Scene parameter optimization method, system and device for smart under-screen

Abstract

The application discloses a scene parameter optimization method, system and device of a smart middle screen, relates to the technical field of data processing, and the method comprises the following steps: acquiring static scene data and dynamic scene data of a scene in which the smart middle screen is located; performing feature extraction on the static scene data and the dynamic scene data respectively to obtain scene static features and scene dynamic features; calling a pre-generated static comfort prediction model, taking the lowest comfort degree and device parameter space as constraints, and based on the scene static features, performing reverse optimization by using a Bayesian algorithm to generate a global comfort baseline parameter; determining parameter constraint conditions based on the global comfort baseline parameter and device parameter boundaries; calling a dynamic comfort prediction model, taking a target comfort degree as an optimization target, taking the parameter constraint conditions as constraints, and based on the scene dynamic features, performing reverse optimization by using a differential evolution algorithm to generate partition comfort adjustment parameters. The application has the effect of improving user experience.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to methods, systems and devices for optimizing scene parameters of smart screens. Background Technology

[0002] With the popularization of smart homes and smart offices, smart central screens, as core terminals integrating display, interaction, and control functions, are typically linked with various devices such as air conditioners, fresh air systems, lighting, curtains, and humidifiers, undertaking the important task of environmental regulation. To ensure user comfort and reduce energy consumption, the operating parameters of the smart central screen and its linked devices need to be optimized, such as temperature settings, fan speed, light intensity, and curtain opening. However, this type of application scenario has significant unique characteristics: on the one hand, there are long-term stable factors such as room orientation, area, building envelope, and equipment installation location, whose impact on comfort is continuous and fundamental; on the other hand, there are also real-time changing factors such as human activity, outdoor weather, time variations, and local heat sources, which cause dynamic fluctuations in comfort requirements in different areas. Furthermore, due to the diversity of spatial zoning and equipment types, the requirements for the same environmental parameters in different areas may vary significantly, and equipment operating parameters are usually subject to multiple constraints such as hardware boundaries, safety thresholds, and comfort levels.

[0003] Currently, common approaches to parameter optimization for smart screens and interconnected devices are primarily based on preset rules, simple feedback control, or a single global optimization model. For example, constant temperature or humidity control is performed based on user-defined target values, or a single device is adjusted independently using data from a single environmental sensor. These methods have significant limitations in practical applications: First, they often treat the scene as a whole, failing to effectively distinguish between long-term stable characteristics and real-time changing characteristics, resulting in control parameters that cannot simultaneously adapt to basic environmental conditions and temporary disturbances. Second, most methods use globally uniform parameter settings, ignoring the personalized comfort needs of different zones, easily leading to problems such as excessively cold or hot areas, and uneven illumination. Furthermore, existing optimization methods have low optimization efficiency when facing multiple constraints such as device parameter boundaries and minimum comfort levels, making it difficult to quickly generate optimal parameters that meet constraints under real-time changes, resulting in response lag or parameter infeasibility. In addition, some methods rely on simple linear or empirical models to predict comfort, making it difficult to accurately characterize the nonlinear coupling relationship between multiple environmental factors and device parameters, further affecting the rationality and robustness of parameter optimization.

[0004] In summary, the current technology's shortcomings in practical applications mainly lie in: its inability to simultaneously consider long-term stability and real-time dynamic characteristics; the lack of a refined control mechanism that coordinates global and regional aspects; low optimization efficiency under multiple constraints; and insufficient comfort prediction capabilities. Therefore, this field needs a method that can rapidly generate scenario parameters that balance global basic comfort and regionally differentiated needs, while satisfying constraints such as device parameter boundaries and minimum comfort levels, thereby improving the control effect and user comfort of smart central screens and linked devices. Summary of the Invention

[0005] To improve user experience, this application provides a method, system, and device for optimizing scene parameters of a smart central screen.

[0006] Firstly, this application provides a method for optimizing scene parameters of a smart central screen, employing the following technical solution: The methods for optimizing scene parameters of smart central screens include: Acquire static and dynamic scene data of the scene in which the smart central screen is located, wherein the static scene data is used to characterize the long-term stable state of the scene, and the dynamic scene data is used to characterize the real-time changes of the scene. Feature extraction is performed on the static scene data to obtain static scene features; and feature extraction is performed on the dynamic scene data to obtain dynamic scene features; A pre-generated static comfort prediction model is invoked. With the minimum comfort level and device parameter space as constraints, a Bayesian algorithm is used to perform reverse optimization based on the static features of the scene to generate global comfort baseline parameters. The static comfort prediction model is used to output comfort level based on the static feature data of the scene and the device parameter data. Based on the global comfort baseline parameters and device parameter boundaries, parameter constraints are determined, wherein the device parameter boundaries are the endpoint values ​​of the device parameter space; The pre-generated dynamic comfort prediction model is invoked, with the target comfort as the optimization objective and the parameter constraints as the constraints. Based on the dynamic characteristics of the scene, the differential evolution algorithm is used for reverse optimization to generate partition comfort adjustment parameters. The dynamic comfort prediction model is used to output comfort based on the scene dynamic feature data, global comfort baseline parameter data and partition comfort adjustment parameter data. The zonal comfort adjustment parameters are fused with the global comfort baseline parameters to obtain scene parameters, which are then used to perform zonal control of the smart central screen and linked devices.

[0007] By employing the above technical solution, static and dynamic scene data of the smart central screen's environment are acquired. Static scene data characterizes the long-term stable state of the scene, while dynamic scene data characterizes its real-time changes. Feature extraction is then performed on the static scene data to obtain static scene features; similarly, feature extraction is performed on the dynamic scene data to obtain dynamic scene features. A pre-generated static comfort prediction model is then invoked, using minimum comfort level and device parameter space as constraints. Based on the static scene features, a Bayesian algorithm is employed for backpropagation to generate global comfort baseline parameters. The static comfort prediction model outputs parameters based on the static scene feature data and device parameter data. Comfort is assessed first, and then parameter constraints are determined based on global comfort baseline parameters and device parameter boundaries. The device parameter boundaries are the endpoint values ​​of the device parameter space. A pre-generated dynamic comfort prediction model is then invoked, using the target comfort level as the optimization objective and the parameter constraints as constraints. Based on scene dynamic characteristics, a differential evolution algorithm is used for reverse optimization to generate zoned comfort adjustment parameters. The dynamic comfort prediction model outputs comfort levels based on scene dynamic feature data, global comfort baseline parameter data, and zoned comfort adjustment parameter data. The zoned comfort adjustment parameters are then fused with the global comfort baseline parameters to obtain scene parameters for zoned control of the smart central screen and linked devices. This application distinguishes between static and dynamic scene features and constructs a hierarchical optimization mechanism to achieve synergy between the global comfort baseline and zoned differentiated adjustments. Under constraints such as device parameter boundaries and minimum comfort levels, it improves the efficiency of multi-constraint optimization and the accuracy of comfort prediction, thereby effectively improving the control effect and user experience of the smart central screen and linked devices. This invention distinguishes between static and dynamic scene characteristics, enabling parameter optimization to adapt to both long-term stable factors and real-time changing factors. By constructing static and dynamic comfort prediction models respectively, it accurately depicts the nonlinear coupling relationship between multiple factors and comfort. Based on this, it performs hierarchical reverse optimization of global baseline and zonal adjustments to achieve synergy between global comfort and zonal differentiated needs. Meanwhile, under constraints such as device parameter boundaries and minimum comfort, Bayesian optimization and differential evolution algorithms ensure high optimization efficiency and parameter feasibility. The resulting scene parameters take into account both global basic comfort and zonal personalized needs, effectively improving the control effect and user experience of smart central screens and linked devices.

[0008] Optionally, the step of extracting features from the static scene data to obtain static scene features includes: The static scene data is classified to obtain spatial structure data, equipment configuration data and environmental baseline data. The spatial structure data includes at least one of room layout information, enclosure structure information and space size information. The equipment configuration data includes at least one of equipment type, installation location, rated parameters and controllable range. The environmental baseline data includes historical environmental parameter statistics. Feature extraction is performed on the spatial structure data to obtain spatial layout features; The device configuration data is subjected to feature encoding and attribute mapping to obtain device capability characteristics; The environmental baseline data is subjected to statistical feature construction and dimensionality reduction processing to obtain environmental baseline features; The spatial layout features, the equipment capability features, and the environmental baseline features are fused to obtain the static features of the scene.

[0009] By adopting the above technical solution, in order to generate static scene features, the static scene data is classified to obtain spatial structure data, equipment configuration data, and environmental baseline data. Among them, the spatial structure data includes at least one of room layout information, enclosure structure information, and space size information; the equipment configuration data includes at least one of equipment type, installation location, rated parameters, and controllable range; and the environmental baseline data includes historical environmental parameter statistics. Then, feature extraction is performed on the spatial structure data to obtain spatial layout features. Then, feature encoding and attribute mapping are performed on the equipment configuration data to obtain equipment capability features. Then, statistical feature construction and dimensionality reduction processing are performed on the environmental baseline data to obtain environmental baseline features. Finally, the spatial layout features, equipment capability features, and environmental baseline features are fused to obtain the static scene features.

