Building office interval illumination self-adaptive intelligent control method and system

Abstract

The invention provides a building office interval illumination adaptive intelligent control method and system, and relates to the technical field of building office interval illumination adaptive intelligent control. The method comprises the following steps: collecting personnel density and illuminance data of each office function area; then, based on the personnel density, a plurality of BP neural networks optimized by a particle swarm optimization algorithm are adopted for parallel prediction, and a plurality of initial illuminance values of each region are obtained; performing prediction average voting processing on the initial values of the same area to determine target illuminance, and calculating a deviation value between the target illuminance and actually measured illuminance; and finally, the deviation is adjusted and processed by using fuzzy PID, and a brightness adjusting signal is generated and issued to a corresponding area driving module, so that illumination adaptive control is realized, and on-demand, accurate and adaptive adjustment of the illumination of the office interval of the building can be realized through combination of multi-neural network prediction and fuzzy PID control.

Description

Technical Field

[0001] This application relates to the technical field of adaptive intelligent control of lighting in office buildings, and in particular to a method and system for adaptive intelligent control of lighting in office buildings. Background Technology

[0002] Office buildings typically include open-plan offices, private offices, and public corridors. The amount of natural light entering these areas fluctuates significantly with seasons, time of day, and weather. Office activities also involve diverse scenarios such as fixed work hours, temporary overtime, and unexpected meetings. Under different scenarios and lighting conditions, office workers have varying needs for stable indoor lighting environments. For example, artificial lighting needs to be reduced when natural light is abundant at midday, while uniform and comfortable lighting is required when working late at night. To adapt to the dynamic changes in natural light and diverse office scenarios, an intelligent control method is needed that can coordinate lighting with natural light and the office environment to ensure the stability of the lighting environment while reducing ineffective energy consumption.

[0003] Currently, in the field of lighting control in office buildings, existing technologies deploy light sensors, clock modules, and human presence sensors in various areas. Light sensors collect real-time indoor and outdoor illuminance data, clock modules obtain current time information, and human presence sensors determine whether there is human activity in the area. The collected data is transmitted to the control unit, which has a built-in preset "time period-illuminance threshold" rule library. For example, on weekdays from 9:00 to 18:00, the indoor illuminance is maintained at 400 lux, and then reduced to 300 lux after 18:00. Based on the matching results between the rule library and the sensor data, a lighting control signal is generated using relay switches or basic PWM dimming technology, and transmitted to the lighting equipment driver module to complete the brightness adjustment.

[0004] However, the most significant drawback of existing technologies lies in their insufficient flexibility and environmental adaptability. On the one hand, they rely on a fixed set of preset rules, and when faced with sudden changes in natural light, they must wait for the next sensor data acquisition cycle before adjustments can be triggered, resulting in a noticeable response lag. On the other hand, the preset scene modes cannot cover non-fixed office scenarios such as temporary overtime work or impromptu multi-person meetings, easily leading to situations where the lighting is too strong or too weak. Furthermore, lighting control in each area is mostly independent, lacking a cross-area collaborative adjustment mechanism. For example, when a meeting room is temporarily activated, the lighting in adjacent corridors cannot be synchronously adapted to the guiding brightness, making it difficult to meet the dynamic changing needs of office scenarios. Summary of the Invention

[0005] The purpose of this application is to provide an adaptive intelligent control method and system for lighting in office building areas, in order to solve the problem in the prior art that it is impossible to accurately match the real-time lighting needs caused by dynamic changes in personnel density.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides an adaptive intelligent control method for lighting in building office areas, comprising:

[0007] Collect personnel density and illuminance data for each functional area of ​​the building's office space;

[0008] Based on the personnel density data, multiple BP neural networks optimized by particle swarm optimization algorithm are used to predict the lighting demand of each functional area in the building office area, and multiple initial illuminance values ​​for each functional area are obtained.

[0009] A predictive average voting process is performed on all the initial illuminance values ​​of the same functional area to obtain the target illuminance value of the functional area. The difference between the target illuminance value and the illuminance data is calculated to obtain the illuminance deviation value.

[0010] The illuminance deviation value is processed by fuzzy PID control to generate a lighting brightness adjustment signal for the corresponding functional area. The lighting brightness adjustment signal is then transmitted to the lighting equipment driver module of the corresponding functional area to complete the adaptive control of the lighting brightness.

[0011] Optionally, based on the personnel density data, multiple backpropagation neural networks optimized by particle swarm optimization are used to predict the lighting demand of each functional area in the building's office space, obtaining multiple initial illuminance values ​​for each functional area, including:

[0012] Obtain the lighting requirements baseline for each functional area;

[0013] A lighting equipment deployment model is constructed. Based on the lighting equipment deployment model and light intensity distribution characteristics, the light contribution weight of each lighting equipment to each location in the functional area is calculated to obtain the equipment light influence model.

[0014] Based on the personnel density data and lighting demand benchmarks of each functional area, and combined with the pre-built equipment lighting impact model, multiple BP neural networks optimized by the particle swarm optimization algorithm are used to predict the lighting demand of each functional area in the building office area, and obtain multiple initial illuminance values ​​for each functional area.

[0015] Optionally, based on the personnel density data and lighting demand benchmarks of each functional area, and combined with a pre-built equipment lighting impact model, multiple BP neural networks optimized by particle swarm optimization are used to predict the lighting demand of each functional area in the building's office space, obtaining multiple initial illuminance values ​​for each functional area, including:

[0016] The corresponding features are extracted from the personnel density data, the lighting demand benchmark, and the equipment lighting impact model, respectively.

[0017] According to the preset combination rules, all features are combined to form the neural network input feature vector.

[0018] The input feature vector of the neural network is input into multiple parallel BP neural networks for forward propagation processing to predict the lighting demand of each functional area in the building office space. The initial illuminance value of each functional area is obtained from the output of each BP neural network. The network parameters of each BP neural network are pre-optimized by the particle swarm optimization algorithm. During the optimization process, the comfort of the people in the functional area is taken as the optimization goal. The fitness function designed based on the optimization goal includes a prediction accuracy term and an energy consumption index term, and the parameters of different BP neural networks are different.

[0019] Optionally, the step of performing a predicted average voting process on all the initial illuminance values ​​of the same functional area to obtain the target illuminance value of the functional area includes:

[0020] A confidence assessment is performed on all initial illuminance values ​​of the same functional area to obtain a confidence weight corresponding to each initial illuminance value. The confidence weight is dynamically adjusted based on the historical prediction accuracy of the corresponding BP neural network during the confidence assessment process.

[0021] Based on the confidence weights, calculate the weighted average illuminance value of the corresponding functional area;

[0022] Based on the weighted average illuminance value, the dispersion evaluation process is performed on all the initial illuminance values ​​corresponding to the functional area to identify the abnormal illuminance values ​​of the corresponding functional area.

[0023] Calculate the arithmetic mean of all normal initial illuminance values ​​in the functional area after removing the abnormal illuminance values, and use the arithmetic mean as the target illuminance value for the corresponding functional area.

[0024] Optionally, the step of performing a dispersion assessment on all the initial illuminance values ​​corresponding to the functional area based on the weighted average illuminance value to identify abnormal illuminance values ​​in the corresponding functional area includes:

[0025] The window size is adaptively adjusted based on the rate of change in personnel density in functional areas to construct a dynamic evaluation window;

[0026] Based on the dynamic evaluation window, a relative stability analysis method is used to fuse the weighted average illuminance value and the initial illuminance value to obtain the stability coefficient corresponding to each initial illuminance value.

[0027] Based on the lighting requirements benchmark of the functional area and the personnel density data, the boundary conditions for anomaly detection of the functional area are determined.

