Collaborative optimization intelligent lighting system dynamic energy scheduling method and electronic equipment

Abstract

The invention relates to the technical field of property intelligent management, and provides a collaborative optimization intelligent lighting system dynamic energy scheduling method, which comprises the following steps: constructing an energy concentrator network of an intelligent lighting system, abstracting lighting equipment of a plurality of projects into an energy concentrator model, wherein each energy concentrator describes a conversion relation between electric energy and luminous flux through a coupling matrix; a multi-objective optimization function with maximization of the electric energy utilization rate as the core is established, the multi-objective optimization function integrates comfort, equipment and power grid constraints, and dynamic configuration is carried out based on operation parameters of the energy hub network; solving the multi-objective optimization function by adopting a two-stage optimization process, generating a reference operation plan and performing rolling correction; and issuing a dynamic dimming control signal to each lighting terminal through an energy concentrator network based on the optimized scheduling instruction obtained by solving so as to cooperatively schedule the electric energy of the intelligent lighting system. Therefore, the cross-region and multi-project intelligent lighting system is optimized in a collaborative manner, and the energy utilization rate is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent property management technology, and in particular to a dynamic energy dispatching method and electronic device for a collaboratively optimized intelligent lighting system. Background Technology

[0002] With the popularization of IoT technology, the application of smart lighting systems in individual buildings or projects has become quite mature. These systems typically allow users to control lighting fixtures with a single click (such as setting a meeting mode or a cinema mode) or perform quick control (such as adjusting the brightness of a specific area) via a mobile application or management platform, greatly improving ease of use. At the same time, the system also integrates basic functions such as equipment management (such as lighting status monitoring) and energy consumption statistics (such as monthly electricity consumption reports), helping managers understand the operational status of individual projects.

[0003] However, for large conglomerates with numerous office buildings, industrial parks, and shopping malls, the situation is far more complex. Each project has its own independent smart lighting system, creating isolated "information silos." While group managers can gain access to each project through account management, they struggle to integrate these fragmented systems at a macro level. For example, they cannot coordinate lighting strategies across projects during peak electricity consumption periods to maximize electricity savings for the entire group; nor can they quickly and efficiently replicate energy-saving strategies proven effective in one project to other similar projects.

[0004] At its root, traditional management platforms, while feature-rich, lack a "smart brain" capable of unified computation and coordinated scheduling at the group level. Equipment management and energy consumption statistics across different systems are fragmented, failing to provide data support for overall decision-making; and one-click scenario control and quick control strategies for each project operate independently, failing to create a synergistic effect.

[0005] Therefore, there is an urgent need in this field for an intelligent lighting scheduling method that can uniformly model all lighting devices and perform dynamic collaborative optimization, in order to solve the problem of low energy utilization caused by the inability of lighting systems to perform collaborative optimization scheduling. Summary of the Invention

[0006] The purpose of this invention is to solve the problem of low energy utilization caused by the inability of lighting systems to coordinate and optimize scheduling.

[0007] To address the aforementioned objectives, in one embodiment of the present invention, a collaborative optimization method for dynamic energy scheduling of an intelligent lighting system is proposed, comprising: S1: constructing an energy hub network for the intelligent lighting system, abstracting lighting equipment from multiple projects into energy hub models, wherein each energy hub describes the conversion relationship between electrical energy and light flux through a coupling matrix; S2: establishing a multi-objective optimization function with maximizing energy utilization as its core, the multi-objective optimization function integrating comfort, equipment, and grid constraints, and dynamically configuring it based on the operating parameters of the energy hub network; S3: solving the multi-objective optimization function using a two-stage optimization process, generating a baseline operating plan and continuously revising it; S4: based on the optimized scheduling instructions obtained from the solution, issuing dynamic dimming control signals to each lighting terminal through the energy hub network to collaboratively schedule the electrical energy of the intelligent lighting system.

[0008] Optionally, in some embodiments, the method, in step S2, establishing a multi-objective optimization function with maximizing power utilization as its core includes: S2a: defining a comprehensive evaluation index for power utilization, which integrates demand satisfaction, supply-demand matching accuracy, and load stability; S2b: constructing a comprehensive utility function, which integrates the comprehensive evaluation index for power utilization, the total cost index, and the comfort index, and is expressed as: in, The value of the comprehensive utility function. As an indicator of electricity utilization efficiency, For total cost indicators, For comfort indicators, S2c: Configuring weighting coefficients for the comprehensive utility function, the weighting coefficients being dynamically adjusted based on the time strategy and spatial behavior probability field.

[0009] Optionally, some embodiments of the method further include the following intelligent prediction steps based on behavior perception: S5: Construct a spatial behavior probability field based on a hidden Markov model, and calculate the movement trajectory and dwell probability of people in space through multi-source sensor data fusion; S6: Recursively calculate the state sequence probability of the hidden Markov model using a forward computation process to generate a spatial occupancy probability distribution prediction for future time periods; S7: Input the spatial occupancy probability distribution prediction into the multi-objective optimization function to dynamically adjust the weight coefficients in the illuminance service quality constraint; S8: Based on the spatiotemporal evolution characteristics of the spatial behavior probability field, perform pre-emptive control of the lighting system to make the illumination distribution match the movement pattern of people in advance.

[0010] Optionally, in some embodiments, the method, in step S5, constructing the spatial behavior probability field based on the Hidden Markov Model includes: S5a: constructing a multi-source sensor data fusion module to perform spatiotemporal registration and state estimation on heterogeneous sensor data in the observation vector interwoven from the multi-source sensor data; S5b: establishing a probability propagation model based on the Gaussian diffusion kernel function to dynamically update the spatiotemporal distribution of the spatial behavior probability field according to the personnel's movement speed and direction, wherein the Gaussian diffusion kernel function is expressed as:

[0011] in and These represent the source and target locations of the personnel, respectively. S5c: Design a state recognition mechanism based on maximum a posteriori probability estimation, and analyze the optimal state sequence of the hidden Markov model through the Viterbi decoding process; S5d: Couple the predicted output of the spatial behavior probability field with the dynamic dimming control signal to realize adaptive lighting control based on behavior prediction.

[0012] Optionally, in some embodiments, the method, in step S5b, establishing the probability propagation model based on the Gaussian diffusion kernel function includes an adaptive parameter optimization step: S5b1: Constructing a mapping table of personnel movement speed and diffusion parameters, wherein the diffusion parameters... With the speed of personnel movement Satisfy the inverse relationship , S5b2: The adjustment coefficient is calculated based on the environmental layout characteristics; S5b3: The sliding window real-time estimation method is adopted, based on the personnel movement speed-diffusion parameter mapping table, using the nearest... The adjustment coefficient is dynamically updated based on the personnel location data at each time step. S5b3: Through multi-core function fusion, combined with the updated adjustment coefficient S5b4: The Gaussian kernel and the exponential kernel are weighted and combined to generate an adaptive diffusion kernel function; S5b5: The output of the adaptive diffusion kernel function is integrated into the spatial behavior probability field, and the probability of personnel movement is predicted based on the result of the multi-kernel function fusion method.