[0010] Optionally, the step of extracting features from the dynamic scene data to obtain dynamic scene features includes: The dynamic scene data is classified to obtain environmental sensing data, personnel activity data, equipment operating status data, and external environment data. The environmental sensing data includes at least one of temperature, humidity, and light intensity. The personnel activity data includes at least one of personnel location, number of personnel, and activity type. The equipment operating status data includes at least one of equipment on / off status, operating mode, load parameters, and energy consumption data. The external environment data includes at least one of outdoor temperature, outdoor light intensity, and weather conditions. Based on a preset time window and sliding step size, the dynamic scene data is sliced ​​to obtain multiple dynamic scene segments, wherein the dynamic scene segments include environmental sensing segments, personnel activity segments, equipment operating status segments, and external environment segments. For each dynamic scene segment, feature extraction is performed on the corresponding environmental sensing segment, personnel activity segment, equipment operating status segment, and external environment segment to obtain the corresponding environmental dynamic features, personnel dynamic features, equipment dynamic features, and external dynamic features. The environmental dynamic features, personnel dynamic features, equipment dynamic features, and external dynamic features within the same dynamic scene segment are spliced ​​together to obtain the segment dynamic features; The dynamic features of each segment are time-series aggregated to obtain the scene dynamic features.

[0011] By adopting the above technical solution, in order to generate dynamic scene features, the dynamic scene data is classified to obtain environmental sensing data, personnel activity data, equipment operating status data, and external environment data. The environmental sensing data includes at least one of temperature, humidity, and light intensity; the personnel activity data includes at least one of personnel location, number of personnel, and activity type; the equipment operating status data includes at least one of equipment on / off status, operating mode, load parameters, and energy consumption data; and the external environment data includes at least one of outdoor temperature, outdoor light intensity, and weather conditions. Then, based on a preset time window and sliding step size, the dynamic scene data is sliced ​​to obtain... Multiple dynamic scene segments are obtained, including environmental sensing segments, personnel activity segments, equipment operation status segments, and external environment segments. For each dynamic scene segment, features are extracted from the corresponding environmental sensing segment, personnel activity segment, equipment operation status segment, and external environment segment to obtain corresponding environmental dynamic features, personnel dynamic features, equipment dynamic features, and external dynamic features. Then, the environmental dynamic features, personnel dynamic features, equipment dynamic features, and external dynamic features within the same dynamic scene segment are concatenated to obtain the segment dynamic features. Finally, the dynamic features of each segment are temporally aggregated to obtain the scene dynamic features.

[0012] Optionally, the step of calling a pre-generated static comfort prediction model, constrained by the minimum comfort level and the device parameter space, and generating global comfort baseline parameters by using a Bayesian algorithm for back-optimization based on the static features of the scenario, includes: The pre-generated static comfort prediction model is invoked, and the device parameter space of the smart central screen and the linked devices and the set minimum comfort level are obtained. The device parameter space consists of the adjustable parameters of each linked device and the value range of the adjustable parameters. The static features of the scene are used as the fixed input of the static comfort prediction model, the device parameter space is used as the parameter search domain, and the first constraint condition is that the comfort level output by the static comfort prediction model is not lower than the minimum comfort level. Multiple initial equipment parameter sets are extracted within the parameter search domain, and the static comfort prediction model is called respectively to obtain the corresponding comfort prediction values. An equipment parameter sample set is constructed based on the initial equipment parameter sets and the corresponding comfort prediction values. Construct a Bayesian optimization objective function, wherein, under the condition of satisfying the first constraint, the Bayesian optimization objective function aims to minimize the excess between the comfort prediction value and the minimum comfort value, and imposes a penalty on the equipment parameter samples that do not satisfy the first constraint. Based on the device parameter sample set, a surrogate model for the Bayesian optimization objective function is constructed; Based on the surrogate model and the preset acquisition function, the next set of equipment parameters is determined within the parameter search domain; and, based on the static comfort prediction model, the comfort of the next set of equipment parameters is predicted, and the next set of equipment parameters and the corresponding comfort prediction value are added to the equipment parameter sample set; and, the surrogate model is refitted and iterated until the first iteration termination condition is met, wherein the first iteration termination condition includes the current iteration number reaching a first preset number; Select a group of device parameters from the device parameter sample set that satisfies the first constraint and minimizes the function value of the Bayesian optimization objective function, and use them as global comfort baseline parameters. The global comfort baseline parameters are used to characterize the basic operating state of the smart central screen and linked devices that meets the minimum comfort level and does not violate the device parameter boundaries.

[0013] By adopting the above technical solution, in order to generate global comfort baseline parameters, a pre-generated static comfort prediction model is invoked, and the device parameter space of the smart central screen and linked devices, as well as the set minimum comfort level, are obtained. The device parameter space consists of the adjustable parameters of each linked device and the range of their values. Then, the static features of the scene are used as the fixed input of the static comfort prediction model, and the device parameter space is used as the parameter search domain. The first constraint is that the comfort level output by the static comfort prediction model is not lower than the minimum comfort level. Multiple initial device parameter sets are then extracted from the parameter search domain, and the static comfort prediction model is invoked for each set to obtain the corresponding comfort prediction values. A device parameter sample set is constructed based on the initial device parameter sets and the corresponding comfort prediction values. Finally, a Bayesian optimization objective function is constructed, where, under the first constraint, the Bayesian optimization objective function minimizes the excess between the comfort prediction value and the minimum comfort level. The system aims to determine the target and penalize device parameter samples that do not meet the first constraint. Then, based on the device parameter sample set, a surrogate model of the Bayesian optimization objective function is constructed. Based on the surrogate model and a preset acquisition function, the next set of device parameters is determined within the parameter search domain. Furthermore, based on a static comfort prediction model, comfort prediction is performed on the next set of device parameters, and the next set of device parameters and its corresponding comfort prediction value are added to the device parameter sample set. Finally, the surrogate model is refitted and iterated until the first iteration termination condition is met, where the first iteration termination condition includes the current iteration number reaching a first preset number. Then, the device parameter set that satisfies the first constraint and minimizes the function value of the Bayesian optimization objective function is selected from the device parameter sample set as the global comfort baseline parameter. The global comfort baseline parameter characterizes the basic operating state of the smart central screen and linked devices, which meets the minimum comfort level and does not violate the device parameter boundaries.

[0014] Optionally, the step of determining parameter constraints based on the global comfort baseline parameters and device parameter boundaries includes: Obtain the device parameter boundaries of the smart central screen and its linked devices, wherein the device parameter boundaries include the upper limit and lower limit of adjustable parameters; The global comfort baseline parameters are analyzed to obtain the baseline parameter values ​​of the adjustable parameters; For each adjustable parameter, the difference between the lower limit of the adjustable parameter and the corresponding baseline parameter value is taken as the lower limit of the adjustment amount of the adjustable parameter, and the difference between the upper limit of the adjustable parameter and the corresponding baseline parameter value is taken as the upper limit of the adjustment amount of the adjustable parameter. Parameter constraints are generated based on the lower limit and the upper limit of the adjustment amount of the adjustable parameter.

[0015] By adopting the above technical solution, in order to generate parameter constraints, the device parameter boundaries of the smart central screen and linked devices are obtained. The device parameter boundaries include the upper and lower limits of adjustable parameters. Then, the global comfort baseline parameters are parsed to obtain the baseline parameter values ​​of the adjustable parameters. For each adjustable parameter, the difference between the lower limit of the adjustable parameter and the corresponding baseline parameter value is used as the lower limit of the adjustment amount of the adjustable parameter, and the difference between the upper limit of the adjustable parameter and the corresponding baseline parameter value is used as the upper limit of the adjustment amount of the adjustable parameter. Then, parameter constraints are generated based on the lower and upper limits of the adjustment amount of the adjustable parameters.

[0016] Optionally, the step of calling a pre-generated dynamic comfort prediction model, using the target comfort as the optimization objective and parameter constraints as constraints, and generating partition comfort adjustment parameters by using a differential evolution algorithm for reverse optimization based on the scene's dynamic characteristics, includes: Invoke the pre-generated dynamic comfort prediction model and obtain the set target comfort level, wherein the target comfort level is not lower than the minimum comfort level; Based on the parameter constraints, multiple partition comfort adjustment parameter sets are randomly generated to form an initial population, wherein each partition comfort adjustment parameter set includes the adjustment amount of each adjustable parameter; For each partition comfort adjustment parameter group in the initial population, the scene dynamic features, the global comfort baseline parameters, and the partition comfort adjustment parameter group are input into the dynamic comfort prediction model to obtain the corresponding comfort prediction value. Construct a differential evolution fitness function, wherein the fitness function aims to minimize the deviation between the predicted comfort value and the target comfort value, and imposes a penalty on the partition comfort adjustment parameter group that does not meet the parameter constraints; Based on the fitness values ​​of each partition comfort adjustment parameter group in the initial population, differential mutation, crossover and selection operations are performed on the initial population to generate a new generation population. The new generation population is used as the current population and iterated until the second iteration termination condition is met. The second iteration termination condition includes the current iteration number reaching the second preset number or the function value of the differential evolution fitness function converging. The optimal set of partition comfort adjustment parameters with the best fitness value is selected from the final population and used as the partition comfort adjustment parameters.