[0028] Based on the stability coefficient and the anomaly determination boundary conditions, an anomaly marking algorithm is used to filter the initial illuminance values ​​of the anomalies to obtain the abnormal illuminance values ​​of the functional areas.

[0029] Optionally, the step of performing fuzzy PID adjustment processing on the illuminance deviation value to generate a lighting brightness adjustment signal for the corresponding functional area, and transmitting the lighting brightness adjustment signal to the lighting equipment driver module of the corresponding functional area to complete the adaptive control of lighting brightness, includes:

[0030] Based on the preset lighting level classification rules, the illuminance deviation value is fuzzified to obtain the level distribution. The level distribution is then matched with the preset lighting adjustment rules to obtain the adjustment amount of each lighting parameter.

[0031] The adjustment amounts of each lighting parameter are precisely converted to obtain precise adjustment values ​​for each parameter. Based on these precise adjustment values, a dynamic compensation algorithm is used to perform control quantity compensation calculations to generate control signals for the lighting equipment.

[0032] The control signal of the lighting equipment is processed by brightness adjustment encoding to obtain the lighting brightness adjustment signal for each functional area;

[0033] The lighting brightness adjustment signal is transmitted to the lighting device driver module corresponding to the functional area to complete the adaptive control of lighting brightness.

[0034] Optionally, the step of performing precise conversion processing on the adjustment amounts of each lighting parameter to obtain precise adjustment values ​​for each parameter, and based on the precise adjustment values, using a dynamic compensation algorithm to perform control quantity compensation calculations to generate a lighting equipment control signal, includes:

[0035] The centroid method is used to defuzzify the lighting parameters and perform precise conversion to obtain the precise adjustment values ​​corresponding to the proportional coefficient, integral time and derivative coefficient.

[0036] The precise adjustment value is fused with the preset benchmark PID parameters to obtain a set of dynamic control parameters;

[0037] Based on the aforementioned set of dynamic control parameters, a dynamic compensation algorithm is used to perform numerical compensation of the control quantity to obtain the initial control quantity;

[0038] The initial control quantity is subjected to output limiting processing to obtain the lighting equipment control signal.

[0039] Secondly, this application provides an adaptive intelligent control system for lighting in office building areas, comprising:

[0040] The data acquisition module is used to collect personnel density data and illuminance data in various functional areas of the building's office space;

[0041] The prediction module is used to predict the lighting demand of each functional area of ​​the building office space based on the personnel density data and through multiple BP neural networks optimized by the particle swarm optimization algorithm, so as to obtain multiple initial illuminance values ​​for each functional area.

[0042] The voting module is used to perform predictive averaging voting on all the initial illuminance values ​​of the same functional area to obtain the target illuminance value of the functional area, and to calculate the difference between the target illuminance value and the illuminance data to obtain the illuminance deviation value.

[0043] The adjustment module is used to perform fuzzy PID adjustment processing on the illuminance deviation value, generate the lighting brightness adjustment signal of the corresponding functional area, and transmit the lighting brightness adjustment signal to the lighting equipment drive module of the corresponding functional area to complete the adaptive control of lighting brightness.

[0044] Thirdly, this application provides an electronic device, comprising:

[0045] Memory, used to store computer programs;

[0046] A processor is configured to execute the computer program to implement the steps of the adaptive intelligent control method for building office space lighting as described in the first aspect above.

[0047] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the adaptive intelligent control method for building office space lighting as described in the first aspect above.

[0048] The adaptive intelligent control method for lighting in office building areas provided in this application collects personnel density data and illuminance data for each functional area of ​​the office building area; based on the personnel density data, it uses multiple BP neural networks optimized by particle swarm optimization algorithm to perform illuminance demand prediction processing for each functional area of ​​the office building area, obtaining multiple initial illuminance values ​​for each functional area; it performs prediction average voting processing on all the initial illuminance values ​​of the same functional area to obtain the target illuminance value of the functional area, calculates the difference between the target illuminance value and the illuminance data, and obtains the illuminance deviation value; it performs fuzzy PID adjustment processing on the illuminance deviation value to generate a lighting brightness adjustment signal for the corresponding functional area, and transmits the lighting brightness adjustment signal to the lighting equipment driver module of the corresponding functional area to complete the adaptive control of lighting brightness.

[0049] The technical solution of this application has the following beneficial effects:

[0050] This application obtains core foundational data directly related to the lighting needs of office areas, ensuring that subsequent lighting demand prediction, deviation calculation, and lighting adjustment are based on actual environmental conditions. This avoids blind control detached from the actual scenario and provides real and reliable data support for the entire adaptive control process. The particle swarm optimization algorithm improves the problem of BP neural networks easily getting trapped in local optima and lacking convergence, enhancing the stability of network predictions. Simultaneously, parallel prediction by multiple BP neural networks generates multiple sets of initial illuminance values, avoiding the limitations of single-model prediction, enriching the diversity of lighting demand prediction results, and improving the reliability and generalization ability of lighting demand prediction. The prediction averaging voting process can be used to process multiple sets of initial illuminance values. By integrating and filtering data, the error impact of a single prediction result is reduced, making the target illuminance value more closely match the actual lighting needs of the functional area. The calculation of the illuminance deviation value clearly quantifies the gap between the current lighting state and the target requirement, providing a clear direction and basis for subsequent precise lighting adjustment. By using fuzzy PID control, the adaptability of fuzzy control to complex scenarios and the precision of PID control are balanced. The adjustment strategy can be dynamically adjusted for different illuminance deviation values, avoiding overshoot or response lag. By executing the adjustment signal through the drive module, the control strategy can be directly converted into brightness changes of the lighting equipment, realizing closed-loop control from deviation analysis to actual adjustment, and ultimately achieving adaptive matching of office area lighting.

[0051] Furthermore, this application obtains the lighting demand benchmark for each functional area, then constructs a lighting equipment deployment model. Based on the lighting equipment deployment model and light intensity distribution characteristics, it calculates the lighting contribution weight of each lighting equipment to each location within the functional area, thus obtaining the equipment lighting impact model. Finally, based on the personnel density data and lighting demand benchmark for each functional area, combined with the pre-constructed equipment lighting impact model, it uses multiple BP neural networks optimized by the particle swarm optimization algorithm to perform lighting demand prediction processing on each functional area, obtaining multiple initial illuminance values ​​for each functional area.

[0052] This application introduces a lighting demand benchmark, providing a basic reference for lighting demand prediction in each functional area, thus avoiding prediction results that deviate from the inherent lighting adaptation characteristics of the area. The construction of the equipment lighting impact model combines the actual deployment location of lighting equipment with the light intensity distribution pattern, ensuring that the prediction process fully considers the actual differences in the contribution of different lighting equipment to the regional lighting, rather than calculating the overall lighting demand of the area in a general way. Finally, by combining personnel density data, lighting demand benchmark, and equipment lighting impact model for prediction, the matching degree between multiple initial illuminance values ​​and the actual lighting demand of functional areas is further improved, making the initial illuminance values ​​more scenario-specific and practically instructive, laying a better foundation for the determination of subsequent target illuminance values ​​and the improvement of lighting adjustment accuracy.

[0053] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 A flowchart illustrating an adaptive intelligent control method for lighting in office spaces within a building, as provided in this application embodiment;

[0056] Figure 2 This application provides a schematic diagram illustrating a specific implementation of an adaptive intelligent control method for lighting in office building areas.