[0013] Optionally, the method in some embodiments further includes a behavior prediction enhancement step based on digital twins: S9: Constructing a digital twin model of the intelligent lighting system, wherein the digital twin model adopts a deep deterministic strategy gradient approach; S10: Simulating the spatiotemporal evolution of the spatial behavior probability field through the digital twin model to generate a spatial occupancy probability distribution that meets a preset quality standard for future time periods; S11: Fusing the spatial occupancy probability distribution that meets the preset quality standard with the forward computation output of the hidden Markov model to generate an enhanced behavior prediction; S12: Dynamically adjusting the illuminance service quality constraint weights in the multi-objective optimization function based on the enhanced behavior prediction.

[0014] Optionally, the method in some embodiments further includes an enhanced modeling step based on neural radiation fields: S13: Construct a spatial occupancy perception model based on neural radiation fields, and calculate the illuminance and personnel density distribution at each point in the space through volume rendering integral; S14: Couple the neural radiation field model with the hidden Markov model to generate a spatiotemporal occupancy prediction sequence that meets preset accuracy requirements; S15: Dynamically adjust the distribution parameters and evolution law of the spatial behavior probability field based on the spatiotemporal occupancy prediction that meets preset accuracy requirements; S16: Input the optimized spatial behavior probability field into the multi-objective optimization function to anticipate and match the illumination distribution and personnel movement patterns of the intelligent lighting system.

[0015] Optionally, in some embodiments, the method of constructing a spatial occupancy perception model based on neural radiation fields in step S13 includes the following modeling steps: S13a: using a multi-level feature extraction method, extracting spatial geometric features and optical properties based on the original sensor data stream; S13b: designing a spatiotemporal consistency loss function, using the output of the multi-level feature extraction method to ensure a smooth transition of occupancy prediction between consecutive frames; S13c: using an online adaptive rendering method, dynamically correcting the neural radiation field parameters based on the feedback of the spatiotemporal consistency loss function; S13d: establishing a multi-scale fusion method, integrating the output of the online adaptive rendering with traditional sensor data to generate an occupancy distribution that meets preset standards.

[0016] Optionally, in some embodiments, the spatiotemporal consistency loss function designed in step S13b is defined as: in: This represents temporal consistency loss, measuring the difference in occupancy predictions between adjacent frames; This represents the spatial smoothness loss, measuring the spatial gradient of the occupancy prediction. and This is a hyperparameter used to balance temporal consistency and spatial smoothness.

[0017] Optionally, the method in some embodiments further includes a cooperative optimization step based on distributed model predictive control: S17: Construct a distributed optimization architecture, decompose the energy hub network into multiple subsystems and assign local controllers; S18: Design a cooperative optimization method based on alternating direction multiplier processes, and use the distributed optimization architecture to achieve global consistency optimization through the exchange of state information between adjacent subsystems; S19: Establish a distributed rolling optimization framework, based on the cooperative optimization method, each local controller solves the subsystem optimization problem in parallel and exchanges boundary state information; S20: Coordinate the optimization results of each local controller through a consensus protocol, and generate a cooperative control strategy that satisfies the global optimization objective based on the boundary state information.

[0018] Optionally, some embodiments of the method further include a game theory-based multi-item resource allocation optimization step: S21: Establish a game theory model for multi-item resource allocation, and maximize the overall utility function using a Nash bargaining solution, wherein the utility function integrates energy consumption and comfort indices; S22: Adopt a distributed optimization approach, with each local controller solving the local resource allocation problem in parallel, and exchanging boundary state information through an alternating direction multiplier process; S23: Coordinate the optimization results of each local controller through a consensus protocol to generate a global resource allocation strategy; S24: Apply the global resource allocation strategy to the operation plan of the energy hub network to coordinate the scheduling of the multi-item resources.

[0019] Optionally, the method in some embodiments further includes the step of establishing an abstract energy consumption topology model for the energy hub network: S25: Based on the energy hub network, define topology nodes and edges, where nodes represent the energy hub or lighting device, edges represent energy transmission paths, and initialize a topology adjacency matrix to describe the connection relationships between nodes; S26: Establish a topology energy consumption mapping function, the mapping function being expressed as... ,in It is an energy flow vector, representing the amount of energy transferred along each path; It is a topological adjacency matrix that describes the connection weights and transmission efficiency between nodes; It is a power input vector, defined based on the input vector of the energy hub; S27: The state changes of the topology nodes are monitored in real time through the sensor network, and the parameters of the topology adjacency matrix are updated in an adaptive manner to calibrate the abstract energy consumption topology model; S28: The output of the calibrated abstract energy consumption topology model is integrated into the multi-objective optimization function to dynamically optimize the energy scheduling strategy to improve network energy efficiency and stability.

[0020] Optionally, the method in some embodiments further includes a real-time optimization step for energy transmission paths: S29: Define multiple alternative energy transmission paths in the abstract energy consumption topology model, each path corresponding to a different node connection sequence; S30: Establish a path comprehensive evaluation function:

[0021] in For path The overall score, Represents a node arrive Connection efficiency, and They are nodes and power, For transmission delay, S31: Based on the real-time system status, calculate the comprehensive score of each candidate path through the path evaluation function; S32: Select the path whose comprehensive score meets the preset threshold requirement as the preferred energy transmission channel; S33: Send the configuration parameters of the preferred energy transmission channel to the relevant nodes for execution.

[0022] Optionally, the method in some embodiments further includes a high-order topology modeling step based on hypergraph theory: S34: Constructing a hypergraph topology structure for the intelligent lighting system, where vertices represent lighting devices and hyperedges represent multi-device collaborative control tasks or shared physical space regions; S35: Based on the hypergraph topology structure, calculating the spatial light field propagation using a bidirectional scattering distribution function kernel to establish an optical coupling relationship matrix between devices; S36: Using the optical coupling relationship matrix, performing message passing using a physically-informed graph convolution method to aggregate the physical features and optical properties of neighboring nodes; S37: Integrating the message-passing optimized hypergraph topology into the coupling matrix of the energy hub network to enhance the accuracy of multi-device collaborative control.

[0023] Optionally, in some embodiments, the method of issuing the dynamic dimming control signal in step S4 includes a scenario optimization step based on counterfactual reasoning: S4a: Define semantic scenario lighting quality targets in the digital twin environment, including multi-dimensional constraints such as illuminance distribution, uniformity, and glare index; S4b: Calculate the optimized control sequence that satisfies the scenario targets based on the current real-time state of the system through counterfactual optimization; S4c: Compare and fine-tune the optimized control sequence with measured sensor data in real time to ensure the adaptability of the control strategy; S4d: Issue the corrected control command to each lighting terminal through the energy hub network to robustly achieve the scenario targets.