[0017] By adopting the above technical solution, in order to generate partition comfort adjustment parameters, a pre-generated dynamic comfort prediction model is invoked, and a set target comfort level is obtained, wherein the target comfort level is not lower than the minimum comfort level. Then, according to parameter constraints, multiple partition comfort adjustment parameter sets are randomly generated to form an initial population. Each partition comfort adjustment parameter set includes the adjustment amount of each adjustable parameter. Then, for each partition comfort adjustment parameter set in the initial population, the scene dynamic features, global comfort baseline parameters, and the partition comfort adjustment parameter set are input into the dynamic comfort prediction model to obtain the corresponding comfort prediction value. Finally, a differential evolution fitness function is constructed, wherein the fitness function is based on... The objective is to minimize the deviation between the predicted comfort value and the target comfort value. Penalties are imposed on partition comfort adjustment parameter sets that do not meet the parameter constraints. Then, based on the fitness values ​​of each partition comfort adjustment parameter set in the initial population, differential mutation, crossover, and selection operations are performed on the initial population to generate a new generation population. The new generation population is then used as the current population and iterated until the second iteration termination condition is met. The second iteration termination condition includes the current iteration number reaching a second preset number or the function value of the differential evolution fitness function converging. Finally, the partition comfort adjustment parameter set with the best fitness value is selected from the final population as the partition comfort adjustment parameter.

[0018] Optionally, the step of fusing the zone comfort adjustment parameters with the global comfort baseline parameters to obtain scene parameters for controlling the smart central screen and linked devices includes: The global comfort baseline parameters are analyzed to obtain the baseline parameter values ​​of each adjustable parameter; The comfort adjustment parameters of the partition are analyzed to obtain the adjustment amount corresponding to each adjustable parameter; The baseline parameter value corresponding to each adjustable parameter is superimposed with the corresponding adjustment amount to obtain the scene parameters; Based on the scenario parameters, corresponding device control commands are generated to perform zoned control of the smart central screen and linked devices.

[0019] By adopting the above technical solution, in order to generate scene parameters, the global comfort baseline parameters are parsed to obtain the baseline parameter values ​​of each adjustable parameter. Then, the zone comfort adjustment parameters are parsed to obtain the adjustment amount corresponding to each adjustable parameter. Then, the baseline parameter values ​​corresponding to each adjustable parameter are superimposed with the corresponding adjustment amounts to obtain scene parameters. Then, based on the scene parameters, corresponding device control commands are generated to perform zone control on the smart central screen and linked devices.

[0020] Optionally, the generation step of the static comfort prediction model includes: Acquire historical static scene data, historical device parameter data, and corresponding comfort labels, wherein the comfort labels are used to characterize the degree of environmental comfort; Feature extraction is performed on the historical static scene data to obtain historical scene static features; and the historical device parameter data is standardized and encoded to obtain historical device parameter features. The static features of the historical scene and the parameters of the historical equipment corresponding to the same scene are concatenated to obtain comfort sample features, and the comfort sample features are associated with the corresponding comfort labels to construct a comfort sample set; The comfort sample set is divided into a training set and a validation set; Based on the training set, the pre-built neural network model is trained to obtain the trained neural network model; Based on the validation set, the trained neural network model is validated to obtain a validated neural network model, which is then used as a static comfort prediction model.

[0021] By adopting the above technical solution, in order to generate a static comfort prediction model, historical static scene data, historical equipment parameter data, and corresponding comfort labels are obtained. The comfort labels are used to characterize the degree of environmental comfort. Then, feature extraction is performed on the historical static scene data to obtain historical scene static features. In addition, the historical equipment parameter data is standardized and encoded to obtain historical equipment parameter features. Then, the historical scene static features and historical equipment parameter features corresponding to the same scene are concatenated to obtain comfort sample features. The comfort sample features are associated with the corresponding comfort labels to construct a comfort sample set. Then, the comfort sample set is divided into a training set and a validation set. Then, based on the training set, a pre-built neural network model is trained to obtain a trained neural network model. Then, based on the validation set, the trained neural network model is validated to obtain a validated neural network model. The validated neural network model is used as the static comfort prediction model.

[0022] Secondly, this application also provides a scene parameter optimization system for a smart central screen, which adopts the following technical solution: The scene parameter optimization system for the smart central screen includes: The scene data acquisition module is used to acquire static scene data and dynamic scene data of the scene in which the smart central screen is located. The static scene data is used to characterize the long-term stable state of the scene, and the dynamic scene data is used to characterize the real-time changes of the scene. The scene feature extraction module is used to extract features from the static scene data to obtain static scene features; and to extract features from the dynamic scene data to obtain dynamic scene features. The baseline parameter generation module is used to call a pre-generated static comfort prediction model, which, constrained by the minimum comfort level and the device parameter space, uses a Bayesian algorithm to perform reverse optimization based on the static features of the scene to generate global comfort baseline parameters. The static comfort prediction model is used to output comfort based on the static feature data of the scene and the device parameter data. The parameter constraint determination module is used to determine parameter constraint conditions based on the global comfort baseline parameters and the device parameter boundaries, wherein the device parameter boundaries are the endpoint values ​​of the device parameter space; The parameter generation module is used to call a pre-generated dynamic comfort prediction model, with the target comfort as the optimization objective and the parameter constraints as the constraints. Based on the dynamic characteristics of the scene, the differential evolution algorithm is used to perform reverse optimization to generate partition comfort adjustment parameters. The dynamic comfort prediction model is used to output comfort based on the scene dynamic feature data, global comfort baseline parameter data and partition comfort adjustment parameter data. The parameter fusion control module is used to fuse the zone comfort adjustment parameters with the global comfort baseline parameters to obtain scene parameters for zone control of the smart central screen and linked devices.

[0023] Thirdly, this application also provides a computer device, which adopts the following technical solution: A computer device includes a memory and a processor, the memory storing a computer program executable on the processor, the processor executing the computer program to implement the method described in the first aspect.

[0024] In summary, this application includes at least the following beneficial technical effects: acquiring static and dynamic scene data of the scene in which the smart central screen is located, wherein the static scene data is used to characterize the long-term stable state of the scene, and the dynamic scene data is used to characterize the real-time changes of the scene; then, feature extraction is performed on the static scene data to obtain static scene features; and, feature extraction is performed on the dynamic scene data to obtain dynamic scene features, then a pre-generated static comfort prediction model is called, and based on the static scene features and constrained by the minimum comfort level and device parameter space, a Bayesian algorithm is used for back-optimization to generate global comfort baseline parameters, wherein the static comfort prediction model is used to generate global comfort baseline parameters based on the static scene feature data and device parameters. The system takes data, outputs comfort levels, and then determines parameter constraints based on global comfort baseline parameters and device parameter boundaries. The device parameter boundaries are the endpoint values ​​of the device parameter space. A pre-generated dynamic comfort prediction model is then invoked, using the target comfort level as the optimization objective and the parameter constraints as constraints. Based on scene dynamic characteristics, a differential evolution algorithm is used for reverse optimization to generate zonal comfort adjustment parameters. The dynamic comfort prediction model outputs comfort levels based on scene dynamic feature data, global comfort baseline parameter data, and zonal comfort adjustment parameter data. The zonal comfort adjustment parameters are then fused with the global comfort baseline parameters to obtain scene parameters for zonal control of the smart central screen and linked devices. This application, by distinguishing between static and dynamic scene features and constructing a hierarchical optimization mechanism, achieves synergy between the global comfort baseline and zonal differentiated adjustments. Under constraints such as device parameter boundaries and minimum comfort levels, it improves the efficiency of multi-constraint optimization and the accuracy of comfort prediction, thereby effectively improving the control effect and user experience of the smart central screen and linked devices. This invention distinguishes between static and dynamic scene characteristics, enabling parameter optimization to adapt to both long-term stable factors and real-time changing factors. By constructing static and dynamic comfort prediction models respectively, it accurately depicts the nonlinear coupling relationship between multiple factors and comfort. Based on this, it performs hierarchical reverse optimization of global baseline and zonal adjustments to achieve synergy between global comfort and zonal differentiated needs. Meanwhile, under constraints such as device parameter boundaries and minimum comfort, Bayesian optimization and differential evolution algorithms ensure high optimization efficiency and parameter feasibility. The resulting scene parameters take into account both global basic comfort and zonal personalized needs, effectively improving the control effect and user experience of smart central screens and linked devices. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the overall process of an embodiment of this application.

[0026] Figure 2 This is a schematic diagram of the system structure of this application.

[0027] Figure 3 This is a structural block diagram of the computer device described in this application. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0029] This application discloses a method for optimizing scene parameters of a smart central screen.

[0030] Reference Figure 1 The optimization methods for scene parameters of smart central screens include: Step S11: Obtain static scene data and dynamic scene data of the scene where the smart central screen is located. Static scene data is used to characterize the long-term stable state of the scene, and dynamic scene data is used to characterize the real-time changes of the scene.