[0057] Figure 3 A schematic diagram of a building office lighting adaptive intelligent control system provided in this application embodiment;

[0058] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0059] Current lighting control methods in office buildings largely rely on pre-set fixed rules, which are difficult to adapt to dynamic changes in the office environment. On the one hand, when natural light fluctuates suddenly, existing methods require waiting for the next data collection cycle to trigger adjustments, resulting in significant response lag and causing indoor lighting to be too dim or too bright for a period of time. On the other hand, for unplanned office scenarios such as temporary overtime work or multiple people using meeting rooms temporarily, existing methods cannot flexibly adjust lighting, easily leading to lighting that does not meet actual needs and affecting office comfort. In addition, the lighting control of each area is relatively independent, lacking coordinated adaptation to changes in the needs of different areas caused by personnel movement, further reducing the flexibility of overall lighting control.

[0060] To address the issues of lag and poor scene adaptability in existing lighting control methods, this application provides an adaptive intelligent lighting control method for office areas in buildings. This method first collects personnel density data and current illuminance data for each functional office area. Then, based on this data, multiple backpropagation (BP) neural networks optimized by particle swarm optimization (PSO) predict the lighting demand for each area, obtaining multiple sets of initial illuminance values. Subsequently, the multiple sets of initial illuminance values ​​are averaged and voted on to determine a target illuminance value that better reflects the actual needs, and the difference between the target illuminance and the actual illuminance is calculated. Finally, a control signal is generated using fuzzy PID control technology to drive the lighting equipment in the corresponding area to adjust its brightness. In this method, the optimized BP neural networks can capture the dynamic changes in personnel density and illuminance in real time, avoiding the lag problem of fixed rules; the integration of multiple prediction results makes the target illuminance more accurate and adaptable to diverse scenarios such as temporary overtime work and sudden increases in personnel; and the fuzzy PID control technology can quickly adjust the brightness according to the illuminance difference, ensuring timely and stable adjustment, thus effectively solving the shortcomings of existing technologies.

[0061] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0062] The core of this application is to provide an adaptive intelligent control method for lighting in office building areas, and a flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes:

[0063] S101. Collect personnel density data and illuminance data for each functional area of ​​the building's office space.

[0064] In the above scheme, the functional areas of the building office space refer to the subdivided areas within the building used for office activities, including common areas such as open office areas, corridors, and meeting room front areas, which are the specific scope of action of adaptive lighting control; personnel density data refers to data reflecting the relationship between the number of people in each functional area and the area area, including real-time personnel number, area area and other related information, used to determine the intensity of light demand of people in that area; illuminance data refers to data reflecting the brightness of the environment in each functional area, including real-time ambient light intensity, light distribution uniformity and other information, used to determine whether the current lighting meets the needs of personnel.

[0065] In this application example, firstly, the deployment scheme of data acquisition equipment is determined according to the type of each functional area in the office building. Among them, the personnel density data acquisition adopts an infrared array human presence detection module, and the illuminance data acquisition adopts a diffuse reflection light acquisition probe. Based on the spatial scale and layout characteristics of each functional area, the coverage range and deployment location of each group of equipment are determined to ensure that the acquisition range can completely cover the corresponding functional area.

[0066] Next, after equipment deployment, the data acquisition devices are activated for data collection and transmission. The infrared array-type human presence detection module receives infrared radiation signals from the human body within the functional area in real time, captures the activity status of people in the area, and generates raw electrical signals containing information on the location and number of people. The diffuse reflection light acquisition probe receives ambient light diffuse reflection signals within the functional area in real time, detects the level of illumination within the area, and generates raw electrical signals containing light intensity information. After generating the raw electrical signals, both types of devices transmit the signals to a preset data processing unit. The data processing unit performs signal conversion processing on the raw electrical signals, converting the raw electrical signals reflecting human information into human density data that can be directly used for subsequent processing, and converting the raw electrical signals reflecting light information into illuminance data that can be directly used for subsequent processing, thus completing the entire data acquisition process.

[0067] In practical applications, a building office area is used as the implementation scenario. This office area includes three functional areas: open office area A1, corridor A2, and meeting room front area A3. Open office area A1 has an area of ​​200㎡, corridor A2 is 90m long, and meeting room front area A3 has an area of ​​40㎡. When deploying data acquisition equipment, the number of devices of a single type is calculated using the following formula: ,in, This indicates the number of single-type data acquisition devices deployed (unit: groups). Indicates the size parameter of the functional area. This indicates the coverage parameters of a single set of acquisition devices. This represents the floor function. Based on the above formula, calculate the number of devices deployed in each area: For open office area A1, S = 200㎡, C = 100㎡ / group, then N = =2 groups; S=90m, C=50m / group for corridor A2, then N= =2 groups; Area A3 in front of the conference room has S=40㎡, C=100㎡ / group, then N= =1 group, and complete the deployment of all infrared array human presence detection modules and diffuse reflection light acquisition probes according to the calculation results.

[0068] Next, the equipment was activated for data acquisition: the infrared array human presence detection module detected 10 people in open office area A1, 3 people in corridor A2, and 2 people in the front area of ​​meeting room A3; the diffuse reflection light acquisition probe detected an illumination intensity of 280 lux in open office area A1, 220 lux in corridor A2, and 250 lux in the front area of ​​meeting room A3. The personnel density in each area was then calculated using the formula: ,in, Indicates population density (unit: people / m²). This indicates the real-time number of people detected by the infrared array human presence detection module (unit: people). The area of ​​the functional area is represented by m² (for corridors, the average width is the product of the width and length; here, the average width of corridor A2 is 2m, and the area is 90m² × 2m = 180m²). The population density of each area is calculated using the above formula: D = 10 people / 200m² = 0.05 people / m² for A1; D = 3 people / 180m² ≈ 0.017 people / m² for A2; D = 2 people / 40m² = 0.05 people / m² for A3. The final population density data obtained in this step are: A1: 0.05 people / m², A2: 0.017 people / m², A3: 0.05 people / m², and illuminance data are: A1: 280 lux, A2: 220 lux, A3: 250 lux. These will serve as the core input data for the subsequent step S102, "predicting illuminance demand based on population density data using multiple BP neural networks optimized by particle swarm optimization algorithm," providing fundamental data support for subsequent illuminance demand prediction.

[0069] The aforementioned S101 overall solution, by strategically deploying data acquisition devices in each functional area, achieves real-time and comprehensive collection of personnel density and illuminance data. The acquired data accurately reflects the personnel distribution and current lighting conditions in each functional area, avoiding the problem of subsequent processing being out of sync with the actual scene due to incomplete or delayed data collection. At the same time, the process of converting raw electrical signals into usable data provides a standardized basic input for subsequent steps such as predicting lighting demand based on personnel density data and calculating illuminance deviation values. This ensures that the entire adaptive lighting control process can be carried out in an orderly manner based on real and reliable data, laying a data foundation for achieving precise lighting adjustment.

[0070] S102. Based on the personnel density data, multiple BP neural networks optimized by particle swarm optimization algorithm are used to predict the lighting demand of each functional area in the building office area, and multiple initial illuminance values ​​for each functional area are obtained.

[0071] Optionally, in step S102, based on the personnel density data, multiple backpropagation neural networks optimized by particle swarm optimization are used to predict the lighting demand of each functional area in the building's office space, obtaining multiple initial illuminance values ​​for each functional area, including:

[0072] Step 1021: Obtain the lighting requirements baseline for each functional area.

[0073] Step 1022: Construct a lighting equipment deployment model. Based on the lighting equipment deployment model and light intensity distribution characteristics, calculate the light contribution weight of each lighting equipment to each location within the functional area to obtain the equipment light influence model.

[0074] Step 1023: Based on the personnel density data and lighting demand benchmark of each functional area, and combined with the pre-built equipment lighting impact model, multiple BP neural networks optimized by the particle swarm optimization algorithm are used to perform lighting demand prediction processing on each functional area of ​​the building office area, and obtain multiple initial illuminance values ​​for each functional area.

