverse all regions j ( j = 1, ..., N ) and all future time steps k ( k = 1, ..., Apply the mapping rules described above; Ultimately, a forward-looking lighting demand matrix with the same dimensions as the passenger flow forecast is generated. Each element Represents the region j In the future The recommended brightness value at any given time will serve as the core attraction and target basis for swarm intelligence optimization.

[0037] In one alternative implementation, the swarm intelligence optimization algorithm is a hybrid multi-objective particle swarm optimization algorithm that integrates chaotic initialization, the Levy flight mechanism, and the gray wolf cooperative idea.

[0038] In one alternative implementation, a lighting digital twin is used as a virtual evaluation environment. A swarm intelligence optimization algorithm is employed to solve a multi-objective lighting control problem, generating a globally optimal lighting control scheme. Twin calibration and model self-evolution are then performed, including: S2031: Define decision variables that match each area of ​​the rail transit station, and encode these decision variables as location vectors for a hybrid multi-objective particle swarm optimization algorithm. The decision variables are... A decision vector of dimension, where, The total number of areas with rail transit stations. This refers to the number of time steps for the prediction. In this embodiment, the site is divided into Each region, predicting the future. For each time step, the decision vector is... A dimensional vector representing the target brightness of each region at each future time point; the decision vector. X = [ L 11 , L 12 ,..., L 1j , L 21 , ..., L 2j , ..., L i1 , ..., L ij , ..., L NK ],in, L ij Indicates the area i In the future j The target brightness at each time step, with a value range of [ L min , L max (e.g., [0, 100%]), where, n For regional indication; S2032: Define at least one objective function for a hybrid multi-objective particle swarm optimization algorithm; In this embodiment, the objective function includes The formula for minimizing total energy consumption is:

[0039] In the formula, This represents the total energy consumption value. For the region i Total power of all lights at full power; It is the duration (in seconds) of each time step; The formula for minimizing visual discomfort is:

[0040] In the formula, This represents the visual discomfort value. In lighting control scheme X The average uniform glare value of the entire station was simulated by the lighting digital twin. In lighting control scheme X The ratio of the minimum illuminance to the average illuminance of the entire station; w These are weighting coefficients used to balance glare and uniformity; The formula for maximizing bootstrapping efficiency is:

[0041] In the formula, To guide efficiency values; In lighting control scheme X Passengers p Average path length from entrance to destination (e.g., car door); p Indicate the amount for passengers; The total number of passengers is represented by the negative sign because optimization algorithms typically minimize the objective, and maximizing guidance efficiency is achieved by minimizing the negative value of the path length. The formula for maximizing the safety factor is:

[0042] In the formula, This is the safety factor value; In lighting control scheme X Below, area i At time step j The minimum illuminance; by maximizing the negative value of the minimum illuminance, to ensure that there are no blind spots in the lighting throughout the station; S2033: Using a lighting digital twin as a virtual evaluation environment and high-precision passenger flow prediction distribution as the motion input of the virtual evaluation environment, a fitness evaluator based on a hybrid multi-objective particle swarm optimization algorithm is constructed. S2034: Based on the prospective lighting demand prediction results and fitness evaluator, a hybrid multi-objective particle swarm optimization algorithm is used to solve the multi-objective lighting control problem and generate the Pareto optimal solution set. S2035: Based on the actual scenario of rail transit stations, the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) is used to weight each locally optimal lighting control scheme in the Pareto optimal solution set to obtain the relative similarity of each locally optimal lighting control scheme. In this embodiment, different target weighting and reorganization are set for each preset actual scenario. For example, when the actual scenario is a peak period, the weight of the guiding efficiency safety coefficient in the objective function is set to the highest, for example [ w 1 = 0.1, w 2 = 0.1, w 3 = 0.4, w 4 = 0.4]; Based on the actual scenarios of rail transit stations, obtain the corresponding target weighted reorganization. W = [ w 1, w 2, w 3, w 4], and the sum is 1, extract the fitness vector corresponding to each locally optimal lighting control scheme in the Pareto optimal solution set. And reorganize according to the target weighted weight [ w 1, w 2, w 3, w 4] Regarding the fitness vector After weighting, the relative proximity of each locally optimal lighting control scheme is obtained, using the following formula:

[0043] In the formula, Relative proximity; S2036: Output the local optimal lighting control scheme with the highest relative proximity as the global optimal lighting control scheme, and save the simulation data of the global optimal lighting control scheme in the virtual evaluation environment; S2037: Execute the globally optimal lighting control scheme, continuously collect real data after execution, and perform twin calibration and model self-evolution based on simulation data, real data, high-precision passenger flow prediction distribution, and forward-looking lighting demand prediction results.