[0031] It's important to note that acquiring both static and dynamic scene data for the smart screen's environment is crucial. Static scene data characterizes the long-term stability of the scene, such as room orientation, area, thermal performance of the building envelope, and equipment installation location. These factors remain relatively constant over extended periods, forming the foundation for comfort levels. Dynamic scene data, on the other hand, characterizes real-time changes in the scene, such as occupant distribution, indoor and outdoor temperature and humidity, light intensity, and time information. These factors fluctuate rapidly over time, directly impacting the immediate comfort needs of specific areas. Dividing scene data into static and dynamic categories helps in establishing separate predictive models and optimization strategies for influencing factors at different time scales. This avoids a single model simultaneously processing both slowly and rapidly changing data, which could lead to decreased accuracy or delayed response, thus improving the rationality and real-time nature of parameter optimization.

[0032] Step S12: Extract features from static scene data to obtain static scene features; and extract features from dynamic scene data to obtain dynamic scene features.

[0033] It should be noted that the purpose of feature extraction is to filter out independent feature representations that have a significant impact on comfort from the raw data, remove redundant information and noise interference, and reduce data dimensionality. By extracting static and dynamic features of the scene separately, we can provide compact and effective inputs for subsequent static and dynamic comfort predictions, reduce computational burden, and improve the model's sensitivity to key influencing factors, enabling the reverse optimization process to more accurately capture the mapping relationship between the environment and equipment parameters.

[0034] Step S13: Call the pre-generated static comfort prediction model, and use the minimum comfort level and device parameter space as constraints. Based on the static features of the scene, the Bayesian algorithm is used for reverse optimization to generate global comfort baseline parameters. The static comfort prediction model is used to output comfort level based on the static feature data of the scene and the device parameter data.

[0035] It should be noted that in this step, the Bayesian algorithm is used for reverse optimization, mainly based on its efficient sampling characteristics in high-cost black-box function optimization. The static comfort prediction model involves a complex nonlinear mapping between multi-dimensional scene static features and equipment parameters. If traversal or random search methods are used, the computational cost is huge and it is difficult to guarantee the quality of the search. Bayesian optimization approximates the comfort prediction function by constructing surrogate models such as Gaussian processes, and uses the sampling function to achieve a balance between exploring unknown regions and developing known better regions, thereby locating the better parameter combination that meets the constraints within a finite number of model evaluations. At the same time, the minimum comfort level is introduced as a constraint into the optimization process to ensure that the generated parameters do not sacrifice the user's basic comfort experience; the equipment parameter space defines the feasible search domain to prevent the generation of set values ​​that exceed hardware capabilities or safety thresholds. The resulting global comfort baseline parameters reflect the better equipment configuration under long-term stable conditions, providing a reliable starting point and benchmark for subsequent dynamic adjustments. This avoids frequent parameter fluctuations or deviations from the appropriate range caused by relying solely on real-time feedback, which is beneficial for maintaining a basic comfort level while reducing system energy consumption and operational complexity.

[0036] Step S14: Determine parameter constraints based on global comfort baseline parameters and device parameter boundaries, where the device parameter boundaries are the endpoint values ​​of the device parameter space.

[0037] It should be noted that determining parameter constraints based on the global comfort baseline parameters and device parameter boundaries aims to define a search range that is both safe and reasonable for subsequent dynamic adjustments of different zones. Device parameter boundaries, as endpoints of the device parameter space, directly reflect the hardware's allowed limit operating range. Including them in the constraints effectively prevents the optimization process from generating unexecutable parameters that exceed device capabilities or may cause safety issues, ensuring the physical feasibility of the final control commands. The global comfort baseline parameters, obtained through static optimization under long-term stable conditions, represent the balance between overall comfort and energy consumption. Using them as a reference benchmark to limit the magnitude of dynamic adjustments avoids excessive deviation from the global optimum in local areas in pursuit of immediate comfort, thus preventing abnormal increases in overall system energy consumption or control conflicts between different zones. Through the combined effect of these two levels of constraints, dynamic optimization of zones is confined to a clear and finite feasible domain. This retains the flexibility to adjust in response to real-time changes while reducing the search space, improving the convergence speed of the optimization algorithm and the engineering practicality of the parameter results. This allows the subsequently generated zone comfort adjustment parameters to quickly and stably approximate the solution that satisfies the target comfort level.

[0038] Step S15: Call the pre-generated dynamic comfort prediction model, take the target comfort as the optimization objective and the parameter constraints as the constraints, and use the differential evolution algorithm to perform reverse optimization based on the scene dynamic features to generate the partition comfort adjustment parameters. The dynamic comfort prediction model is used to output the comfort level based on the scene dynamic feature data, the global comfort baseline parameter data and the partition comfort adjustment parameter data.

[0039] It should be noted that the differential evolution algorithm is used for reverse optimization in this step because the search space for partition adjustment parameters in dynamic scenarios typically exhibits complex characteristics such as multi-peak, nonlinearity, and constraints. The differential evolution algorithm, based on population iteration, maintains population diversity through mutation, crossover, and selection mechanisms, possessing strong global search capabilities and adaptability to complex constraints. It is suitable for quickly finding partition adjustment amounts that meet the target comfort level within a search range jointly defined by global comfort baseline parameters and device parameter boundaries. Unlike the static stage, the dynamic stage requires real-time response to scene changes. The differential evolution algorithm can evaluate multiple candidate solutions in parallel, improving optimization convergence speed and meeting real-time requirements. Searching for partition adjustment amounts near the global comfort baseline parameters reduces the search space and improves optimization efficiency; it also ensures that partition adjustments do not deviate from the overall comfort baseline, avoiding excessive local adjustments that could lead to system oscillations or abnormally high energy consumption. The final generated partition comfort adjustment parameters provide differentiated compensation for the real-time dynamic characteristics of different regions, solving the problem that traditional global unified control cannot meet local needs or that independent local control disrupts overall coordination. This simultaneously improves user comfort and system operational stability in multi-partition scenarios.

[0040] Step S16: The zone comfort adjustment parameters and the global comfort baseline parameters are fused to obtain scene parameters for zone control of the smart central screen and linked devices.

[0041] It should be noted that the global baseline provides a comfortable foundation under long-term stable conditions, while the zonal adjustment reflects real-time changes and compensation for regional differences. The combination of the two makes the final scenario parameters both stable and adaptable. This fusion approach effectively avoids the shortcomings of a single control mode: if only global control is used, local needs are easily ignored; if each zone adjusts independently according to its own goals without global coordination, it may cause overall imbalance or increased energy consumption. After fusion, the parameters are naturally located within the feasible domain of the device and can be directly sent to the smart central screen and linked devices for execution, improving the practicality of the control strategy and the user comfort experience.

[0042] In the above implementation, static and dynamic scene data of the scene where the smart screen is located are acquired. Static scene data characterizes the long-term stable state of the scene, while dynamic scene data characterizes the real-time changes in the scene. Then, feature extraction is performed on the static scene data to obtain static scene features; and feature extraction is performed on the dynamic scene data to obtain dynamic scene features. A pre-generated static comfort prediction model is then invoked. Constrained by minimum comfort level and device parameter space, a Bayesian algorithm is used for backpropagation based on the static scene features to generate global comfort baseline parameters. The static comfort prediction model outputs comfort parameters based on the static scene feature data and device parameter data. The process involves several steps. First, based on the global comfort baseline parameters and device parameter boundaries, parameter constraints are determined. The device parameter boundaries are the endpoint values ​​of the device parameter space. Then, a pre-generated dynamic comfort prediction model is invoked, using the target comfort level as the optimization objective and the parameter constraints as constraints. Based on the scene's dynamic characteristics, a differential evolution algorithm is used for reverse optimization to generate zoned comfort adjustment parameters. The dynamic comfort prediction model outputs comfort levels based on scene dynamic feature data, global comfort baseline parameter data, and zoned comfort adjustment parameter data. The zoned comfort adjustment parameters are then fused with the global comfort baseline parameters to obtain scene parameters for zoned control of the smart central screen and linked devices. This application distinguishes between static and dynamic scene features and constructs a hierarchical optimization mechanism to achieve synergy between the global comfort baseline and zoned differentiated adjustments. Under constraints such as device parameter boundaries and minimum comfort levels, it improves the efficiency of multi-constraint optimization and the accuracy of comfort prediction, thereby effectively improving the control effect and user experience of the smart central screen and linked devices. This invention distinguishes between static and dynamic scene characteristics, enabling parameter optimization to adapt to both long-term stable factors and real-time changing factors. By constructing static and dynamic comfort prediction models respectively, it accurately depicts the nonlinear coupling relationship between multiple factors and comfort. Based on this, it performs hierarchical reverse optimization of global baseline and zonal adjustments to achieve synergy between global comfort and zonal differentiated needs. Meanwhile, under constraints such as device parameter boundaries and minimum comfort, Bayesian optimization and differential evolution algorithms ensure high optimization efficiency and parameter feasibility. The resulting scene parameters take into account both global basic comfort and zonal personalized needs, effectively improving the control effect and user experience of smart central screens and linked devices.

[0043] As a further implementation of the method, the step of extracting features from static scene data to obtain static scene features includes: Step S21: Classify the static scene data to obtain spatial structure data, equipment configuration data and environmental baseline data. The spatial structure data includes at least one of room layout information, enclosure structure information and space size information. The equipment configuration data includes at least one of equipment type, installation location, rated parameters and controllable range. The environmental baseline data includes historical environmental parameter statistics.