[0075] Step 1023 may specifically include the following steps: extracting corresponding features from the personnel density data, the lighting demand benchmark, and the equipment lighting impact model respectively; performing feature combination processing on all features according to preset combination rules to form a neural network input feature vector; inputting the neural network input feature vector into multiple parallel BP neural networks for forward propagation processing to predict the lighting demand of each functional area in the building office area, obtaining the initial illuminance value of each functional area output by each BP neural network, wherein the network parameters of each BP neural network are pre-optimized by the particle swarm optimization algorithm. During the optimization process, the comfort of personnel in the functional area is taken as the optimization objective. The fitness function designed based on the optimization objective includes a prediction accuracy term and an energy consumption index term, and the parameters of different BP neural networks are different. The aforementioned preset combination rules are designed to systematically integrate features extracted from personnel density data, lighting demand benchmarks, and equipment lighting impact models into an input vector recognizable by the neural network. Specifically, these rules include: extracting real-time personnel density values ​​from personnel density data; extracting minimum illuminance thresholds and suitable illuminance upper limits from lighting demand benchmarks; and extracting the lighting contribution weights of each lighting device from the equipment lighting impact model. The extracted features are arranged in a fixed order: "personnel density features → lighting demand benchmark features → equipment lighting impact features," ensuring structural consistency of the input vector. All extracted features are normalized, mapping values ​​to the 0-1 range to avoid affecting the neural network's learning effect due to differences in feature magnitude. If the number of devices in different functional areas differs, zeros are padded to ensure consistent dimensions of the equipment lighting contribution weight features across all areas, guaranteeing a uniform input vector length.

[0076] In the above scheme, the lighting demand benchmark refers to the basic lighting intensity reference value determined based on the usage scenarios of each functional area in the building's office space, industry lighting comfort standards, and historical usage data. This includes the minimum lighting intensity required by people in the area during regular activities and the suitable lighting intensity range, providing a basic reference for subsequent lighting demand prediction. The lighting equipment deployment model is a digital model reflecting the installation location, quantity, model, and installation height of lighting equipment in each functional area, used to simulate the distribution of lighting equipment within the area. The equipment lighting impact model is a model built based on the lighting equipment deployment model and light intensity distribution characteristics. The lighting contribution weight refers to the proportion of influence of a single lighting device on the lighting intensity at a certain location within the area, used to quantify the actual effect of different devices on the area's lighting. The particle swarm optimization algorithm optimizes multiple B... A P-type neural network (BP neural network) refers to a network that optimizes the initial weights and thresholds of multiple BP neural networks using a particle swarm optimization algorithm. This results in higher prediction accuracy and faster convergence. Multiple BP neural networks working in parallel can generate multiple sets of prediction results, avoiding the limitations of a single model. The neural network input feature vector is a vector data formed by combining key information extracted from personnel density data, lighting demand benchmarks, and equipment lighting impact models according to preset rules. This vector is used as the input to the BP neural network. The fitness function is a function used in the particle swarm optimization algorithm to evaluate the performance of the BP neural network parameters. This embodiment does not specify the expression of this function. The prediction accuracy term measures the degree of agreement between the neural network's lighting demand prediction results and actual demand. The energy consumption index term measures the energy consumption level of the lighting equipment corresponding to the prediction results.

[0077] In this application example, firstly, the lighting requirements baseline for each functional area is obtained through step 1021. This process requires reference to national office lighting design standards, combined with the core purpose of each functional area, such as open office areas for long-term office work, corridors for personnel passage, and meeting room front areas for temporary waiting. Simultaneously, lighting usage feedback data for each area over the past three months is retrieved, such as personnel satisfaction records regarding lighting levels. This information is then used to determine the basic range of lighting requirements for each functional area, clarifying the minimum and suitable intensity ranges for different areas, providing a basic reference for subsequent lighting requirement prediction.

[0078] Next, after obtaining the lighting demand baseline, a lighting equipment deployment model is constructed in step 1022 to obtain the equipment lighting impact model. First, on-site surveys are conducted for each functional area to record the area's dimensions, internal layout, and the installation location, quantity, model, and installation height of the lighting equipment. Based on these measured data, a digital lighting equipment deployment model is constructed, clearly showing the distribution of lighting equipment within the area. Then, combining the light intensity distribution characteristics of the lighting equipment, such as the beam angle and the attenuation law of luminous flux with distance, the light intensity superposition analysis method is used to calculate the proportion of influence of a single lighting equipment on the light intensity at different locations within the area, i.e., the light contribution weight. These weights are combined with the lighting equipment deployment model to form an equipment lighting impact model that can quantify the lighting effect of the equipment.

[0079] Finally, after obtaining the lighting demand baseline and the equipment lighting impact model, lighting demand prediction is completed in step 1023, and multiple initial illuminance values ​​are obtained. First, real-time personnel density values ​​are extracted from the personnel density data collected in S101. A suitable lighting intensity range is extracted from the lighting demand baseline obtained in 1021. The lighting contribution weights of each lighting device are extracted from the equipment lighting impact model constructed in 1022. Following a preset combination rule of "personnel density - baseline range - contribution weight," these extracted key information are integrated to form a neural network input feature vector. This feature vector is then input to multiple parallel BP neural networks. The initial weights and thresholds of these BP neural networks have been pre-optimized using a particle swarm optimization algorithm. During the optimization process, the comfort of personnel in the functional area is the core objective. A fitness function containing prediction accuracy and energy consumption indicators is designed, and iterative adjustments are made to make the neural network's prediction results more closely match actual needs and energy consumption control requirements. Finally, each BP neural network, through forward propagation, passes the input feature vector layer by layer and undergoes nonlinear transformation to output the corresponding functional area's lighting demand prediction value, i.e., each functional area obtains multiple sets of initial illuminance values.

[0080] In practical application, the open office area A1, corridor A2, and meeting room front area A3 in the office building were taken as the implementation targets. First, referring to the requirement of 300 lux for minimum illumination in office areas in the "Standard for Lighting Design of Buildings", and considering the needs of A1 for long-term computer and paper document use, A2 for personnel passage only, and A3 for temporary waiting, the illumination feedback data of A1, A2 and A3 over the past three months were retrieved. It was found that 90% of the personnel in A1 felt comfortable at around 350 lux, 85% of the personnel in A2 felt suitable at around 250 lux, and 88% of the personnel in A3 felt appropriate at around 300 lux. Subsequently, the illumination requirements of A1 were determined to be 300-400 lux, A2 200-300 lux, and A3 250-350 lux.

[0081] Next, a field survey was conducted, revealing that A1 measures 20m × 15m with no internal obstructions. Six LED lights will be installed, each with a beam angle of 120° and a rated luminous flux of 3500lm. These lights will be evenly distributed across the ceiling at 6m intervals and a height of 3m. Based on this, a lighting deployment model for A1 will be constructed. Considering the principle that luminous flux decreases by 10% for every 1m increase in LED light distance, the impact of a single light on its position within the area will be calculated using a light contribution weighting formula. The formula is as follows: ,in, This represents the weight of the illumination contribution of the i-th LED light to the j-th position within A1. This represents the actual luminous intensity of the i-th LED at position j. The calculation method is to multiply the rated luminous flux by the distance attenuation coefficient. For example, if a workstation in A1 is 4m away from the first lamp, then... =3500lm × (1 - 4 × 10%) = 2100lm, where n represents the number of LEDs affecting the j-th position. Here, the workstation is affected by 3 lights, so n = 3. The contribution weights of the 1st, 2nd, and 3rd lights at this workstation are calculated to be 0.42, 0.33, and 0.25, respectively. Similarly, the weights of all positions in A1 are calculated to form the equipment lighting impact model of A1. The lighting equipment deployment model and equipment lighting impact model of A2 and A3 are constructed in the same way. The size of A2 is 50m × 2m, and it is equipped with 8 C-type LEDs. The size of A3 is 8m × 6m, and it is equipped with 2 C-type LEDs.