[0024] Optionally, the method in some embodiments further includes a design step of a nonlinear observer with disturbance suppression characteristics: S38: Based on the dynamic characteristics of the energy hub network, construct a disturbance estimation framework based on an extended state observer, and enhance the tracking capability for unmodeled dynamics through a nonlinear gain function; S39: Design an adaptive disturbance compensation method to feedforward the total disturbance estimated by the extended state observer to the control law of the dynamic dimming control signal; S40: Ensure the convergence of the nonlinear observer through stability analysis, and ensure the consistent eventual boundedness of the disturbance estimation error; S41: Input the disturbance-compensated control quantity into the discretized state-space model, thereby improving the anti-interference performance of the model predictive control framework.

[0025] Optionally, the method in some embodiments further includes an adaptive gain scheduling step for the nonlinear observer: S42: based on the observation error vector of the extended state observer Design a gain scheduling method: in It is the adaptive gain matrix. As the reference gain, For adaptive rate, S43: Using the aforementioned adaptive gain matrix, the observation error vector is used. Update the gain parameters of the extended state observer to enhance the tracking capability for unmodeled dynamics; S44: Monitor the observation error vector in real time. Based on the convergence characteristics, the adaptive rate is dynamically adjusted. The parameters; S45: Feedforward the disturbance estimate output by the extended state observer after gain scheduling optimization to the control law, and feed back the compensation effect to the gain scheduling method to form a closed-loop optimization.

[0026] Optionally, the method in some embodiments further includes a perturbation compensation optimization step based on deep reinforcement learning: S46: Construct a deep reinforcement learning perturbation compensator coupled to the extended state observer, wherein the perturbation compensator adopts an actor-critic network structure; S47: Learn system dynamics and perturbation characteristics through the actor-critic network to generate an adaptive feedforward compensation signal; S48: Fuse the adaptive feedforward compensation signal with the total perturbation feedforward compensation signal estimated by the extended state observer to generate a composite control law; S49: Based on the execution effect of the composite control law, update the parameters of the actor-critic network through time difference error to perform online optimization of perturbation compensation.

[0027] Optionally, some embodiments of the method further include a collaborative optimization step with the active distribution network: S50: Based on the real-time operating status of the energy hub network, assess the demand response potential of each project and construct a flexible load aggregation model for the active distribution network; S51: Establish a two-layer, two-stage scheduling model, wherein the upper-layer model aims to minimize the operating cost of the distribution network, and the lower-layer model is coupled with the upper-layer model based on the multi-objective optimization function through price signals and load reduction; S52: Use KKT conditions to transform the lower-layer model into constraints for the upper-layer model, and use the Big M method to linearize the transformed complementary constraints; S53: Integrate the solved collaborative optimization strategy with the dynamic dimming control signal, so that the intelligent lighting system participates in the optimized operation of the distribution network as a flexible load.

[0028] Optionally, some embodiments of the method further include a distributed observer training step based on federated learning: S54: Construct a federated learning framework, where each edge node trains the model parameters of the expanded state observer locally; S55: Aggregate the model updates of each node into a global model through a secure aggregation method to protect data privacy; S56: Distribute the aggregated global model to each edge node to collaboratively optimize the observer parameters; S57: Periodically evaluate the federated learning effect, dynamically adjust the learning rate and aggregation frequency to ensure stable convergence of the training process.

[0029] In other embodiments of this application, an electronic device is also proposed, comprising: a processor, a computer-readable storage medium having a computer program stored thereon, the program being executed by the processor to implement the steps of the method of any other embodiment. [Technical Effects]: In one embodiment, the method constructs an energy hub network for the intelligent lighting system in step S1, abstracting the lighting equipment of multiple projects into energy hub models. Each model describes the conversion relationship between electrical energy and luminous flux through a coupling matrix. This step establishes a unified energy flow topology, providing a structural foundation for cross-project collaboration. Step S2 establishes a multi-objective optimization function with maximizing energy utilization as its core, integrating comfort, equipment, and grid constraints, and dynamically configuring it based on the operating parameters of the energy hub network, ensuring the comprehensiveness and adaptability of the optimization objectives. Step S3 employs a two-stage optimization process: first, a baseline operating plan is generated, and then rolling adjustments are made to address real-time changes, improving the stability and response speed of the optimization results. Step S4, based on the optimized scheduling instructions obtained from the solution, issues dynamic dimming control signals through the energy hub network, realizing coordinated energy scheduling. These steps, through interaction and cooperation, solve technical problems: the energy hub network provides a modeling framework, the multi-objective optimization function balances multiple constraints, the two-stage optimization ensures the combination of planning and real-time operation, and dynamic scheduling executes the optimization decisions. Through this system-level design, the technical solution solves the problem of low energy utilization caused by the inability of cross-regional and multi-project intelligent lighting systems to coordinate and optimize, and ultimately achieves overall energy efficiency improvement and operating cost reduction.

[0030] In one embodiment, the method constructs a personnel movement speed-diffusion parameter mapping table in step S5b1, linking the diffusion parameter σ with the personnel movement speed v through an inverse relationship, and introducing an adjustment coefficient k calculated based on environmental layout features. This step establishes the foundation for parameterization, enabling the probability propagation model to be initialized based on personnel movement speed and environmental features, providing a starting point for subsequent adaptive adjustments. In step S5b2, a sliding window real-time estimation method is used to dynamically update the adjustment coefficient k based on the personnel movement speed-diffusion parameter mapping table using personnel position data from the most recent N time steps. This step ensures that the model parameters can respond in real time to short-term changes in personnel movement patterns, avoiding model lag caused by environmental dynamism. In step S5b3, a multi-kernel function fusion method is used, combining the updated adjustment coefficient k, to generate an adaptive diffusion kernel function by weighted combination of Gaussian and exponential kernels. This process balances local accuracy and long-distance influence, enhancing the model's expressive power in complex spatial structures. Finally, in step S5b4, the output of the adaptive diffusion kernel function is integrated into the spatial behavior probability field, and personnel movement probability prediction is performed based on the result of the multi-kernel function fusion method. These steps work together: the mapping table provides a parameterized framework, real-time estimation ensures dynamic adaptability, and multi-kernel fusion improves model robustness, ultimately enabling the probability propagation model to accurately capture the movement patterns of people in space. This technical solution, through this step-by-step optimization mechanism, solves the problem of insufficient real-time performance and accuracy of the probability propagation model under changing personnel movement patterns and environmental layouts, thus providing a reliable basis for behavior prediction in adaptive lighting control.