[0044] Step S22: Extract features from the spatial structure data to obtain spatial layout features.

[0045] Step S23: Perform feature encoding and attribute mapping on the device configuration data to obtain the device capability characteristics.

[0046] Step S24: Statistical feature construction and dimensionality reduction are performed on the environmental baseline data to obtain environmental baseline features.

[0047] Step S25: The spatial layout features, equipment capability features, and environmental baseline features are fused to obtain the static features of the scene.

[0048] It should be noted that static scene data encompasses multiple stable factors affecting long-term comfort. Processing it as a whole can easily reduce the accuracy of subsequent predictions due to feature mixing or information redundancy. Therefore, static scene data is first divided into three categories: spatial structure, equipment configuration, and environmental baseline. Features are then extracted separately for each category based on its unique characteristics, thus comprehensively and efficiently characterizing the fundamental conditions for long-term comfort. Specifically, spatial structure data reflects the geometric shape and enclosure characteristics of rooms, and its extracted spatial layout features can demonstrate the fundamental impact of spatial factors on the distribution of heat, light, and airflow. Equipment configuration data combines discrete categories with continuous parameters; the equipment capability features obtained through feature encoding and attribute mapping can accurately describe the adjustment capabilities and constraint boundaries of each piece of equipment. Environmental baseline data represents high-dimensional historical statistics; statistical feature construction and dimensionality reduction are used to effectively compress feature dimensions while preserving the main environmental patterns, reducing the computational burden. Finally, the three types of features are merged into scene static features, which not only maintains the independent contribution of each dimension of information to comfort, but also improves the compactness and representativeness of the features through redundancy reduction and dimensionality reduction. This provides reliable input for static comfort prediction and reverse optimization of global comfort baseline parameters, and enhances the accuracy and stability of long-term parameter optimization.

[0049] In the above embodiments, in order to generate static scene features, the static scene data is classified to obtain spatial structure data, equipment configuration data, and environmental baseline data. The spatial structure data includes at least one of room layout information, enclosure structure information, and space size information. The equipment configuration data includes at least one of equipment type, installation location, rated parameters, and controllable range. The environmental baseline data includes historical environmental parameter statistics. Then, feature extraction is performed on the spatial structure data to obtain spatial layout features. Then, feature encoding and attribute mapping are performed on the equipment configuration data to obtain equipment capability features. Then, statistical feature construction and dimensionality reduction processing are performed on the environmental baseline data to obtain environmental baseline features. Finally, the spatial layout features, equipment capability features, and environmental baseline features are fused to obtain the static scene features.

[0050] As a further implementation of the method, the step of extracting features from dynamic scene data to obtain dynamic scene features includes: Step S31: Classify the dynamic scene data to obtain environmental sensing data, personnel activity data, equipment operating status data, and external environment data. The environmental sensing data includes at least one of temperature, humidity, and light intensity; the personnel activity data includes at least one of personnel location, number of personnel, and activity type; the equipment operating status data includes at least one of equipment on / off status, operating mode, load parameters, and energy consumption data; and the external environment data includes at least one of outdoor temperature, outdoor light intensity, and weather conditions.

[0051] Step S32: Based on a preset time window and sliding step size, the dynamic scene data is sliced ​​to obtain multiple dynamic scene segments, including environmental sensing segments, personnel activity segments, equipment operating status segments, and external environment segments.

[0052] Step S33: For each dynamic scene segment, extract features from the corresponding environmental sensing segment, personnel activity segment, equipment operation status segment, and external environment segment to obtain the corresponding environmental dynamic features, personnel dynamic features, equipment dynamic features, and external dynamic features.

[0053] Step S34: The environmental dynamic features, personnel dynamic features, equipment dynamic features and external dynamic features within the same dynamic scene segment are spliced ​​together to obtain the segment dynamic features.

[0054] Step S35: Perform temporal aggregation on the dynamic features of each segment to obtain the scene dynamic features.

[0055] It should be noted that dynamic scene data is characterized by its multi-source heterogeneity, strong real-time nature, and interconnectedness. Directly processing it as a whole not only fails to highlight the immediate impact of each factor on comfort but also easily reduces the efficiency of subsequent prediction and optimization due to high data dimensionality and noise. Therefore, dynamic scene data is first divided into four categories: environmental sensing, personnel activity, equipment operating status, and external environment. This allows different types of influencing factors to be processed using methods that match their data characteristics during the feature extraction stage, avoiding information mixing. Based on this, a preset time window and sliding step size are introduced for slicing. The aim is to discretize the continuously changing dynamic process into several temporally related segments, preserving the temporal continuity of the data while providing a suitable granularity for segment-by-segment feature extraction and real-time processing. Extracting features separately for each type of data within each segment allows for a more accurate characterization of short-term fluctuations in environmental parameters, the instantaneous impact of personnel distribution and activity, changes in equipment operating status, and the intensity of external disturbances. This ensures that the contribution of each type of dynamic factor to comfort is expressed independently and fully. Subsequently, multiple features within the same segment are concatenated to form a complete representation of the scene state at that moment. This allows subsequent models to comprehensively consider the comfort level under the coupled effects of multiple factors, rather than viewing a single factor in isolation. Finally, the dynamic features of each segment are temporally aggregated to further capture the trends, delay effects, and short-term fluctuation patterns of comfort changes. This ensures that the final scene dynamic features contain both local instantaneous information and global temporal expressiveness, thus providing high-quality input for the dynamic comfort prediction model and supporting the stable optimization of subsequent zoned comfort adjustment parameters.

[0056] In the above implementation, to generate dynamic scene features, the dynamic scene data is classified to obtain environmental sensing data, personnel activity data, equipment operating status data, and external environment data. The environmental sensing data includes at least one of temperature, humidity, and light intensity; the personnel activity data includes at least one of personnel location, number of personnel, and activity type; the equipment operating status data includes at least one of equipment on / off status, operating mode, load parameters, and energy consumption data; and the external environment data includes at least one of outdoor temperature, outdoor light intensity, and weather conditions. Then, based on a preset time window and sliding step size, the dynamic scene data is sliced ​​to obtain... Multiple dynamic scene segments are generated, including environmental sensing segments, personnel activity segments, equipment operation status segments, and external environment segments. For each dynamic scene segment, features are extracted from the corresponding environmental sensing segment, personnel activity segment, equipment operation status segment, and external environment segment to obtain corresponding environmental dynamic features, personnel dynamic features, equipment dynamic features, and external dynamic features. Then, the environmental dynamic features, personnel dynamic features, equipment dynamic features, and external dynamic features within the same dynamic scene segment are concatenated to obtain the segment dynamic features. Finally, the dynamic features of each segment are temporally aggregated to obtain the scene dynamic features.

[0057] As a further implementation of the method, the step of calling a pre-generated static comfort prediction model, constrained by the minimum comfort level and the equipment parameter space, and generating global comfort baseline parameters by using a Bayesian algorithm for back-optimization based on static scene features includes: Step S41: Call the pre-generated static comfort prediction model and obtain the device parameter space of the smart central screen and the linked devices, as well as the set minimum comfort level. The device parameter space consists of the adjustable parameters of each linked device and the range of values ​​of the adjustable parameters.

[0058] Step S42: Use the static features of the scene as the fixed input of the static comfort prediction model, use the device parameter space as the parameter search domain, and set the comfort level output by the static comfort prediction model to be no less than the minimum comfort level as the first constraint condition.

[0059] Step S43: Extract multiple initial equipment parameter sets within the parameter search domain, and call the static comfort prediction model respectively to obtain the corresponding comfort prediction values. Construct an equipment parameter sample set based on the initial equipment parameter sets and the corresponding comfort prediction values.

[0060] Step S44: Construct a Bayesian optimization objective function, wherein the Bayesian optimization objective function aims to minimize the excess between the predicted comfort value and the minimum comfort value under the first constraint condition, and imposes a penalty on the equipment parameter samples that do not meet the first constraint condition.

[0061] Step S45: Based on the device parameter sample set, construct a surrogate model for the Bayesian optimization objective function; Step S46: Based on the surrogate model and the preset acquisition function, determine the next set of equipment parameters within the parameter search domain; and based on the static comfort prediction model, predict the comfort of the next set of equipment parameters, add the next set of equipment parameters and the corresponding comfort prediction value to the equipment parameter sample set; and refit the surrogate model and iterate until the first iteration termination condition is met, wherein the first iteration termination condition includes the current iteration number reaching the first preset number.

[0062] Step S47: Select a set of device parameters from the device parameter sample set that satisfies the first constraint condition and minimizes the function value of the Bayesian optimization objective function, as the global comfort baseline parameters. The global comfort baseline parameters are used to characterize the basic operating state of the smart central screen and linked devices that meets the minimum comfort level and does not violate the boundaries of the device parameters.