[0082] Then, density values ​​are extracted from the personnel densities of A1 (0.04 people / m²), A2 (0.015 people / m²), and A3 (0.06 people / m²) obtained from S101. Range values ​​of A1 (300-400 lux), A2 (200-300 lux), and A3 (250-350 lux) are extracted from the baseline determined by 1021. Lighting contribution weights for each region are extracted from the equipment lighting influence model. These are combined according to the rule of "density value - baseline range - contribution weight" to form an input feature vector, such as the vector for A1 being [0.04, 300, 400, 0.42, 0.33, 0.25]. This vector is then input into three parallel BP neural networks. The parameters of these networks are optimized using a particle swarm optimization algorithm, with the optimization using a fitness function. Where F is the fitness value, For example, F is a weighting coefficient, which can be set to 0.6 to balance prediction and energy consumption. P is the prediction accuracy, which is the ratio of the predicted value to the actual comfortable lighting. E is the relative energy consumption, which is the ratio of the energy consumption of the device corresponding to the predicted lighting to the rated energy consumption. F is stabilized above 0.88 through 60 iterations. Finally, the three BP neural networks output the initial illuminance values ​​of A1 (335 lux, 350 lux, 342 lux), A2 (225 lux, 238 lux, 230 lux), and A3 (265 lux, 278 lux, 270 lux). These initial illuminance values ​​will be used for the prediction average voting process in the subsequent S103 step to determine the target illuminance value for each area.

[0083] The aforementioned S102 overall solution provides a basic reference for illumination prediction that conforms to regional characteristics by obtaining a baseline of illumination demand, thus avoiding prediction results that deviate from the actual usage scenario. By constructing a lighting equipment deployment model and an equipment illumination impact model, it quantifies the actual impact of different lighting equipment on regional illumination, making the prediction process more consistent with the actual distribution of equipment and light intensity characteristics. The prediction is performed using multiple BP neural networks optimized by the particle swarm optimization algorithm, which not only solves the problem that a single BP neural network is prone to getting trapped in local optima and insufficient prediction accuracy, but also enriches the reference dimensions of illumination demand through multiple sets of prediction results, ensuring that the prediction results can meet both the comfort needs of personnel and the rationality of lighting energy consumption. Overall, it provides accurate and diverse initial data support for subsequent steps to determine the target illuminance value, effectively improving the scientificity and adaptability of the entire adaptive lighting control process.

[0084] S103. Perform prediction average voting processing on all the initial illuminance values ​​of the same functional area to obtain the target illuminance value of the functional area, calculate the difference between the target illuminance value and the illuminance data, and obtain the illuminance deviation value.

[0085] Optionally, step S103, performing a prediction average voting process on all the initial illuminance values ​​of the same functional area to obtain the target illuminance value of the functional area, includes:

[0086] Step 1031: Perform confidence assessment on all initial illuminance values ​​of the same functional area to obtain the confidence weight corresponding to each initial illuminance value. The confidence weight is dynamically adjusted based on the historical prediction accuracy of the corresponding BP neural network during the confidence assessment process.

[0087] Step 1032: Based on the confidence weight, calculate the weighted average illuminance value of the corresponding functional area.

[0088] Step 1033: Based on the weighted average illuminance value, perform dispersion evaluation processing on all the initial illuminance values ​​corresponding to the functional area to identify abnormal illuminance values ​​of the corresponding functional area.

[0089] Step 1033 may specifically include the following steps: adaptively adjusting the window size based on the rate of change of personnel density in the functional area to construct a dynamic evaluation window; based on the dynamic evaluation window, using a relative stability analysis method to perform a fusion analysis on the weighted average illuminance value and the initial illuminance value to obtain a stability coefficient corresponding to each initial illuminance value; determining the anomaly judgment boundary conditions of the functional area based on the illuminance demand benchmark of the functional area and the personnel density data; and using an anomaly labeling algorithm to filter out abnormal initial illuminance values ​​according to the stability coefficient and the anomaly judgment boundary conditions to obtain abnormal illuminance values ​​of the functional area.

[0090] The relative stability analysis method is based on a dynamic evaluation window. It assesses the stability of initial illuminance values ​​by quantifying the relative deviation between the initial illuminance value and the weighted average illuminance value, and the fluctuation of this deviation within the dynamic evaluation window. The specific algorithm is as follows: First, using the weighted average illuminance value as a benchmark, calculate the relative deviation of each initial illuminance value from the benchmark value; the calculation formula is... ,in, The relative deviation of the i-th initial illuminance value. Let i be the initial illuminance value. The initial illuminance value is weighted and averaged. Then, historical relative deviation data corresponding to this initial illuminance value are extracted within the dynamic evaluation window, and the standard deviation of the historical deviation is calculated to reflect the degree of deviation fluctuation. Finally, the formula is used to calculate the weighted average illuminance value. Calculate the stability coefficient, where, Let be the stability coefficient for the i-th initial illuminance value. is the standard deviation of the historical relative bias, and After normalization to the 0-1 range, the closer the coefficient is to 1, the smaller the deviation between the initial illuminance value and the baseline value, and the more stable the fluctuation. The closer it is to 0, the larger the deviation and the more unstable the fluctuation. The anomaly marking algorithm is based on the preset anomaly judgment boundary conditions and the stability coefficient of the initial illuminance value. It is an algorithm that filters out abnormal initial illuminance values ​​that do not meet the actual lighting requirements of the functional area. The specific algorithm is as follows: First, each initial illuminance value is compared with the illuminance range boundary in the anomaly judgment boundary conditions, and the initial value that exceeds the illuminance range is marked. Then, the stability coefficient corresponding to each initial illuminance value is compared with the stability coefficient threshold in the anomaly judgment boundary conditions, and the initial value with the coefficient lower than the threshold is marked. Finally, a logical "OR" operation is performed on the two marking results. That is, the initial illuminance value that satisfies either "exceeds the illuminance range" or "stability coefficient is lower than the threshold" is judged as an abnormal illuminance value, thus completing the screening of anomalies.

[0091] Step 1034: Calculate the arithmetic mean of all normal initial illuminance values ​​in the functional area after removing the abnormal illuminance values, and use the arithmetic mean as the target illuminance value for the corresponding functional area.

[0092] In the above scheme, the prediction average voting process refers to the process of calculating the average value after reliability assessment and anomaly screening of multiple initial illuminance values ​​for the same functional area to determine the target illuminance value, which is used to integrate multiple sets of prediction results and improve the accuracy of the target value; the confidence weight refers to the weight value reflecting the reliability of each initial illuminance value, which is dynamically adjusted based on the historical prediction accuracy of the BP neural network outputting the initial value, with a larger weight for higher accuracy; the weighted average illuminance value refers to the average value obtained by multiplying each initial illuminance value by its corresponding confidence weight and summing the results, which is used to initially integrate multiple sets of prediction results; the dispersion assessment process refers to the process of identifying and eliminating abnormal initial illuminance values ​​by analyzing the difference between the initial illuminance value and the weighted average illuminance value and combining it with the actual needs of the area; the dynamic assessment window refers to the adaptive adjustment based on the rate of change of personnel density in the functional area. The resizing analysis window shrinks when the rate of change is high to focus on real-time data, and expands when the rate of change is low to reference more historical data. The stability coefficient measures the stability of the difference between a single initial illuminance value and the weighted average illuminance value. The smaller the difference and the smaller the fluctuation, the closer the coefficient is to 1, and vice versa. The anomaly judgment boundary condition is a judgment standard determined based on the functional area's lighting demand benchmark and personnel density data, used to define whether the initial illuminance value is abnormal. The target illuminance value is the arithmetic mean of the remaining normal initial illuminance values ​​after removing abnormal initial illuminance values, representing the actual lighting intensity required for the area. The illuminance deviation value is the difference between the target illuminance value and the real-time illuminance data collected by S101, used to quantify the gap between the current lighting status and the target demand, providing a basis for subsequent lighting adjustments.