[0031] In another embodiment, the method constructs a digital twin model of the intelligent lighting system in step S9, using a deep deterministic strategy gradient approach to create a virtual mirror of the physical system, providing a high-fidelity simulation environment for behavior prediction. Step S10 simulates the spatiotemporal evolution of the spatial behavior probability field using the digital twin model, generating a spatial occupancy probability distribution that meets preset quality standards for future time periods. This step utilizes the predictive capabilities of the digital twin to provide forward-looking data that surpasses traditional sensors. Step S11 fuses the accurate spatial occupancy probability distribution with the forward computation output of the hidden Markov model to generate enhanced behavior prediction, combining the advantages of data-driven and model-driven approaches and reducing the uncertainty of a single method. Step S12 dynamically adjusts the illuminance service quality constraint weights in the multi-objective optimization function based on the enhanced behavior prediction, enabling the lighting strategy to be optimized in real time according to the predicted behavior. In these steps: the digital twin model provides a virtual testing platform, behavior prediction fusion enhances data reliability, and weight adjustment enables adaptive control strategies. Through this integration, the technical solution solves the problem of unstable lighting control effects caused by high uncertainty in behavior prediction, ultimately improving the system's intelligence level and energy efficiency.

[0032] In another embodiment, the method employs a multi-level feature extraction approach in step S13a to extract spatial geometric features and optical properties based on the original sensor data stream, providing rich input information for the neural radiation field model and ensuring that the model can capture subtle changes in the environment. Step S13b designs a spatiotemporal consistency loss function, utilizing the output of the multi-level feature extraction approach to ensure a smooth transition in occupancy prediction between consecutive frames. This step, through loss function constraints, reduces temporal jitter and spatial inconsistency in the prediction results. Step S13c uses online adaptive rendering, based on the feedback of the spatiotemporal consistency loss function, to dynamically correct the neural radiation field parameters, enabling the model to self-optimize based on real-time data and adapt to dynamic environmental changes. Step S13d establishes a multi-scale fusion approach, integrating the output of the online adaptive rendering with traditional sensor data to generate an occupancy distribution that meets preset quality standards, fusing data from different sources to improve overall accuracy. The coordination of these steps reflects a closed-loop optimization from feature extraction to model correction: multi-level features provide basic information, the loss function ensures spatiotemporal consistency, online rendering enables real-time adaptation, and multi-scale fusion enhances the complementarity of data. The technical solution addresses the issues of inaccurate and inconsistent predictions by neural radiation field models in dynamic environments through this collaborative mechanism, thereby providing high-precision spatial perception capabilities for lighting systems.

[0033] In one embodiment, the method defines topology nodes and edges based on the energy hub network in step S25 and initializes the topology adjacency matrix, establishing a structured representation of energy flow and providing a visual basis for energy consumption analysis. Step S26 establishes a topology energy consumption mapping function, representing the energy flow vector as the product of the adjacency matrix and the power input vector, quantifying the energy transmission efficiency between nodes and making energy consumption relationships computable. Step S27 monitors the state changes of the topology nodes in real time through the sensor network and updates the parameters of the topology adjacency matrix in an adaptive manner, calibrating the abstract energy consumption topology model and ensuring the consistency between the model and the physical system. Step S28 integrates the output of the calibrated abstract energy consumption topology model into the multi-objective optimization function to dynamically optimize the energy scheduling strategy, realizing a closed-loop application from model to decision. These steps work together: topology modeling provides an analytical framework, the mapping function realizes energy consumption quantification, real-time monitoring maintains model accuracy, and integrated optimization improves scheduling efficiency. Through this step-by-step construction and calibration process, the technical solution solves the problem of lack of data support for scheduling decisions due to the opacity of energy flow, ultimately enhancing network energy efficiency and operational stability.

[0034] In another embodiment, the method constructs a spatial occupancy perception model based on a neural radiation field in step S13. It calculates the illuminance and personnel density distribution at various points in the space through volume rendering integrals, providing high-resolution environmental perception data and laying a fine foundation for behavior prediction. Step S14 couples the neural radiation field model with the Hidden Markov Model to generate a spatiotemporal occupancy prediction sequence that meets preset accuracy requirements, combining the high precision of deep learning with the temporal reasoning capability of probabilistic models. Step S15 dynamically adjusts the distribution parameters and evolution patterns of the spatial behavior probability field based on the spatiotemporal occupancy prediction, enabling the behavior model to be updated according to real-time perception data, thus improving the adaptability of the prediction. Step S16 inputs the optimized spatial behavior probability field into the multi-objective optimization function to achieve precise advance matching between illumination distribution and personnel movement patterns, ensuring that lighting control can respond to environmental changes in advance. These steps work together: the neural radiation field provides detailed perception, the coupled model enhances prediction reliability, the dynamic adjustment optimizes the behavior model, and the advance matching achieves proactive control. The proposed technical solution, through this integrated optimization, solves the problems of inaccurate environmental perception leading to lag in lighting control and low energy efficiency, thereby improving the system's response speed and energy utilization efficiency.

[0035] In another embodiment, the method constructs a hypergraph topology for the intelligent lighting system in step S34, where vertices represent lighting devices and hyperedges represent multi-device collaborative control tasks or shared physical space regions. This breaks through the binary limitations of traditional graph theory and can describe complex multi-coupling relationships between devices. Step S35, based on the hypergraph topology, uses a bidirectional scattering distribution function kernel to calculate spatial light field propagation, establishing an optical coupling relationship matrix between devices and quantifying the mutual influence of devices at the optical level. Step S36 utilizes the optical coupling relationship matrix and employs a physically-informed graph convolution method for message passing, aggregating the physical characteristics and optical properties of neighboring nodes to achieve collaborative reasoning and information fusion of device states. Step S37 integrates the message-passing-optimized hypergraph topology into the coupling matrix of the energy hub network, enhancing the accuracy of multi-device collaborative control and enabling global optimization to consider local interaction effects. These steps interact with each other: the hypergraph structure provides high-order modeling capabilities, light field propagation calculates physical interactions, message passing promotes information sharing, and integrated application strengthens control collaboration. Through this chain from structural modeling to control optimization, the technical solution solves the problems of complex relationships leading to optimization difficulties and inaccurate execution in multi-device collaborative control, and ultimately realizes the efficient collaborative operation of the lighting system in diverse scenarios.

[0036] In one embodiment, the method combines temporal consistency loss and spatial smoothness loss by defining a spatiotemporal consistency loss function. Temporal consistency loss measures the difference in occupancy predictions between adjacent frames, while spatial smoothness loss measures the spatial gradient of occupancy predictions, and the two are balanced through hyperparameters. This function, as an optimization objective during the training of the neural radiation field model, forces the model to produce temporally and spatially consistent predictions. Specifically, temporal consistency loss reduces inter-frame jitter in the prediction results, ensuring the temporal continuity of the occupancy distribution; spatial smoothness loss avoids sharp changes in the prediction graph, improving the naturalness of the spatial dimension. By minimizing this loss function, the model can learn more stable spatiotemporal features, reducing inconsistencies caused by environmental noise or missing data. The underlying logic of this technical solution is that the loss function imposes constraints on the model output from both spatiotemporal dimensions, ensuring the smoothness and reliability of the predictions, thereby solving the problems of inconsistency and insufficient robustness of neural radiation field models in dynamic scenes. Ultimately, this solution enhances the accuracy and practicality of spatial occupancy perception, providing more reliable environmental perception data for lighting control. Attached Figure Description

[0037] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings: Figure 1 A flowchart illustrating a dynamic energy scheduling method for an intelligent lighting system according to an embodiment of the present invention is shown. Figure 2 A flowchart illustrating an embodiment of the intelligent lighting scheduling method based on behavior prediction of the present invention is shown. Figure 3 This is a structural block diagram of an exemplary intelligent lighting system dynamic energy dispatching device. Detailed Implementation

[0038] The following are several embodiments to specifically implement the corresponding technical solutions of the present invention.