[0044] In one alternative implementation, based on forward-looking lighting demand predictions and a fitness evaluator, a hybrid multi-objective particle swarm optimization algorithm is used to solve the multi-objective lighting control problem, generating a Pareto optimal solution set, including: S20341: Based on the forward-looking lighting demand prediction results, generate the core attraction point of the initial particle swarm in the hybrid multi-objective particle swarm optimization algorithm; In this embodiment, the forward-looking lighting demand forecast result is: R = { r d | d = 1, 2, ..., D},in, r d In the firstd Recommended brightness values ​​in dimensions For the size of the decision variable; S20342: Select the Logistic map as the chaos generator to generate... A two-dimensional chaotic matrix is ​​obtained and linearly mapped to the interval [-1, 1] to obtain a normalized offset direction matrix, where, M For particle swarm scale, For the size of the decision variable; The formula is:

[0045] In the formula, For the first +1、 One chaotic variable; The stability coefficient controls the chaotic behavior when When = 4, it is in a completely chaotic state; this sequence has ergodicity and randomness, which can ensure that the initial particle swarm is uniformly distributed in the solution space, avoid getting trapped in local optima, and is superior to traditional random initialization; It is a chaotic indicator; Iterative Logistic Mapping Next, generate a one-dimensional chaotic sequence. Z seq ={ , , ..., }; Reshape the one-dimensional sequence into a The two-dimensional chaotic matrix is ​​linearly mapped to the interval [-1, 1] to obtain the standardized offset direction matrix, as shown in the formula:

[0046] In the formula, For reshaping functions; It is a two-dimensional chaotic matrix; This is the offset direction matrix; S20343: Generate, based on the core attraction point and offset direction matrix, including... M An initial swarm of particles, where each particle in the swarm corresponds to an alternative lighting control scheme. In this embodiment, for the initial particle swarm The initial position of each particle and each dimension is calculated by adding a chaotic offset to the value of the core attraction point in that dimension. The magnitude of the offset is determined by the maximum offset radius and the chaos value, as shown in the formula:

[0047] In the formula, For the first The particle in the first d The initial position of the dimension; The maximum offset radius; Offset direction matrix The Middle The particle in the first d The chaotic value of the dimension; Initial velocity: the initial velocity vector of each particle. You can [- , Randomly generated within the range, where It is the maximum permissible speed, for example = ; S20344: Load the candidate lighting control schemes corresponding to each particle in the initial particle swarm into the lighting digital twin of the fitness evaluator. Input the high-precision passenger flow prediction distribution into the lighting digital twin for simulation. After the simulation, use the objective function to calculate the corresponding fitness vector based on the key data of the lighting digital twin. The particle with the optimal fitness vector is stored as the optimal particle in an external archive. S20345: Using the fast non-dominated sorting algorithm, the initial particle swarm and the best particles in the external archive are divided into several non-dominated layers. F 1, F 2, ...; where, F The solution in 1 is Pareto optimal and is not dominated by any other solution. The crowding distance of each particle in each non-dominated layer is calculated. This value measures the density of the solutions around it. The larger the value, the sparser the region where the solution is located and the better the diversity. S20346: Introducing the Levy flight mechanism and the gray wolf cooperative idea, the particle state of the initial or previous iteration particle swarm is updated to obtain the updated particle swarm. The fitness evaluator is used to obtain the fitness vector of each particle, and the particle with the best fitness vector is stored as the optimal particle in the external archive. In this embodiment, a random flight probability between [0,1] is generated for each particle. ,if Flight probability threshold (like If (e.g., = 0.1), then perform the particle state update for Levy flight, using the following formula:

[0048] In the formula, Number of iterations t+ 1 ,t The The position of each particle is updated; for Levy Distributed random numbers; b for Levy Step length, and b ∈[1,2]; t This represents the current iteration number; For flight control step size; Otherwise, implement the particle state update improved by the gray wolf cooperative approach, starting from the current optimal non-dominated layer. F In step 1, based on the congestion level, three optimal solutions are identified, defined as Alpha ( α Beta β ) and Delta ( δ The wolf positions, expanding the historical best position gbest of particles in the standard PSO update to three wolves, are given by the formula:

[0049] In the formula, Number of iterations t+ 1 ,t The The position of each particle is updated; This is the collaboration coefficient, which can be set to the same or different values ​​to adjust the influence weight. A random number in the range [0,1]. Alpha in the initial or previous iteration of the particle swarm ( α Beta β ) and Delta ( δ The position of the particle corresponding to the wolf; Number of iterations t+ 1 ,t The The velocity of each particle; The convergence factor; For the first The historical optimal position of each particle;

[0050] In the formula, This represents the maximum number of iterations. , To adjust the parameters; It is the hyperbolic tangent function; This represents the maximum value of the convergence factor. This represents the minimum convergence factor. S20347: Repeat the iterative update steps until the number of iterations reaches the iteration threshold, and output all the best particles in the external archive as the Pareto optimal solution set.

[0051] In one optional implementation, a globally optimal lighting control scheme is executed, real-world data is continuously collected after execution, and twin calibration and model self-evolution are performed based on simulated data, real-world data, high-precision passenger flow forecasts, and forward-looking lighting demand forecasts. This includes: S20371: Execute the globally optimal lighting control scheme and continuously collect real data after execution. The real data includes real ambient light data and real passenger flow data collected by the sensor network after execution. S20372: Obtain the error value between the real ambient light data and the simulated data. If the error value exceeds the error threshold, use the Kalman filter algorithm to automatically adjust the twin parameters in the lighting digital twin to obtain an updated lighting digital twin. The simulated data is the simulated ambient light data obtained by simulating the globally optimal lighting control scheme in the lighting digital twin. S20373: Pair real data, simulated data, and high-precision passenger flow forecast distribution and forward-looking lighting demand forecast results to form new training samples and store them in the training sample library. S20374: Periodically select a number of training samples from the training sample library to continuously train the passenger flow prediction model and obtain an updated passenger flow prediction model.

[0052] This invention also provides a data analysis-based rail transit lighting control device, referring to... Figure 3 The device may include the following units: The data acquisition and twin construction unit 301 is used to collect multi-dimensional environmental and operational data in real time through a sensor network deployed in rail transit stations, and to construct a lighting digital twin that is mapped and dynamically evolves in real time with the physical station. Passenger flow and lighting demand prediction unit 302 is used to input environmental and operational data into a pre-built passenger flow prediction model to generate high-precision passenger flow prediction distribution and corresponding forward-looking lighting demand prediction results for each area of ​​the rail transit station. The lighting control scheme optimization unit 303 is used to use the lighting digital twin as a virtual evaluation environment, use swarm intelligence optimization algorithm to solve multi-objective lighting control problems, generate the globally optimal lighting control scheme, and perform twin calibration and model self-evolution.

[0053] Based on the same inventive concept, another embodiment of the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. Memory, used to store computer programs; The processor, when executing the program stored in the memory, implements the data analysis-based rail transit lighting control method of the present invention.

[0054] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EI) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus. The communication interface is used for communication between the aforementioned terminal and other devices. The memory can include Random Access Memory (RAM), or non-volatile memory, such as at least one disk storage device. Optionally, the memory can also be at least one storage device located remotely from the aforementioned processor.

[0055] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0056] Furthermore, to achieve the above objectives, embodiments of the present invention also propose a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the data analysis-based rail transit lighting control method of the embodiments of the present invention.

[0057] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable hardware devices (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0058] The embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (apparatus), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0059] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0060] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0061] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. "And / or" indicates that either one or both can be chosen. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes the element.

[0062] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A data analysis-based method for controlling rail transit lighting, characterized in that, The method includes: By deploying sensor networks within rail transit stations, multi-dimensional environmental and operational data are collected in real time, and a lighting digital twin that is mapped and dynamically evolves in real time with the physical station is constructed. By inputting environmental and operational data into a pre-built passenger flow prediction model, a high-precision passenger flow prediction distribution and corresponding forward-looking lighting demand prediction results are generated for each area of ​​the rail transit station. Using a lighting digital twin as a virtual evaluation environment, a swarm intelligence optimization algorithm is used to solve a multi-objective lighting control problem, generate a globally optimal lighting control scheme, and perform twin calibration and model self-evolution.