[0093] In this application example, firstly, confidence assessment is performed on all initial illuminance values ​​of the same functional area through step 1031. This process requires retrieving the historical prediction accuracy data of the BP neural network outputting each initial illuminance value over the past month, and using a normalization algorithm to convert the historical prediction accuracy into confidence weights, ensuring that the sum of the confidence weights corresponding to all initial illuminance values ​​is 1, thereby clarifying the reliability of different initial illuminance values.

[0094] Next, after obtaining the confidence weights, the weighted average illuminance value of the corresponding functional area is calculated based on the confidence weights using 1032. Each initial illuminance value is multiplied by its corresponding confidence weight, and then all the product results are summed to obtain the weighted average illuminance value of the functional area, thus achieving the initial integration of multiple sets of initial illuminance values.

[0095] Then, after obtaining the weighted average illuminance value, step 1033 is used to perform dispersion assessment on all initial illuminance values ​​of the functional area. The first step involves adaptively adjusting the size of the dynamic assessment window based on the rate of change in population density within the functional area: first, the rate of change in population density within the functional area is calculated per unit time. If the rate of change is higher than a preset threshold, the size of the dynamic assessment window is reduced to focus on real-time data; if the rate of change is lower than the preset threshold, the size of the dynamic assessment window is expanded to reference more historical data. The second step uses a relative stability analysis method to calculate the stability coefficient corresponding to each initial illuminance value: using the weighted average illuminance value as a benchmark, the relative deviation between each initial illuminance value and the benchmark value is calculated, and then the fluctuation of this relative deviation within the dynamic assessment window is considered. The first step is to obtain the stability coefficient of each initial illuminance value. The second step is to determine the anomaly judgment boundary conditions based on the illuminance demand benchmark and personnel density data of the functional area: according to the minimum and suitable illuminance range of the illuminance demand benchmark, combined with the impact of the current personnel density on the illuminance demand, the normal range of the initial illuminance value and the qualified threshold of the stability coefficient are set, which together constitute the anomaly judgment boundary conditions. The third step is to use an anomaly marking algorithm to screen abnormal illuminance values: each initial illuminance value and its stability coefficient are compared with the anomaly judgment boundary conditions. If the initial illuminance value exceeds the normal range or the stability coefficient is lower than the qualified threshold, it is judged as an abnormal illuminance value.

[0096] Finally, after identifying abnormal illuminance values, the target illuminance value for the functional area is calculated in step 1034. First, all initial illuminance values ​​deemed abnormal within the functional area are removed. Then, the remaining normal initial illuminance values ​​are averaged, and the resulting arithmetic mean is used as the target illuminance value for that functional area. After determining the target illuminance value, the illuminance deviation value is calculated: the target illuminance value of the functional area is compared with the real-time illuminance data collected in step S101 to obtain the illuminance deviation value for that functional area, providing a basis for subsequent lighting adjustments.

[0097] In practical applications, taking the open office area A1, corridor A2, and meeting room front area A3 of a building as the implementation targets, the initial illuminance values ​​obtained by A1 from step S102 are 335 lux, 350 lux, and 342 lux, respectively. The historical prediction accuracy of the BP neural network for these three initial illuminance values ​​is 928,890. The initial illuminance values ​​obtained by A2 from step S102 are 225 lux, 238 lux, and 230 lux, respectively. The corresponding historical prediction accuracy of the BP neural network is 899,187. The initial illuminance values ​​obtained by A3 from step S102 are 265 lux, 278 lux, and 270 lux, respectively. The corresponding historical prediction accuracy of the BP neural network is 938,690.

[0098] First, the confidence weights are calculated using the formula... ,in, This represents the confidence weight corresponding to the i-th initial illuminance value. This represents the historical prediction accuracy of the BP neural network that outputs the i-th initial illuminance value. The sum of the historical prediction accuracies of the BP neural network for all output initial illuminance values, where n represents the number of initial illuminance values, yields:

[0099] A1 ,but , , ;

[0100] A2 ,but , , ;

[0101] A3 ,but , , .

[0102] Next, the weighted average illuminance value is calculated using the formula... ,in, This represents the weighted average illuminance value. Representing the i-th initial illuminance value, we get:

[0103] A1 =335×0.34+350×0.32+342×0.34=113.9+112+116.28=342.18lux;

[0104] A2 =225×0.33+238×0.34+230×0.33=74.25+80.92+75.9=231.07lux;

[0105] A3 =265×0.35+278×0.32+270×0.33=92.75+88.96+89.1=270.81lux.

[0106] Next, abnormal illuminance values ​​are identified. The first step is to calculate the rate of change in population density, using the following formula: Where r represents the rate of change in population density, Indicates the current population density. This indicates the population density 5 minutes ago. (A1) =0.042 people / m² =0.04 people / m², then r = (0.04² - 0.04) / 0.04 = 0.05; A2 =0.016 people / ㎡ =0.015 people / m², then r = (0.016 - 0.015) / 0.015 ≈ 0.067; A3 =0.063 people / m² =0.06 people / m², then r = (0.063 - 0.06) / 0.06 = 0.05. The preset threshold for the rate of change in population density is 0.2. Since the rates of change for all three factors are below the threshold, the dynamic evaluation window size is set to 5 data sets. The second step is to calculate the stability coefficient using the formula... ,in, The stability coefficient for the i-th initial illuminance value is obtained as follows:

[0107] A1 =1-|335-342.18| / 342.18≈0.98, =1-|350-342.18| / 342.18≈0.97, =1-|342-342.18| / 342.18≈0.99;

[0108] A2 =1-|225-231.07| / 231.07≈0.97, =1-|238-231.07| / 231.07≈0.97, =1-|230-231.07| / 231.07≈0.99;

[0109] A3 =1-|265-270.81| / 270.81≈0.98, =1-|278-270.81| / 270.81≈0.97, =1-|270-270.81| / 270.81≈0.99;

[0110] The third step is to determine the boundary conditions for anomaly detection: Based on the A1 light requirement baseline of 300-400 lux, the A2 light requirement baseline of 200-300 lux, and the A3 light requirement baseline of 250-350 lux, along with the current population density, the boundary conditions for anomaly detection are set as follows: the initial illuminance value is either 20 lux lower than the lower limit of the baseline or 20 lux higher than the upper limit of the baseline, and the stability coefficient is lower than 0.9. The fourth step is to screen for outliers. The initial illuminance values ​​in all three areas are within the normal range, and the stability coefficient is higher than 0.9; therefore, there are no abnormal illuminance values.

[0111] Finally, the target illuminance value is calculated using the arithmetic mean formula. ,in, This represents the target illuminance value, and m represents the number of normal initial illuminance values. A1's... =(335+350+342) / 3=342.33lux; A2's =(225+238+230) / 3=231lux; A3's =(265+278+270) / 3=271 lux; Then calculate the illuminance deviation value, the formula for calculating the deviation value is... ,in, Indicates the illuminance deviation value. This represents the real-time illuminance data collected in step S101. A1 =280 lux, then =342.33-280=62.33lux; A2's =220 lux, then =231-220=11 lux; A3's =250 lux, then =271-250=21 lux. The illuminance deviation values ​​A1 (62.33 lux), A2 (11 lux), and A3 (21 lux) obtained in this step will be used as input data for the fuzzy PID adjustment process in the subsequent S104 step to generate the lighting brightness adjustment signal corresponding to each functional area, thereby realizing adaptive control of lighting brightness.