[0039] Example 1: like Figure 1 As shown, in a specific implementation scenario, the collaborative optimization method for dynamic energy scheduling of intelligent lighting systems is implemented according to the following steps: First, in step S1, an energy hub network for the intelligent lighting system is constructed. The lighting equipment distributed across three office parks A, B, and C in three different geographical locations is uniformly abstracted as an energy hub model. Each office park is considered a macro-level energy hub, while independent lighting areas within each park (such as open office areas, meeting rooms, and corridors) are modeled as micro-level energy hubs. For each micro-level energy hub, its coupling matrix is ​​established to describe the conversion relationship between electrical energy and light flux. Taking a certain type of LED lamp as an example, its coupling matrix is ​​represented in diagonal matrix form:

[0040] in Indicates the first The inherent luminous efficacy of each channel luminaire is obtained through equipment calibration; Indicates in The dimming level at any given time, with a value range of [value range missing]. The energy hub network is described by a weighted directed graph, where nodes include source nodes (total power inlet) and micro-energy hub nodes, and edges represent energy transmission paths with line loss coefficients as weights.

[0041] Next, in step S2, a multi-objective optimization function is established with maximizing energy utilization as its core. This optimization function integrates three main objectives: energy utilization rate, total cost, and comfort level. Specifically, a comprehensive utility function is constructed:

[0042] in It is calculated as the ratio of effective luminous flux to total power consumption. Includes electricity costs and equipment operation and maintenance costs. The uniformity of illumination and the deviation from the target illumination were quantified. The weighting coefficients... , , Dynamic adjustments are made based on real-time electricity pricing strategies, personnel activity patterns, and grid demand response signals.

[0043] Then, in step S3, a two-stage optimization process is used to solve the multi-objective optimization function. The first stage is executed every morning at dawn. Based on weather forecasts, work schedules, and electricity price information, the NSGA-II multi-objective evolutionary algorithm is used to generate a baseline operating plan for the next 24 hours. This plan includes dimming level settings for each lighting area every 15 minutes. The second stage is executed every minute during actual operation. Using a model predictive control framework, the baseline plan is continuously revised based on real-time sensor data and the deviation between actual and predicted data to ensure that the system always operates in an optimal state.

[0044] Finally, in step S4, based on the optimized scheduling instructions obtained from the solution, dynamic dimming control signals are sent to each lighting terminal through the energy hub network. Specifically, the cloud optimization engine sends the scheduling instructions to the edge gateways of each park. The gateways parse the instructions and convert them into dimming instructions for specific lighting controllers, which are then accurately transmitted to each target lamp through the Thread wireless network protocol, realizing the coordinated scheduling of power for the lighting systems of the three parks.

[0045]

Example 2

[0046] In this embodiment, the system first performs behavior prediction based on a Hidden Markov Model (HMM). In step S5, the spatial behavior probability field is constructed: the movement trajectories and dwell points of personnel are collected in real time by using multi-source sensors (including infrared grid sensors and cameras) uniformly deployed on the top of the office area. A simplified HMM is established, with its states defined as three spatial occupancy modes: "high-density clustering," "medium-density distribution," and "low-density / idle." Through step S5a, Kalman filtering is used to perform spatiotemporal registration and fusion of the raw coordinate data from different types of sensors to eliminate the measurement noise and blind zone effects of individual sensors, resulting in a more reliable estimation of personnel positions.

[0047] Next, the system optimizes and enhances the parameters of the behavior prediction model. In step S5b, a probability propagation model based on the Gaussian diffusion kernel function is established: a Gaussian kernel function is defined.

[0048] in It is the distance to the current location of the person, the diffusion parameter. The initial values ​​are set based on the average walking speed of people in the office area. Adaptive parameter optimization is implemented through steps S5b1 to S5b3: the system maintains a lookup table that maps the range of people's movement speeds to different... The system calculates the average movement speed every 5 minutes using historical position data within a sliding time window, and adjusts the lookup table accordingly. The value is then used to update the spatial behavior probability field, predicting the probability of people appearing in each area within the next few minutes.

[0049] Then, the system introduces a digital twin for prediction simulation and optimization. In step S9, a digital twin model of the office area is constructed: based on the building's BIM model, a three-dimensional scene is created in virtual space containing the location, orientation, light distribution curves, and work surface height of all lighting fixtures. In step S10, the spatiotemporal evolution of the spatial behavior probability field is simulated using the digital twin model: the personnel probability distribution predicted in step S5b is used as the input for the activities of virtual personnel in the digital twin, and a fast lighting simulation based on ray tracing is run to calculate the illuminance distribution cloud map of the entire office area in the future. In step S11, this high-precision illuminance distribution generated by the digital twin simulation is weighted and fused with the relatively coarse occupancy probability distribution directly output by the Hidden Markov Model (for example, in areas with poor sensor coverage, the digital twin prediction is given higher weight) to generate the final enhanced behavior prediction.

[0050] Furthermore, the system establishes an energy consumption topology model to support refined energy scheduling. In step S25, based on the energy hub network, an abstract energy consumption topology model for the office area is defined: each lighting circuit is defined as a topology node, and the edges between nodes represent the connection relationships of power lines. In step S26, a topology adjacency matrix is ​​initialized. Its elements The initial values ​​are preset based on the physical distance between loops and cable specifications. Through step S27, the system continuously monitors the input power and output luminous flux of each loop, and uses a simple exponential smoothing method to update the elements of the topology adjacency matrix online to calibrate the model so that it more accurately reflects the actual energy consumption transmission efficiency.

[0051] Finally, the system performs comprehensive optimization and decision-making. In step S12, based on the enhanced behavior prediction, the illuminance service quality constraint weights in the multi-objective optimization function are dynamically adjusted: for areas predicted to be densely populated and with insufficient illuminance, their comfort weights are significantly increased; for areas predicted to be uninhabited or with excessive illuminance, their comfort weights are appropriately reduced to prioritize energy-saving targets. Subsequently, the optimization engine, combined with the latest energy efficiency data provided by the calibrated abstract energy consumption topology model, solves the multi-objective optimization problem, calculates the optimal dimming command for each lamp, and sends it to the actuator to achieve dynamic and precise control of the lighting environment.