2. The rail transit lighting control method based on data analysis according to claim 1, characterized in that, By deploying sensor networks within rail transit stations, multi-dimensional environmental and operational data are collected in real time, and a lighting digital twin that is mapped and dynamically evolves in real time with the physical station is constructed, including: By deploying a sensor network within rail transit stations, multi-dimensional environmental and operational data are collected in real time, and the data is preprocessed to obtain preprocessed environmental and operational data. Using 3D simulation modeling technology, a 3D geometric model of a rail transit station is created, and preprocessed environmental and operational data is dynamically mapped to the 3D geometric model to obtain an initial lighting digital twin that is mapped in real time to the physical station. Integrate a ray tracing engine into the initial lighting digital twin to obtain a lighting digital twin that is mapped to the physical site in real time and evolves dynamically.

3. The rail transit lighting control method based on data analysis according to claim 2, characterized in that, The environmental and operational data includes passenger flow data, ambient light data, train operation data, and time and event data.

4. The rail transit lighting control method based on data analysis according to claim 3, characterized in that, The passenger flow prediction model is constructed based on the ST-GAT algorithm and includes an input layer, a graph attention layer, a gated loop layer, and an output layer.

5. The rail transit lighting control method based on data analysis according to claim 4, characterized in that, Environmental and operational data are input into a pre-built passenger flow prediction model to generate high-precision passenger flow forecasts for each area of ​​the rail transit station and corresponding forward-looking lighting demand forecasts, including: The preprocessed environmental and operational data is converted into a station spatiotemporal map. The nodes in the station spatiotemporal map are the areas of rail transit stations. The node features of the nodes include passenger flow data features, ambient light data features, train operation data features, and time and event data features of the corresponding area. The weights of the edges of the nodes are set according to geographical distance. The input layer of the passenger flow prediction model receives the spatiotemporal graph of the station, converts it into a three-dimensional tensor, and inputs it into the first GAT layer within the graph attention layer of the passenger flow prediction model. Using several stacked GAT layers within the graph attention layer, the graph structure features of the site spatiotemporal graph are weighted and aggregated several times based on the attention coefficient of each GAT layer to obtain aggregated graph structure features. The aggregated graph structure feature sequence at several time steps is input into the gated loop unit of the gated loop layer of the passenger flow prediction model in chronological order to obtain the hidden state output sequence. The hidden state output sequence is input into the fully connected network of the output layer of the passenger flow prediction model, and the output function is called to generate a high-precision passenger flow prediction distribution for each area of ​​the rail transit station. Based on the preset passenger flow lighting mapping rule library, the high-precision passenger flow prediction distribution is mapped to the corresponding forward-looking lighting demand prediction results.

6. The rail transit lighting control method based on data analysis according to claim 5, characterized in that, The swarm intelligence optimization algorithm is a hybrid multi-objective particle swarm optimization algorithm that integrates chaotic initialization, Levy flight mechanism and gray wolf cooperative idea.

7. The rail transit lighting control method based on data analysis according to claim 6, characterized in that, Using a lighting digital twin as a virtual evaluation environment, a swarm intelligence optimization algorithm is employed to solve a multi-objective lighting control problem, generating a globally optimal lighting control scheme. Twin calibration and model self-evolution are then performed, including: Define decision variables that match each area of ​​the rail transit station, and encode these decision variables as location vectors for a hybrid multi-objective particle swarm optimization algorithm. The decision variables are... A decision vector of dimension, where, The total number of areas with rail transit stations. This refers to the number of time steps for the prediction. Define at least one objective function for a hybrid multi-objective particle swarm optimization algorithm; Using a lighting digital twin as a virtual evaluation environment and high-precision passenger flow prediction distribution as the motion input of the virtual evaluation environment, a fitness evaluator with a hybrid multi-objective particle swarm optimization algorithm is constructed. Based on the forward-looking lighting demand prediction results and fitness evaluator, a hybrid multi-objective particle swarm optimization algorithm is used to solve the multi-objective lighting control problem and generate the Pareto optimal solution set; Based on the actual scenario of rail transit stations, the weighted TOPSIS method is used to weight each locally optimal lighting control scheme in the Pareto optimal solution set to obtain the relative proximity of each locally optimal lighting control scheme. The local optimal lighting control scheme with the highest relative proximity is output as the global optimal lighting control scheme, and the simulation data of the global optimal lighting control scheme in the virtual evaluation environment is saved. The system executes the globally optimal lighting control scheme, continuously collects real data after execution, and performs twin calibration and model self-evolution based on simulation data, real data, high-precision passenger flow prediction distribution, and forward-looking lighting demand prediction results.