[0063] It should be noted that from steps S41 to S47, the static comfort prediction model involves complex nonlinear relationships between multidimensional static features and equipment parameters. If a traversal or random search method is used to find parameter combinations that satisfy the constraints, the computational cost is high and it is difficult to guarantee the quality of the search. Therefore, this step introduces a Bayesian optimization framework, using the equipment parameter space as the search domain and the minimum comfort level as a hard constraint, ensuring that the generated baseline settings meet the user's basic comfort requirements without exceeding the limits allowed by the equipment hardware. In the objective function design, the goal is to minimize the excess between the predicted comfort value and the minimum comfort level, and to penalize samples that violate the constraints. This ensures that the optimization process avoids excessive pursuit of high comfort, which could lead to unnecessary energy consumption increases or excessive equipment load, while still satisfying basic comfort requirements, thus obtaining baseline parameters that balance comfort and energy saving. Bayesian optimization uses a surrogate model to fit existing samples and a collection function to achieve a balance between exploring unknown regions and developing known optimal regions. It is suitable for efficiently searching for globally optimal solutions on static comfort prediction models where evaluation costs are high. After iterative optimization, the set of equipment parameters that meet the minimum comfort constraint and have the smallest objective function value is selected from the sample set as the global comfort baseline parameter. This parameter reflects the optimal configuration of the basic operating state under long-term stable conditions, and provides a reliable benchmark and constraint basis for subsequent dynamic adjustment of zones.

[0064] In the above implementation, to generate global comfort baseline parameters, a pre-generated static comfort prediction model is invoked, and the device parameter space of the smart central screen and linked devices, as well as the set minimum comfort level, are obtained. The device parameter space consists of the adjustable parameters of each linked device and their value ranges. Then, the static features of the scene are used as the fixed input to the static comfort prediction model, and the device parameter space is used as the parameter search domain. The first constraint is that the comfort level output by the static comfort prediction model is not lower than the minimum comfort level. Multiple initial device parameter sets are then extracted from the parameter search domain, and the static comfort prediction model is invoked for each set to obtain the corresponding comfort prediction values. A device parameter sample set is constructed based on the initial device parameter sets and the corresponding comfort prediction values. Finally, a Bayesian optimization objective function is constructed, whereby, under the first constraint, the Bayesian optimization objective function aims to minimize the excess between the comfort prediction value and the minimum comfort level. The system identifies and penalizes device parameter samples that do not meet the first constraint. Then, based on the device parameter sample set, a surrogate model of the Bayesian optimization objective function is constructed. Based on the surrogate model and a preset acquisition function, the next set of device parameters is determined within the parameter search domain. Furthermore, based on a static comfort prediction model, comfort prediction is performed on the next set of device parameters, and the next set of device parameters and its corresponding comfort prediction value are added to the device parameter sample set. The surrogate model is refitted and iterated until the first iteration termination condition is met, where the first iteration termination condition includes the current iteration number reaching a first preset number. Then, the device parameter set that satisfies the first constraint and minimizes the function value of the Bayesian optimization objective function is selected from the device parameter sample set as the global comfort baseline parameter. The global comfort baseline parameter characterizes the basic operating state of the smart central screen and linked devices, which meets the minimum comfort level and does not violate the device parameter boundaries.

[0065] As a further implementation of the method, the step of determining parameter constraints based on global comfort baseline parameters and device parameter boundaries includes: Step S51: Obtain the device parameter boundaries of the smart central screen and linked devices, wherein the device parameter boundaries include the upper limit and lower limit of the adjustable parameters.

[0066] Step S52: Analyze the global comfort baseline parameters to obtain the baseline parameter values ​​of the adjustable parameters.

[0067] Step S53: For each adjustable parameter, the difference between the lower limit of the adjustable parameter and the corresponding baseline parameter value is taken as the lower limit of the adjustment amount of the adjustable parameter, and the difference between the upper limit of the adjustable parameter and the corresponding baseline parameter value is taken as the upper limit of the adjustment amount of the adjustable parameter.

[0068] Step S54: Generate parameter constraints based on the lower limit and upper limit of the adjustable parameter adjustment amount.

[0069] It should be noted that the core idea from steps S51 to S54 is to transform the search space for subsequent dynamic adjustments of partitions from the absolute range of device parameters to the range of adjustments relative to the global comfort baseline parameters. The global comfort baseline parameters represent the basic settings that meet minimum comfort while also considering energy consumption under long-term stable conditions. If each device parameter is allowed to be adjusted arbitrarily within its entire range, it can easily lead to excessive local adjustments, damaging the overall comfort baseline or causing abnormal increases in energy consumption. By analyzing the baseline parameter values ​​and using the difference between the device parameter boundary and the baseline value as the upper and lower limits of the adjustment amount for each adjustable parameter, a reasonable allowable deviation can be determined for each parameter, ensuring that dynamic adjustments always revolve around the global comfort baseline. This constraint setting, on the one hand, retains the flexibility of partition optimization in responding to real-time changes, allowing device parameters to be moderately corrected near the baseline according to local needs; on the other hand, it limits the search range to a neighborhood that is permissible by the device hardware and where overall comfort is controllable, avoiding the generation of parameters that exceed the device's capabilities or deviate significantly from the basic comfort level, thereby improving the convergence speed of subsequent optimization algorithms and the engineering feasibility of the generated parameters.

[0070] In the above implementation, in order to generate parameter constraints, the device parameter boundaries of the smart central screen and the linked devices are obtained. The device parameter boundaries include the upper limit and lower limit of the adjustable parameters. Then, the global comfort baseline parameters are parsed to obtain the baseline parameter values ​​of the adjustable parameters. Then, for each adjustable parameter, the difference between the lower limit of the adjustable parameter and the corresponding baseline parameter value is used as the lower limit of the adjustment amount of the adjustable parameter, and the difference between the upper limit of the adjustable parameter and the corresponding baseline parameter value is used as the upper limit of the adjustment amount of the adjustable parameter. Then, parameter constraints are generated based on the lower limit and upper limit of the adjustment amount of the adjustable parameters.

[0071] As a further implementation of the method, the steps of calling a pre-generated dynamic comfort prediction model, using the target comfort as the optimization objective and parameter constraints as constraints, and employing a differential evolution algorithm for reverse optimization based on scene dynamic characteristics to generate partition comfort adjustment parameters include: Step S61: Call the pre-generated dynamic comfort prediction model and obtain the set target comfort level, wherein the target comfort level is not lower than the minimum comfort level.

[0072] Step S62: Based on the parameter constraints, randomly generate multiple partition comfort adjustment parameter sets to form an initial population. Each partition comfort adjustment parameter set includes the adjustment amount of each adjustable parameter.

[0073] Step S63: For each partition comfort adjustment parameter group in the initial population, input the scene dynamic features, global comfort baseline parameters and the partition comfort adjustment parameter group into the dynamic comfort prediction model to obtain the corresponding comfort prediction value.

[0074] Step S64: Construct a differential evolution fitness function, wherein the fitness function aims to minimize the deviation between the predicted comfort value and the target comfort value, and imposes a penalty on the partition comfort adjustment parameter group that does not meet the parameter constraints.

[0075] Step S65: Based on the fitness values ​​of each partition comfort adjustment parameter group in the initial population, perform differential mutation, crossover, and selection operations on the initial population to generate a new generation population.

[0076] Step S66: The new generation population is used as the current population and iterated until the second iteration termination condition is met. The second iteration termination condition includes the current iteration number reaching the second preset number or the function value of the differential evolution fitness function converging.

[0077] Step S67: Select the partition comfort adjustment parameter group with the best fitness value from the final population as the partition comfort adjustment parameter.

[0078] It should be noted that from steps S61 to S67, the core task is to locally correct the partition parameters based on the global comfort baseline parameters and incorporating real-time dynamic features, so that the actual comfort level of each region is as close as possible to the set target comfort level. Since the dynamic comfort prediction model involves complex nonlinear relationships between dynamic features, baseline parameters, and adjustment parameters, and the search space is limited by parameter constraints, it is difficult to quickly obtain a satisfactory solution that meets the constraints under real-time changes using simple rules or single-point search methods. Therefore, this step employs the differential evolution algorithm, utilizing its population parallel search and global optimization capabilities to efficiently explore combinations of adjustment amounts within the feasible domain defined by parameter constraints. In the initialization phase, individuals in the population are randomly generated according to the parameter constraints, ensuring that the initial solutions are all within the feasible adjustment range, thereby reducing invalid searches. The fitness function focuses on minimizing the deviation between the predicted comfort value and the target comfort level, and penalizes individuals that violate the constraints, ensuring that the optimization process, while satisfying the equipment parameter boundaries and baseline deviation limits, prioritizes approaching the target comfort level, balancing comfort and feasibility. Through mutation, crossover, and selection operations, the population gradually converges towards better regions during iterations, ultimately selecting the individual with the best fitness as the partition comfort adjustment parameter. This generated partition adjustment parameter preserves the stability of the global baseline while providing differentiated compensation for the real-time dynamic characteristics of different regions, thereby improving overall control performance and user comfort.