[0112] The aforementioned S103 overall scheme, through confidence assessment and weighting of each initial illuminance value by combining the historical prediction accuracy of the BP neural network, fully considers the reliability differences of different prediction results and avoids the accumulation of errors caused by treating all initial values ​​"equally." The combination of weighted averaging and dispersion assessment not only initially integrates multiple sets of prediction results, but also effectively eliminates possible abnormal initial values ​​through dynamic window adjustment, stability analysis, and anomaly screening, ensuring that the target illuminance value calculated subsequently is more in line with the actual lighting needs of the functional area. The final calculated illuminance deviation value clearly quantifies the gap between the current lighting and the target requirement, providing a clear and accurate basis for subsequent lighting adjustment, and comprehensively improving the scientificity and reliability of the "target determination" link in adaptive lighting control, laying the foundation for achieving precise lighting adjustment.

[0113] S104. Perform fuzzy PID adjustment processing on the illuminance deviation value to generate a lighting brightness adjustment signal for the corresponding functional area, and transmit the lighting brightness adjustment signal to the lighting equipment drive module of the corresponding functional area to complete the adaptive control of lighting brightness.

[0114] Optionally, step S104 involves performing fuzzy PID adjustment processing on the illuminance deviation value to generate a lighting brightness adjustment signal for the corresponding functional area, and transmitting the lighting brightness adjustment signal to the lighting equipment drive module of the corresponding functional area to complete adaptive control of the lighting brightness, including:

[0115] Step 1041: Based on the preset lighting level classification rules, the illuminance deviation value is fuzzified to obtain the level distribution. The level distribution is then matched with the preset lighting adjustment rules to obtain the adjustment amount of each lighting parameter.

[0116] The lighting level classification rule is a pre-set rule designed to simplify the fuzzy processing of illuminance deviation values ​​and clarify the approximate range of deviations. Its core is to divide illuminance deviation values ​​into several fixed levels based on the adjustment capabilities of lighting equipment in office building areas and the sensitivity of functional areas to lighting needs. Each level corresponds to a specific deviation value range. This rule allows for quick determination of the current illuminance deviation range, providing a basis for subsequent matching of lighting adjustment rules and determination of parameter adjustment direction. The lighting adjustment rule is a pre-set "deviation level - parameter adjustment amount" correspondence rule designed to quickly determine the adjustment range of core lighting control parameters. Its core is based on the degree of influence of different deviation levels on lighting comfort and the effect of PID control parameters on lighting adjustment. It matches a corresponding parameter adjustment range for each deviation level. This rule directly converts the distribution of illuminance deviation levels into specific parameter adjustment amounts, bridging the fuzzy processing and subsequent precise adjustment, ensuring that parameter adjustments meet the actual needs of the current deviation state.

[0117] Step 1042: Perform precise conversion processing on the adjustment amounts of each lighting parameter to obtain precise adjustment values ​​for each parameter. Specifically, the lighting parameter adjustment amounts, which previously obtained through fuzzy matching and possess range attributes, are converted into single, specific, and precise values ​​using a specific algorithm. This process first clarifies the fuzzy set corresponding to each parameter adjustment amount, then calculates the centroid position of the fuzzy set to ultimately obtain precise values ​​that can be directly used for PID control parameter adjustment. This solves the problem that fuzzy adjustment amounts cannot be directly used as control parameter inputs, providing accurate parameter basis for subsequent dynamic compensation calculations and lighting equipment control signal generation. Based on the precise parameter adjustment values, a dynamic compensation algorithm is used to perform control quantity compensation calculations to generate lighting equipment control signals.

[0118] Step 1042 may specifically include the following steps: using the centroid method to defuzzify the lighting parameters, the adjustment values ​​of each parameter are precisely converted to obtain the precise adjustment values ​​corresponding to the proportional coefficient, integral time, and derivative coefficient; the precise adjustment values ​​are fused with preset benchmark PID parameters to obtain a set of dynamic control parameters, that is, the benchmark PID parameters adapted to the conventional characteristics and functional area requirements of the lighting equipment are used as the base values, and then linear operations are performed on the precise adjustment values ​​for the current illuminance deviation: the proportional coefficient adopts the method of "benchmark Kp and precise adjustment value", increasing the coefficient to speed up the response speed to the current deviation; the integral time adopts the method of "benchmark Ti - precise adjustment value", decreasing the time to enhance the integral action and accelerate the deviation convergence; the derivative coefficient adopts the method of "benchmark Td and precise adjustment value", increasing the coefficient to strengthen the derivative action and reduce overshoot during brightness adjustment; based on the set of dynamic control parameters, a dynamic compensation algorithm is used to perform numerical compensation of the control quantity to obtain the initial control quantity; the initial control quantity is output-limited to obtain the lighting equipment control signal.

[0119] Step 1043: Perform brightness adjustment encoding processing on the lighting equipment control signal to obtain the lighting brightness adjustment signal for each functional area.

[0120] Step 1044: Transmit the lighting brightness adjustment signal to the lighting device driver module corresponding to the functional area to complete the adaptive control of lighting brightness.

[0121] In the above scheme, fuzzy PID control refers to a comprehensive control method combining fuzzy control and PID control, used to balance the adaptability and accuracy of lighting adjustment; lighting level classification rules refer to preset rules that divide illuminance deviation values ​​into different levels according to their magnitude, used to simplify the fuzzification of deviation values; level distribution refers to the distribution of illuminance deviation values ​​corresponding to their respective lighting levels, reflecting the approximate range of the deviation; lighting adjustment rules refer to preset "deviation level - lighting parameter adjustment amount" correspondence rules, used to determine the adjustment range of proportional coefficient, integral time, and derivative coefficient based on the deviation level; the adjustment amount of each lighting parameter refers to the adjustment range of proportional coefficient, integral time, and derivative coefficient obtained based on the lighting adjustment rules, providing a basis for subsequent precise adjustment; the centroid method for defuzzification refers to a method that converts fuzzy adjustment amounts into precise values ​​by calculating the centroid position of the fuzzy set, ensuring the accuracy of the adjustment parameters; proportional coefficient, integral time, and derivative coefficient are the core parameters of PID control, used to adjust the response speed of lighting adjustment, eliminate steady-state deviation, and suppress interference fluctuations, respectively; the reference PID parameters are determined according to the characteristics of the lighting equipment and the functional area requirements. The system defines preset PID base parameters to provide a reference for dynamic parameter adjustment; the dynamic control parameter set refers to the combination of PID parameters obtained by fusing precise adjustment values ​​with baseline PID parameters to adapt to the current deviation state; the dynamic compensation algorithm is an algorithm used to compensate for control deviations caused by environmental changes, ensuring the accuracy of the control quantity; the initial control quantity is the preliminary control value obtained after dynamic compensation algorithm calculation, reflecting the basic adjustment range required for the lighting equipment; output limiting processing is the process of limiting the initial control quantity within the allowable adjustment range of the lighting equipment to avoid exceeding the equipment load; the lighting equipment control signal is the numerical signal that can be directly used to control the lighting equipment after output limiting processing; brightness adjustment encoding processing is the process of converting the lighting equipment control signal into an encoding format that the lighting equipment driver module can recognize, ensuring that the signal can be received by the equipment; the lighting brightness adjustment signal is the signal obtained after encoding processing that the driver module can directly respond to, used to clarify the direction and magnitude of lighting brightness adjustment; the lighting equipment driver module is the hardware module that receives the lighting brightness adjustment signal and drives the lighting equipment to adjust the brightness, and is the core component connecting the control signal and the equipment action.