[0052]

Example 3

[0053] Reference Figure 3 The coordination control device 300 may include one or more of the following components: processing component 302, memory 304, power supply component 306, multimedia component 308, audio component 310, input / output (I / O) interface 312, sensor component 314, and communication component 316.

[0054] Figure 3 The coordinated control device 300 shown, with its processing component 302 and memory 304, is particularly suitable for executing the coordinated control method described in this invention. For example, the processing component 302 can perform the steps of data fusion, passenger flow prediction, and collaborative decision-making, and the memory 304 can store the fused data, prediction models, optimization algorithms, and computer programs. The communication component 316 is used for data interaction with the multimodal sensor network and elevator group control system within the building.

[0055] Processing component 302 typically controls and coordinates the overall operation of control device 300, such as operations associated with display, telephone calls, data communication, camera operation, and recording. Processing component 302 may include one or more processors 320 to execute instructions to complete all or part of the steps of the methods described above. Furthermore, processing component 302 may include one or more modules to facilitate interaction between processing component 302 and other components. For example, processing component 302 may include a multimedia module to facilitate interaction between multimedia component 308 and processing component 302.

[0056] Memory 304 is configured to store various types of data to support the operation of the coordination control device 300. Examples of this data include instructions for any application or method operating on the coordination control device 300, contact data, phone book data, messages, pictures, videos, etc. Memory 304 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0057] Power supply component 306 provides power to various components of the coordination control device 300. Power supply component 306 may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power to the coordination control device 300.

[0058] The multimedia component 308 includes a screen that provides an output interface between the coordination control device 300 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of the touch or swipe action but also the duration and pressure associated with the touch or swipe operation. In some embodiments, the multimedia component 308 includes a front-facing camera and / or a rear-facing camera. When the coordination control device 300 is in an operating mode, such as a shooting mode or a video mode, the front-facing camera and / or the rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0059] Audio component 310 is configured to output and / or input audio signals. For example, audio component 310 includes a microphone (MIC) configured to receive external audio signals when the coordination control device 300 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 304 or transmitted via communication component 316. In some embodiments, audio component 310 also includes a speaker for outputting audio signals.

[0060] I / O interface 312 provides an interface between processing component 302 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0061] Sensor assembly 314 includes one or more sensors for providing status assessments of various aspects of the coordinated control device 300. For example, sensor assembly 314 may detect the on / off state of the coordinated control device 300, the relative positioning of components such as the display and keypad of the coordinated control device 300, changes in the position of the coordinated control device 300 or one of its components, the presence or absence of user contact with the coordinated control device 300, the orientation or acceleration / deceleration of the coordinated control device 300, and temperature changes of the coordinated control device 300. Sensor assembly 314 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 314 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 314 may also include an accelerometer, a gyroscope, a magnetometer, a pressure sensor, or a temperature sensor.

[0062] Communication component 316 is configured to facilitate wired or wireless communication between coordination control device 300 and other devices. Coordination control device 300 can access wireless networks based on communication standards, such as WiFi, 2G, or 3G, or combinations thereof. In one exemplary embodiment, communication component 316 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 316 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0063] In an exemplary embodiment, the coordination control device 300 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the methods described above.

[0064] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 304 including instructions, which can be executed by a processor 320 of a coordination control device 300 to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0065] [Example 4]: Museum Lighting Control Based on Neural Radiation Field and Behavior Prediction This embodiment describes a dynamic energy scheduling method for an intelligent lighting system applied to a large, comprehensive museum. The museum's exhibition halls have complex structures, and the exhibits are sensitive to light. It is necessary to provide visitors with the optimal viewing environment while strictly preventing excessive lighting from damaging the exhibits.

[0066] In this embodiment, the method begins with step S13: constructing a spatial occupancy perception model based on neural radiation fields. Specifically, the system utilizes multiple low-light cameras deployed in the corners of the exhibition hall (outputting only binary motion contours to protect privacy) as the primary sensors. Through step S13, a lightweight neural radiation field model is constructed, which learns to infer the density distribution and approximate direction of movement of people in the entire three-dimensional space by combining this limited two-dimensional contour information with the three-dimensional model of the exhibition hall.

[0067] Further, in step S14, the neural radiation field model is coupled with a simplified Hidden Markov Model (HMM). The states of this HMM correspond to different visiting patterns, such as "clustering in front of famous paintings," "flowing in the exhibition hall aisle," and "staying in the rest area." Through step S14, the system uses real-time, high-precision occupancy data provided by the neural radiation field as observation input to drive the HMM to perform state decoding, thereby generating a spatiotemporal occupancy prediction sequence for the next 15 minutes and predicting the migration trend of people in the exhibition hall.

[0068] Then, step S15 is executed to dynamically adjust the spatial behavior probability field based on the high-precision spatiotemporal occupancy prediction. When the model predicts that a flow of people will arrive at a certain display case in 5 minutes, through step S15, the system will increase the weight of that area in the behavior probability field in advance and mark it as a "high-probability activity area" that needs to be given special attention during that period.

[0069] In this embodiment, in step S16, the optimized spatial behavior probability field is input into the multi-objective optimization function. The lighting control strategy is thus configured in advance. Through step S16, for areas that are about to become crowded, the system will slightly increase the ambient illuminance in advance to ensure the viewing effect; for areas that are about to be evacuated, a slow dimming process will be initiated in advance, achieving energy savings without the visitors noticing.

[0070] Optionally, the scenario optimization steps S4a-S4d based on counterfactual reasoning from other embodiments can also be combined to achieve exhibit protection. In step S4a, a scenario objective is defined for a specific precious silk painting display case: "When no one approaches, the illuminance is maintained at the minimum value L_min allowed by international standards; when it is predicted that someone is approaching, the illuminance smoothly increases to the optimal viewing value L_view before the visitor arrives." Through steps S4b and S4c, the system calculates and verifies the optimal control sequence based on the current state and predictions, and finally issues an instruction in step S4d to achieve precise and smooth control of "lights on when people approach, lights off when people leave," maximizing exhibit protection while ensuring a good user experience.

[0071] [Example 5]: Lighting Coordination in Large Airport Waiting Halls Based on Hypergraph Theory This embodiment describes a dynamic energy scheduling method for an intelligent lighting system applied to the waiting hall of a large international airport terminal. The waiting hall is a large space, including multiple functional areas such as check-in area, security check area, commercial area, and boarding gate rest area. The number of lighting devices is huge, and coordinated scene lighting between areas needs to be achieved based on flight information and passenger flow status.

[0072] In this embodiment, the method begins with step S34: constructing the hypergraph topology of the intelligent lighting system. Specifically, each luminaire or group of luminaires is defined as a vertex. Moving beyond traditional graph theory, a hyperedge is defined as the set of all devices participating in the same collaborative control task. For example, a hyperedge can connect all devices serving the "Flight XYZ Boarding Reminder" scenario: including the accent lighting in the boarding gate area, the directional ground lights leading to the boarding gate passage, and the background lighting for commercial area signs. Step S34 establishes a hypergraph model describing these complex, multi-faceted collaborative relationships.