8. The rail transit lighting control method based on data analysis according to claim 7, characterized in that, Based on forward-looking lighting demand predictions and a fitness evaluator, a hybrid multi-objective particle swarm optimization algorithm is used to solve the multi-objective lighting control problem, generating a Pareto optimal solution set, including: Based on the forward-looking lighting demand forecast, the core attraction point of the initial particle swarm in the hybrid multi-objective particle swarm optimization algorithm is generated. Choosing the Logistic map as the chaos generator generates... A two-dimensional chaotic matrix is ​​obtained and linearly mapped to the interval [-1,1] to obtain a normalized offset direction matrix, where, M For particle swarm scale, For the size of the decision variable; Based on the core attraction point and offset direction matrix, generate including M An initial swarm of particles, wherein each particle in the swarm corresponds to an alternative lighting control scheme; The alternative lighting control schemes corresponding to each particle in the initial particle swarm are loaded into the lighting digital twin of the fitness evaluator. The high-precision passenger flow prediction distribution is input into the lighting digital twin for simulation. After the simulation, the objective function is used to calculate the corresponding fitness vector based on the key data of the lighting digital twin. The particle with the best fitness vector is stored as the optimal particle in the external archive. Using the fast non-dominated sorting algorithm, the initial particle swarm and the best particles in the external archive are divided into several non-dominated layers, and the crowding distance of each particle in each non-dominated layer is calculated. By introducing the Levy flight mechanism and the gray wolf cooperative idea, the particle state of the initial or previous iteration particle swarm is updated to obtain the updated particle swarm. The fitness evaluator is used to obtain the fitness vector of each particle, and the particle with the best fitness vector is stored as the optimal particle in the external archive. Repeat the iterative update steps until the number of iterations reaches the iteration threshold, and output all the best particles in the external archive as the Pareto optimal solution set.

9. The rail transit lighting control method based on data analysis according to claim 8, characterized in that, The system executes a globally optimal lighting control scheme, continuously collects real-world data after execution, and performs twin calibration and model self-evolution based on simulated data, real-world data, high-precision passenger flow forecasts, and forward-looking lighting demand predictions. This includes: The system executes the globally optimal lighting control scheme and continuously collects real data after execution. The real data includes real ambient light data and real passenger flow data collected by the sensor network after execution. The error value between the real ambient light data and the simulated data is obtained. If the error value exceeds the error threshold, the Kalman filter algorithm is used to automatically adjust the twin parameters in the lighting digital twin to obtain an updated lighting digital twin. The simulated data is the simulated ambient light data obtained by simulating the globally optimal lighting control scheme in the lighting digital twin. Real data, simulated data, and high-precision passenger flow forecast distribution and forward-looking lighting demand forecast results are paired to form new training samples, which are then stored in the training sample library. Periodically select a number of training samples from the training sample library to continuously train the passenger flow prediction model and obtain an updated passenger flow prediction model.

10. A data analysis-based rail transit lighting control device, used to implement the rail transit lighting control method as described in any one of claims 1-9, characterized in that, The device includes: The data acquisition and twin construction unit is used to collect multi-dimensional environmental and operational data in real time through a sensor network deployed in rail transit stations, and to construct a lighting digital twin that is mapped and dynamically evolves in real time with the physical station. The passenger flow and lighting demand forecasting unit is used to input environmental and operational data into a pre-built passenger flow forecasting model to generate high-precision passenger flow forecasting distribution and corresponding forward-looking lighting demand forecasting results for each area of ​​the rail transit station. The lighting control scheme optimization unit is used to use the lighting digital twin as a virtual evaluation environment, employ swarm intelligence optimization algorithms to solve multi-objective lighting control problems, generate globally optimal lighting control schemes, and perform twin calibration and model self-evolution.

Images