[0079] In the above implementation, to generate partition comfort adjustment parameters, a pre-generated dynamic comfort prediction model is invoked, and a set target comfort level is obtained, wherein the target comfort level is not lower than the minimum comfort level. Then, according to parameter constraints, multiple partition comfort adjustment parameter sets are randomly generated to form an initial population. Each partition comfort adjustment parameter set includes the adjustment amount of each adjustable parameter. Then, for each partition comfort adjustment parameter set in the initial population, the scene dynamic features, global comfort baseline parameters, and the partition comfort adjustment parameter set are input into the dynamic comfort prediction model to obtain the corresponding comfort prediction value. Then, a differential evolution fitness function is constructed, wherein the fitness function is based on the highest comfort level. The goal is to minimize the deviation between the predicted comfort value and the target comfort value. Penalties are imposed on partition comfort adjustment parameter sets that do not meet the parameter constraints. Then, based on the fitness values ​​of each partition comfort adjustment parameter set in the initial population, differential mutation, crossover, and selection operations are performed on the initial population to generate a new generation population. The new generation population is then used as the current population and iterated until the second iteration termination condition is met. The second iteration termination condition includes the current iteration number reaching a second preset number or the function value of the differential evolution fitness function converging. Finally, the partition comfort adjustment parameter set with the best fitness value is selected from the final population as the partition comfort adjustment parameter.

[0080] As a further implementation of the method, the step of fusing the zone comfort adjustment parameters and the global comfort baseline parameters to obtain scene parameters for controlling the smart central screen and linked devices includes: Step S71: Analyze the global comfort baseline parameters to obtain the baseline parameter values ​​of each adjustable parameter.

[0081] Step S72: Analyze the comfort adjustment parameters of each zone to obtain the adjustment amount corresponding to each adjustable parameter.

[0082] Step S73: The baseline parameter values ​​corresponding to each adjustable parameter are superimposed with the corresponding adjustment amounts to obtain the scene parameters.

[0083] Step S74: Based on the scene parameters, generate corresponding device control commands to perform zoned control of the smart central screen and linked devices.

[0084] It should be noted that from steps S71 to S74, the fusion of global comfort baseline parameters and zone comfort adjustment parameters essentially involves superimposing the basic settings under long-term stable conditions with local compensations under real-time dynamic conditions. This results in scenario parameters that possess both overall stability and regional adaptability. The global comfort baseline parameters provide the basic values ​​for each adjustable parameter while meeting minimum comfort and energy consumption requirements. The zone comfort adjustment parameters reflect the correction amounts determined within the parameter constraints to address real-time dynamic characteristics and zone-specific needs. After parameter-by-parameter superposition, the resulting scenario parameters neither deviate excessively from the long-term optimal baseline due to complete reliance on real-time adjustments, nor neglect local needs by using only globally unified settings. Furthermore, since the zone adjustment amounts are optimized within constraints centered on the baseline value and limited by the device parameter boundaries, the superposition result naturally meets the requirements for device feasibility and safe operation, requiring no additional boundary corrections. Finally, device control commands are generated based on the fused scenario parameters. These commands can be directly mapped to the smart central screen and various linked devices, enabling refined control for different zones. This maintains overall comfort while enhancing the comfort experience for individual users.

[0085] In the above implementation, in order to generate scene parameters, the global comfort baseline parameters are parsed to obtain the baseline parameter values ​​of each adjustable parameter. Then, the zone comfort adjustment parameters are parsed to obtain the adjustment amount corresponding to each adjustable parameter. Then, the baseline parameter values ​​corresponding to each adjustable parameter and the corresponding adjustment amounts are superimposed to obtain scene parameters. Then, based on the scene parameters, corresponding device control commands are generated to perform zone control on the smart central screen and linked devices.

[0086] As a further implementation of the method, the step of generating the static comfort prediction model includes: Step S81: Obtain historical static scene data, historical device parameter data, and corresponding comfort labels, wherein the comfort labels are used to characterize the level of environmental comfort.

[0087] Step S82: Extract features from historical static scene data to obtain historical scene static features; and standardize and encode historical equipment parameter data to obtain historical equipment parameter features.

[0088] Step S83: Concatenate the historical scene static features and historical device parameter features corresponding to the same scene to obtain comfort sample features, and associate the comfort sample features with the corresponding comfort labels to construct a comfort sample set.

[0089] Step S84: Divide the comfort sample set into a training set and a validation set.

[0090] Step S85: Based on the training set, train the pre-built neural network model to obtain the trained neural network model.

[0091] Step S86: Based on the validation set, the trained neural network model is validated to obtain a validated neural network model, and the validated neural network model is used as a static comfort prediction model.

[0092] It should be noted that the generation principle of the dynamic comfort prediction model is basically the same as that of the static comfort prediction model. The generation method of the static comfort prediction model can be referred to steps S81 to S86.

[0093] In the above implementation, in order to generate a static comfort prediction model, historical static scene data, historical equipment parameter data, and corresponding comfort labels are obtained, wherein the comfort labels are used to characterize the degree of environmental comfort. Then, feature extraction is performed on the historical static scene data to obtain historical scene static features; and the historical equipment parameter data is standardized and encoded to obtain historical equipment parameter features. Then, the historical scene static features and historical equipment parameter features corresponding to the same scene are concatenated to obtain comfort sample features, and the comfort sample features are associated with the corresponding comfort labels to construct a comfort sample set. Then, the comfort sample set is divided into a training set and a validation set. Then, based on the training set, a pre-constructed neural network model is trained to obtain a trained neural network model. Then, based on the validation set, the trained neural network model is validated to obtain a validated neural network model, and the validated neural network model is used as the static comfort prediction model.

[0094] This application also discloses a scene parameter optimization system for a smart central screen.

[0095] refer to Figure 2 The scene parameter optimization system for the smart central screen includes: The scene data acquisition module is used to acquire static scene data and dynamic scene data of the scene in which the smart screen is located. Static scene data is used to represent the long-term stable state of the scene, while dynamic scene data is used to represent the real-time changes of the scene. The scene feature extraction module is used to extract features from static scene data to obtain static scene features; and to extract features from dynamic scene data to obtain dynamic scene features. The baseline parameter generation module is used to call the pre-generated static comfort prediction model. With the minimum comfort level and equipment parameter space as constraints, it uses the Bayesian algorithm to perform reverse optimization based on the static features of the scene to generate global comfort baseline parameters. The static comfort prediction model is used to output the comfort level based on the static feature data of the scene and the equipment parameter data. The parameter constraint determination module is used to determine parameter constraints based on global comfort baseline parameters and device parameter boundaries, where the device parameter boundaries are the endpoint values ​​of the device parameter space. The parameter generation module is used to call the pre-generated dynamic comfort prediction model, with the target comfort as the optimization objective and the parameter constraints as the constraints. Based on the scene dynamic features, the differential evolution algorithm is used for reverse optimization to generate the partition comfort adjustment parameters. The dynamic comfort prediction model is used to output the comfort level based on the scene dynamic feature data, the global comfort baseline parameter data, and the partition comfort adjustment parameter data. The parameter fusion control module is used to fuse the zone comfort adjustment parameters with the global comfort baseline parameters to obtain scene parameters, so as to perform zone control on the smart central screen and linked devices.

[0096] The scene parameter optimization system for the smart central screen of the present invention can implement any of the methods for optimizing scene parameters of the smart central screen, and the specific working process of the scene parameter optimization system for the smart central screen of the present invention can refer to the corresponding process in the above-mentioned methods for optimizing scene parameters of the smart central screen.

[0097] This application also discloses a computer device.

[0098] refer to Figure 3 A computer device includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement any of the above-mentioned methods for optimizing scene parameters of a smart central screen.

[0099] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A method for optimizing scene parameters of a smart central screen, characterized in that, include: Acquire static and dynamic scene data of the scene in which the smart central screen is located, wherein the static scene data is used to characterize the long-term stable state of the scene, and the dynamic scene data is used to characterize the real-time changes of the scene. Feature extraction is performed on the static scene data to obtain static scene features; and feature extraction is performed on the dynamic scene data to obtain dynamic scene features; A pre-generated static comfort prediction model is invoked. With the minimum comfort level and device parameter space as constraints, a Bayesian algorithm is used to perform reverse optimization based on the static features of the scene to generate global comfort baseline parameters. The static comfort prediction model is used to output comfort level based on the static feature data of the scene and the device parameter data. Based on the global comfort baseline parameters and device parameter boundaries, parameter constraints are determined, wherein the device parameter boundaries are the endpoint values ​​of the device parameter space; The pre-generated dynamic comfort prediction model is invoked, with the target comfort as the optimization objective and the parameter constraints as the constraints. Based on the dynamic characteristics of the scene, the differential evolution algorithm is used for reverse optimization to generate partition comfort adjustment parameters. The dynamic comfort prediction model is used to output comfort based on the scene dynamic feature data, global comfort baseline parameter data and partition comfort adjustment parameter data. The zonal comfort adjustment parameters are fused with the global comfort baseline parameters to obtain scene parameters, which are then used to perform zonal control of the smart central screen and linked devices.

2. The method for optimizing scene parameters of a smart central screen according to claim 1, characterized in that, The step of extracting features from the static scene data to obtain static scene features includes: The static scene data is classified to obtain spatial structure data, equipment configuration data and environmental baseline data. The spatial structure data includes at least one of room layout information, enclosure structure information and space size information. The equipment configuration data includes at least one of equipment type, installation location, rated parameters and controllable range. The environmental baseline data includes historical environmental parameter statistics. Feature extraction is performed on the spatial structure data to obtain spatial layout features; The device configuration data is subjected to feature encoding and attribute mapping to obtain device capability characteristics; The environmental baseline data is subjected to statistical feature construction and dimensionality reduction processing to obtain environmental baseline features; The spatial layout features, the equipment capability features, and the environmental baseline features are fused to obtain the static features of the scene.