[0073] Further, in step S35, the spatial light field propagation is calculated based on the hypergraph topology. For each set of devices represented by a hyperedge, the system uses a simplified bidirectional scattering distribution function kernel to calculate the overall lighting environment effect under their combined influence. Through step S35, an optical coupling relationship matrix is ​​established, which quantifies the degree of impact on the entire cooperative lighting environment (such as uniformity and accent lighting prominence) when the brightness of a device in the hyperedge changes.

[0074] Then, step S36 is executed, using a graph convolutional approach based on physical information for message passing. When the system decides to execute the "Flight XYZ Boarding Reminder" scenario, the corresponding hyperedge is activated through step S36. Device nodes within this hyperedge begin exchanging state information (such as current brightness and color temperature) through the graph convolutional network, and, referring to the optical coupling relationship matrix obtained in step S35, aggregate the optical characteristics of neighboring nodes to collaboratively calculate a new set of device parameters that best achieves the scenario objective.

[0075] In this embodiment, in step S37, the message-passing optimized hypergraph topology is integrated into the coupling matrix of the power hub network. This means that the collaboratively optimized device control commands are transformed into power input commands for the corresponding nodes in the power hub network. Through step S37, the system issues not control commands for a single device, but a complete, highly collaborative "scenario package" between devices, thereby achieving highly consistent and dynamic lighting effects in a large space and enhancing the accuracy and integrity of multi-device collaborative control.

[0076] Optionally, it can also participate in grid demand response by combining with the collaborative optimization steps S50-S53 of the active distribution network in other embodiments. When the airport receives a load reduction request from the grid, the system evaluates the adjustable lighting potential of each area through steps S50 and S51. In steps S52 and S53, it does not simply turn off all lights globally, but intelligently selects to reduce the brightness of advertising light boxes in commercial areas or turn off some background lighting in non-critical passages based on a hypergraph model, while ensuring that the lighting in core functional areas such as boarding gates and security checks is not affected, thus participating in the optimized operation of the distribution network as a high-quality flexible load.

[0077] In some embodiments of this application, some terms may be interpreted as follows: Energy Hub Network: An energy hub network is a topology model that unifies and interconnects multiple physically dispersed energy conversion and consumption units. In this network, each basic unit is modeled as an energy hub, and the network is used to describe and analyze cross-unit energy transmission, distribution, and collaborative optimization by defining the connection relationships between nodes and the direction of energy flow.

[0078] Energy Hub: The energy hub is the basic functional unit in an energy hub network, used to abstract the energy conversion and consumption behavior of one or more physical devices (such as lighting fixtures, sensors, and controllers). It mathematically describes the conversion relationship between one or more energy inputs and one or more energy service outputs through a coupling matrix.

[0079] Coupling Matrix: The coupling matrix is ​​a mathematical matrix used to quantify the transformation relationship between the input vector and the output vector of an energy hub. The elements in the matrix represent the transformation efficiency or transformation coefficient from a specific input to a specific output. These elements can be fixed parameters or functions that change dynamically with the system's operating state.

[0080] Spatial Behavior Probability Field: The spatial behavior probability field is a spatiotemporal distribution model that probabilistically describes the presence, movement, and dwelling of people or targets within a specified spatial region. This model generates the probability density distribution of behavioral events occurring at various points in space within a future time period by fusing multi-source sensor data with behavior prediction algorithms.

[0081] Digital twin model: The digital twin model is a dynamic mirror of the physical lighting system throughout its entire lifecycle in virtual space. It integrates geometric, physical, behavioral, and rule models, and achieves state synchronization, process simulation, performance prediction, and optimization decision-making through real-time data interaction with the physical system.

[0082] Neural Radiation Field: The neural radiation field is a deep learning-based implicit representation method for 3D scenes. It learns a mapping function from spatial location and viewing angle to scene color and density through a neural network, which is used for new viewpoint synthesis, 3D reconstruction, and high-precision inference of spatial attributes.

[0083] Abstract Energy Consumption Topology Model: The abstract energy consumption topology model is a graph theory-based model that provides a structured description of the energy flow paths and relationships in an energy hub network. It formally represents the energy transmission paths, connection weights, and transmission efficiency in the network by defining topological nodes, edges, and associated adjacency matrices.

[0084] Hypergraph topology: The hypergraph topology is a mathematical graph model used to describe multi-entity relationships within a system. In this structure, a hyperedge can connect any number of vertices, thus representing complex high-order cooperative or coupled relationships between multiple entities that cannot be described by traditional binary edges.

[0085] Extended State Observer: The extended state observer is a control theory component used to estimate system dynamics. Its core idea is to treat the sum of unknown dynamics of the system model, parameter changes, and external disturbances as a "total disturbance," and extend it into a new state variable, which is then estimated and compensated for in real time through the observer.

[0086] Counterfactual optimization algorithm: The counterfactual optimization algorithm is a decision-making method based on counterfactual reasoning. In a digital twin or simulation environment, it simulates and evaluates the possible system performance under virtual conditions different from historical facts (i.e., "counterfactual" scenarios) to find the optimal decision sequence that satisfies a specific objective.

[0087] Federated Learning Framework: The federated learning framework is a distributed machine learning paradigm. Under this framework, multiple participants (such as edge nodes) collaboratively train a shared global model without uploading their local data to a central server. Collaborative learning is achieved only by exchanging encrypted model parameter updates, aiming to protect data privacy.

[0088] Spatial occupancy perception model: The spatial occupancy perception model is a computational model that generates quantitative estimates of the presence, density, and distribution of people or targets in a space based on multi-source sensor data streams.

[0089] Occupancy distribution: The occupancy distribution is the output of the spatial occupancy perception model. It is a spatiotemporal distribution map that represents the probability or density of the presence of people or targets at different locations within a specified spatial region.

Claims

1. A synergistically optimized intelligent lighting system dynamic energy scheduling method, comprising: S1: Construct an energy hub network for an intelligent lighting system, abstracting the lighting equipment of multiple projects into an energy hub model, wherein each energy hub describes the conversion relationship between electrical energy and light flux through a coupling matrix; S2: Establish a multi-objective optimization function with maximizing power utilization as the core. The multi-objective optimization function integrates comfort, equipment and grid constraints, and is dynamically configured based on the operating parameters of the energy hub network. S3: A two-stage optimization process is used to solve the multi-objective optimization function, generate a baseline operating plan, and make rolling corrections; S4: Based on the optimized scheduling instructions obtained from the solution, dynamic dimming control signals are sent to each lighting terminal through the energy hub network to coordinate the scheduling of the power of the intelligent lighting system.

2. The method of claim 1, wherein, Step S2, establishing a multi-objective optimization function with maximizing energy utilization as its core, includes: S2a: defining a comprehensive evaluation index for energy utilization, integrating demand satisfaction, supply-demand matching accuracy, and load stability; S2b: constructing a comprehensive utility function, fusing the comprehensive evaluation index for energy utilization, the total cost index, and the comfort index, and expressing it as: wherein, is a total cost index, is an electrical energy utilization index, is a total cost index, is a comfort index, is a weight coefficient; S2c: is a weight coefficient configured for the comprehensive utility function, which is dynamically adjusted based on a time strategy and a spatial behavior probability field.

3. The method according to claim 1, characterized in that, It also includes the following intelligent prediction steps based on behavior perception: S5: Construct a spatial behavior probability field based on a hidden Markov model, and calculate the movement trajectory and dwell probability of people in space by fusing multi-source sensor data; S6: Recursively calculate the state sequence probability of the hidden Markov model using a forward computation process to generate a prediction of the spatial occupancy probability distribution for future time periods; S7: Input the predicted space occupancy probability distribution into the multi-objective optimization function to dynamically adjust the weight coefficients in the illumination service quality constraint; S8: Based on the spatiotemporal evolution characteristics of the spatial behavior probability field, perform pre-action control of the lighting system to make the illumination distribution match the movement pattern of personnel in advance.

4. The method according to claim 3, characterized in that, Step S5, constructing the spatial behavior probability field based on the Hidden Markov Model, includes: S5a: constructing a multi-source sensor data fusion module to perform spatiotemporal registration and state estimation on heterogeneous sensor data in the observation vector interwoven from the multi-source sensor data; S5b: establishing a probability propagation model based on the Gaussian diffusion kernel function to dynamically update the spatiotemporal distribution of the spatial behavior probability field according to the personnel's movement speed and direction, wherein the Gaussian diffusion kernel function is expressed as: in and These represent the source and target locations of the personnel, respectively. S5c: Design a state recognition mechanism based on maximum a posteriori probability estimation, and analyze the optimal state sequence of the hidden Markov model through the Viterbi decoding process; S5d: Couple the predicted output of the spatial behavior probability field with the dynamic dimming control signal to realize adaptive lighting control based on behavior prediction.

5. The method according to claim 4, characterized in that, The step S5b, which establishes a probability propagation model based on a Gaussian diffusion kernel function, includes an adaptive parameter optimization step: S5b1: Construct a mapping table of personnel movement speed and diffusion parameters, where the diffusion parameters... With the speed of personnel movement Satisfy the inverse relationship , S5b2: The adjustment coefficient is calculated based on the environmental layout characteristics; S5b3: The sliding window real-time estimation method is adopted, based on the personnel movement speed-diffusion parameter mapping table, using the nearest... The adjustment coefficient is dynamically updated based on the personnel location data at each time step. ; S5b3: Through multi-core function fusion, combined with the updated adjustment coefficients S5b4: The Gaussian kernel and the exponential kernel are weighted and combined to generate an adaptive diffusion kernel function; S5b5: The output of the adaptive diffusion kernel function is integrated into the spatial behavior probability field, and the probability of personnel movement is predicted based on the result of the multi-kernel function fusion method.

6. The method according to claim 3, characterized in that, It also includes a behavior prediction enhancement step based on digital twins: S9: Construct a digital twin model of the intelligent lighting system, wherein the digital twin model adopts a deep deterministic strategy gradient approach; S10: Simulate the spatiotemporal evolution of the spatial behavior probability field through the digital twin model to generate a spatial occupancy probability distribution that meets a preset quality standard for future time periods; S11: Fuse the spatial occupancy probability distribution that meets the preset quality standard with the forward computation output of the hidden Markov model to generate an enhanced behavior prediction; S12: Dynamically adjust the illuminance service quality constraint weights in the multi-objective optimization function based on the enhanced behavior prediction.

7. The method according to claim 3, characterized in that, It also includes an enhanced modeling step based on neural radiation fields: S13: Construct a spatial occupancy perception model based on neural radiation fields, and calculate the illuminance and personnel density distribution at each point in the space through volume rendering integral; S14: Couple the neural radiation field model with the hidden Markov model to generate a spatiotemporal occupancy prediction sequence that meets the preset accuracy requirements; S15: Dynamically adjust the distribution parameters and evolution law of the spatial behavior probability field based on the spatiotemporal occupancy prediction that meets the preset accuracy requirements; S16: Input the optimized spatial behavior probability field into the multi-objective optimization function to anticipate and match the illumination distribution and personnel movement patterns of the intelligent lighting system.

8. The method according to claim 7, characterized in that, The construction of the spatial occupancy perception model based on the neural radiation field in step S13 includes the following modeling steps: S13a: Using a multi-level feature extraction method, based on the original sensor data stream, extract spatial geometric features and optical properties; S13b: Design a spatiotemporal consistency loss function and use the output of the multi-level feature extraction method to ensure a smooth transition in occupancy prediction between consecutive frames; S13c: Dynamically correct neural radiation field parameters based on the feedback of the spatiotemporal consistency loss function using online adaptive rendering; S13d: Establish a multi-scale fusion method to integrate the output of the online adaptive rendering with traditional sensor data to generate an occupancy distribution that meets preset standards.

9. The method according to claim 8, characterized in that, The spatiotemporal consistency loss function designed in step S13b is defined as follows: in: This represents temporal consistency loss, measuring the difference in occupancy predictions between adjacent frames; This represents the spatial smoothness loss, measuring the spatial gradient of the occupancy prediction. and This is a hyperparameter used to balance temporal consistency and spatial smoothness.

10. The method according to claim 1, characterized in that, It also includes a collaborative optimization step based on distributed model predictive control: S17: Construct a distributed optimization architecture, decompose the energy hub network into multiple subsystems and assign local controllers; S18: Design a collaborative optimization method based on alternating direction multiplier processes, and use the distributed optimization architecture to achieve global consistency optimization through the exchange of state information between adjacent subsystems; S19: Establish a distributed rolling optimization framework, based on the collaborative optimization method, each local controller solves the subsystem optimization problem in parallel and exchanges boundary state information; S20: Coordinate the optimization results of each local controller through a consensus protocol, and generate a collaborative control strategy that satisfies the global optimization objective based on the boundary state information.

11. The method according to claim 10, characterized in that, It also includes a game theory-based multi-project resource allocation optimization step: S21: Establish a game theory model for multi-project resource allocation, and maximize the overall utility function with a Nash bargaining solution, wherein the utility function integrates energy consumption and comfort indicators; S22: Adopt a distributed optimization approach, in which each local controller solves the local resource allocation problem in parallel, and exchanges boundary state information through an alternating direction multiplier process; S23: Coordinate the optimization results of each local controller through a consensus protocol to generate a global resource allocation strategy; S24: Apply the global resource allocation strategy to the operation plan of the energy hub network to coordinate the scheduling of the multi-project resources.

12. An electronic device comprising: A processor and a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by the processor, it implements the steps of the method as described in any one of claims 1 to 11.

Images