3. The method for optimizing scene parameters of a smart central screen according to claim 1, characterized in that, The step of extracting features from the dynamic scene data to obtain dynamic scene features includes: The dynamic scene data is classified to obtain environmental sensing data, personnel activity data, equipment operating status data, and external environment data. The environmental sensing data includes at least one of temperature, humidity, and light intensity. The personnel activity data includes at least one of personnel location, number of personnel, and activity type. The equipment operating status data includes at least one of equipment on / off status, operating mode, load parameters, and energy consumption data. The external environment data includes at least one of outdoor temperature, outdoor light intensity, and weather conditions. Based on a preset time window and sliding step size, the dynamic scene data is sliced ​​to obtain multiple dynamic scene segments, wherein the dynamic scene segments include environmental sensing segments, personnel activity segments, equipment operating status segments, and external environment segments. For each dynamic scene segment, feature extraction is performed on the corresponding environmental sensing segment, personnel activity segment, equipment operating status segment, and external environment segment to obtain the corresponding environmental dynamic features, personnel dynamic features, equipment dynamic features, and external dynamic features. The environmental dynamic features, personnel dynamic features, equipment dynamic features, and external dynamic features within the same dynamic scene segment are spliced ​​together to obtain the segment dynamic features; The dynamic features of each segment are time-series aggregated to obtain the scene dynamic features.

4. The method for optimizing scene parameters of a smart central screen according to claim 1, characterized in that, The step of calling the pre-generated static comfort prediction model, constrained by the minimum comfort level and the device parameter space, and generating global comfort baseline parameters by using a Bayesian algorithm for back-optimization based on the static features of the scenario, includes: The pre-generated static comfort prediction model is invoked, and the device parameter space of the smart central screen and the linked devices and the set minimum comfort level are obtained. The device parameter space consists of the adjustable parameters of each linked device and the value range of the adjustable parameters. The static features of the scene are used as the fixed input of the static comfort prediction model, the device parameter space is used as the parameter search domain, and the first constraint condition is that the comfort level output by the static comfort prediction model is not lower than the minimum comfort level. Multiple initial equipment parameter sets are extracted within the parameter search domain, and the static comfort prediction model is called respectively to obtain the corresponding comfort prediction values. An equipment parameter sample set is constructed based on the initial equipment parameter sets and the corresponding comfort prediction values. Construct a Bayesian optimization objective function, wherein, under the condition of satisfying the first constraint, the Bayesian optimization objective function aims to minimize the excess between the comfort prediction value and the minimum comfort value, and imposes a penalty on the equipment parameter samples that do not satisfy the first constraint. Based on the device parameter sample set, a surrogate model for the Bayesian optimization objective function is constructed; Based on the surrogate model and the preset acquisition function, the next set of equipment parameters is determined within the parameter search domain; and, based on the static comfort prediction model, the comfort of the next set of equipment parameters is predicted, and the next set of equipment parameters and the corresponding comfort prediction value are added to the equipment parameter sample set; and, the surrogate model is refitted and iterated until the first iteration termination condition is met, wherein the first iteration termination condition includes the current iteration number reaching a first preset number; Select a group of device parameters from the device parameter sample set that satisfies the first constraint and minimizes the function value of the Bayesian optimization objective function, and use them as global comfort baseline parameters. The global comfort baseline parameters are used to characterize the basic operating state of the smart central screen and linked devices that meets the minimum comfort level and does not violate the device parameter boundaries.

5. The method for optimizing scene parameters of a smart central screen according to claim 1, characterized in that, The step of determining parameter constraints based on the global comfort baseline parameters and device parameter boundaries includes: Obtain the device parameter boundaries of the smart central screen and its linked devices, wherein the device parameter boundaries include the upper limit and lower limit of adjustable parameters; The global comfort baseline parameters are analyzed to obtain the baseline parameter values ​​of the adjustable parameters; For each adjustable parameter, the difference between the lower limit of the adjustable parameter and the corresponding baseline parameter value is taken as the lower limit of the adjustment amount of the adjustable parameter, and the difference between the upper limit of the adjustable parameter and the corresponding baseline parameter value is taken as the upper limit of the adjustment amount of the adjustable parameter. Parameter constraints are generated based on the lower limit and the upper limit of the adjustment amount of the adjustable parameter.

6. The method for optimizing scene parameters of a smart central screen according to claim 1, characterized in that, The step of calling the pre-generated dynamic comfort prediction model, using the target comfort as the optimization objective and parameter constraints as constraints, and generating partition comfort adjustment parameters by using the differential evolution algorithm for reverse optimization based on the dynamic characteristics of the scene, includes: Invoke the pre-generated dynamic comfort prediction model and obtain the set target comfort level, wherein the target comfort level is not lower than the minimum comfort level; Based on the parameter constraints, multiple partition comfort adjustment parameter sets are randomly generated to form an initial population, wherein each partition comfort adjustment parameter set includes the adjustment amount of each adjustable parameter; For each partition comfort adjustment parameter group in the initial population, the scene dynamic features, the global comfort baseline parameters, and the partition comfort adjustment parameter group are input into the dynamic comfort prediction model to obtain the corresponding comfort prediction value. Construct a differential evolution fitness function, wherein the fitness function aims to minimize the deviation between the predicted comfort value and the target comfort value, and imposes a penalty on the partition comfort adjustment parameter group that does not meet the parameter constraints; Based on the fitness values ​​of each partition comfort adjustment parameter group in the initial population, differential mutation, crossover and selection operations are performed on the initial population to generate a new generation population. The new generation population is used as the current population and iterated until the second iteration termination condition is met. The second iteration termination condition includes the current iteration number reaching the second preset number or the function value of the differential evolution fitness function converging. The optimal set of partition comfort adjustment parameters with the best fitness value is selected from the final population and used as the partition comfort adjustment parameters.

7. The method for optimizing scene parameters of a smart central screen according to claim 1, characterized in that, The step of fusing the zone comfort adjustment parameters with the global comfort baseline parameters to obtain scene parameters for controlling the smart central screen and linked devices includes: The global comfort baseline parameters are analyzed to obtain the baseline parameter values ​​of each adjustable parameter; The comfort adjustment parameters of the partition are analyzed to obtain the adjustment amount corresponding to each adjustable parameter; The baseline parameter value corresponding to each adjustable parameter is superimposed with the corresponding adjustment amount to obtain the scene parameters; Based on the scenario parameters, corresponding device control commands are generated to perform zoned control of the smart central screen and linked devices.

8. The method for optimizing scene parameters of a smart central screen according to claim 1, characterized in that, The steps for generating the static comfort prediction model include: Acquire historical static scene data, historical device parameter data, and corresponding comfort labels, wherein the comfort labels are used to characterize the degree of environmental comfort; Feature extraction is performed on the historical static scene data to obtain historical scene static features; and the historical device parameter data is standardized and encoded to obtain historical device parameter features. The static features of the historical scene and the parameters of the historical equipment corresponding to the same scene are concatenated to obtain comfort sample features, and the comfort sample features are associated with the corresponding comfort labels to construct a comfort sample set; The comfort sample set is divided into a training set and a validation set; Based on the training set, the pre-built neural network model is trained to obtain the trained neural network model; Based on the validation set, the trained neural network model is validated to obtain a validated neural network model, which is then used as a static comfort prediction model.

9. A scene parameter optimization system for a smart central screen, characterized in that, include: The scene data acquisition module is used to acquire static scene data and dynamic scene data of the scene in which the smart central screen is located. The static scene data is used to characterize the long-term stable state of the scene, and the dynamic scene data is used to characterize the real-time changes of the scene. The scene feature extraction module is used to extract features from the static scene data to obtain static scene features; and to extract features from the dynamic scene data to obtain dynamic scene features. The baseline parameter generation module is used to call a pre-generated static comfort prediction model, which, constrained by the minimum comfort level and the device parameter space, uses a Bayesian algorithm to perform reverse optimization based on the static features of the scene to generate global comfort baseline parameters. The static comfort prediction model is used to output comfort based on the static feature data of the scene and the device parameter data. The parameter constraint determination module is used to determine parameter constraint conditions based on the global comfort baseline parameters and the device parameter boundaries, wherein the device parameter boundaries are the endpoint values ​​of the device parameter space; The parameter generation module is used to call a pre-generated dynamic comfort prediction model, with the target comfort as the optimization objective and the parameter constraints as the constraints. Based on the dynamic characteristics of the scene, the differential evolution algorithm is used to perform reverse optimization to generate partition comfort adjustment parameters. The dynamic comfort prediction model is used to output comfort based on the scene dynamic feature data, global comfort baseline parameter data and partition comfort adjustment parameter data. The parameter fusion control module is used to fuse the zone comfort adjustment parameters with the global comfort baseline parameters to obtain scene parameters for zone control of the smart central screen and linked devices.

10. A computer device, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the method of any one of claims 1 to 8.

Images