[0122] In this application example, firstly, the illuminance deviation value is fuzzified and matched with lighting adjustment rules in step 1041. This process first calls a preset lighting level classification rule to divide the illuminance deviation value into different levels according to its magnitude, obtaining a level distribution that reflects the approximate range of the deviation; then, it calls a preset lighting adjustment rule, which is a correspondence rule of "deviation level - lighting parameter adjustment amount", and matches the level distribution with the lighting adjustment rule to determine the adjustment range of the proportional coefficient, integral time, and derivative coefficient, thereby obtaining the adjustment amount of each lighting parameter.

[0123] Next, after obtaining the adjustment amounts of each lighting parameter, step 1042 converts the adjustment amounts into lighting equipment control signals. The first step uses a centroid-based defuzzification method for precise conversion. Based on the fuzzy sets corresponding to the adjustment amounts of each lighting parameter, the centroid position of the fuzzy sets is calculated, converting the fuzzy adjustment amounts into precise adjustment values ​​corresponding to the proportional coefficient, integral time, and derivative coefficient. The second step calls a preset reference PID parameter, which is a basic parameter set according to the characteristics of the lighting equipment and the functional area requirements. The precise adjustment values ​​are fused with the reference PID parameter according to preset rules to obtain a dynamic control parameter set adapted to the current deviation state. The third step uses a dynamic compensation algorithm to identify the impact of current environmental changes on control, calculates the compensation amount, and combines it with the initial control amplitude corresponding to the dynamic control parameter set to obtain the initial control amount. The fourth step, based on the allowable adjustment range of the lighting equipment, performs output limiting processing on the initial control amount to prevent the control amount from exceeding the equipment load, thus obtaining the lighting equipment control signal.

[0124] Then, after acquiring the lighting equipment control signal, the lighting equipment control signal is encoded for brightness adjustment in step 1043. Using the encoding protocol supported by the lighting equipment driver module, the lighting equipment control signal is converted into an encoding format recognizable by the driver module, obtaining the lighting brightness adjustment signals for each functional area, ensuring that the signals can be correctly received by the driver module.

[0125] Finally, adaptive control of lighting brightness is completed in step 1044. Using wired or wireless communication, the lighting brightness adjustment signals for each functional area are transmitted to the corresponding lighting equipment driver module. After receiving the signals, the driver module drives the lighting equipment to adjust its brightness according to the signal instructions, making the illuminance of the functional area approach the target illuminance value, thus completing the entire adaptive lighting control process.

[0126] In practical applications, the open office area A1, corridor A2, and meeting room front area A3 of the building are taken as the implementation objects. The illuminance deviation value obtained by A1 from S103 is 62.33 lux, A2 is 11 lux, and A3 is 21 lux.

[0127] First, the preset lighting level classification rule is that a deviation value of 0-10 lux is a small deviation, 11-30 lux is a medium deviation, and 31 lux and above is a large deviation. Based on this, deviation value A1 is classified as a large deviation, and A2 and A3 are classified as medium deviations. The preset lighting adjustment rule is that a large deviation corresponds to a proportional coefficient adjustment of 0.6-1.0, an integral time adjustment of 5-8s, and a derivative coefficient adjustment of 0.3-0.5; a medium deviation corresponds to a proportional coefficient adjustment of 0.3-0.6, an integral time adjustment of 3-5s, and a derivative coefficient adjustment of 0.15-0.3. Based on this, A1 has a proportional coefficient adjustment of 0.8, an integral time adjustment of 6s, and a derivative coefficient adjustment of 0.4; A2 has a proportional coefficient adjustment of 0.3, an integral time adjustment of 3s, and a derivative coefficient adjustment of 0.15; and A3 has a proportional coefficient adjustment of 0.5, an integral time adjustment of 4s, and a derivative coefficient adjustment of 0.25.

[0128] Next, the centroid method for fuzzy resolution is used to calculate the precise adjustment value. The formula for the centroid method is as follows: Where u is the precise adjustment value, For elements in a fuzzy set, Let n be the membership degree of an element and n be the number of elements. The fuzzy set of A1 proportional coefficient adjustment values ​​is 0.6, 0.7, 0.8, 0.9, 1.0, corresponding to membership degrees of 0.1, 0.2, 0.4, 0.2, 0.1. The calculated value is u = (0.6 × 0.1 + 0.7 × 0.2 + 0.8 × 0.4 + 0.9 × 0.2 + 1.0 × 0.1) ÷ (0.1 + 0.2 + 0.4 + 0.2 + 0.1) = 0.8, which is the precise adjustment value of the A1 proportional coefficient. The value is 0.8. Similarly, the precise adjustment value of the integral time of A1 is calculated to be 6s, and the precise adjustment value of the derivative coefficient is 0.4. The preset reference PID parameters are a proportional coefficient of 2.0, an integral time of 10s, and a derivative coefficient of 0.5. According to the rule of "dynamic parameter = reference parameter + precise adjustment value", the dynamic control parameter set of A1 is a proportional coefficient of 2.8, an integral time of 16s, and a derivative coefficient of 0.9. Then, a dynamic compensation algorithm is used to identify that the natural light of A1 is reduced by 5 lux compared with 5 minutes ago. The preset compensation coefficient is 0.02, and the compensation amount = 5 × 0.02 = 0.1. The initial control amount calculated based on the dynamic parameters is 65, and the initial control amount after compensation is 65.1. The preset maximum control amount of A1 lighting equipment is 70, and the minimum control amount is 5. Output limiting processing is performed on 65.1 to obtain the lighting equipment control signal of A1, which is 65.1. And according to the same method, the precise adjustment value of the proportional coefficient of A2 is calculated to be 0.3, the integral time is 3s, and the derivative coefficient is 0.15. The dynamic control parameter set is a proportional coefficient of 0.3. With a coefficient of 2.3, an integral time of 13s, a derivative coefficient of 0.65, a compensation of 0.08, an initial control quantity of 18.12, and an output limiting control signal of 18.2, A3 calculates a proportional coefficient precise adjustment value of 0.5, an integral time of 4s, a derivative coefficient of 0.25, a dynamic control parameter set of a proportional coefficient of 2.5, an integral time of 14s, a derivative coefficient of 0.75, a compensation of 0.1, an initial control quantity of 32.4, and an output limiting control signal of 32.5.

[0129] Then, using the RS485 encoding protocol, the lighting control signal 65.1 of A1 is encoded into binary code 01100101 to obtain the lighting brightness adjustment signal of A1; the lighting control signal 18.2 of A2 is encoded into binary code 00010010 to obtain the lighting brightness adjustment signal of A2; the lighting control signal 32.5 of A3 is encoded into binary code 00100000 to obtain the lighting brightness adjustment signal of A3.

[0130] Finally, via Ethernet communication, the lighting brightness adjustment signals for A1, A2, and A3 are transmitted to the LED driver modules in areas A1, A2, and A3, respectively. Upon receiving the signals, the LED driver module for A1 adjusts its brightness from approximately 280 lux to approximately 342 lux; the driver module for A2 adjusts its brightness from approximately 220 lux to approximately 231 lux; and the driver module for A3 adjusts its brightness from approximately 250 lux to approximately 271 lux, ensuring that the illuminance in each area reaches the target value, thus completing the adaptive control of lighting brightness. The lighting brightness adjustment signals generated in this step are directly used to drive the modules, ultimately achieving precise adjustment of lighting in each area without requiring subsequent data retrieval.

[0131] The aforementioned S104 overall solution, through fuzzy PID control, balances the flexible classification of illuminance deviation values ​​by fuzzy control with the precise adjustment capability of PID control, avoiding the problems of insufficient adaptability or low adjustment accuracy of single control methods in complex office environments. The centroid method defuzzification process converts fuzzy adjustment values ​​into precise values, and combined with dynamic compensation algorithms and output limiting processing, ensures the accuracy of lighting equipment control signals and equipment compatibility, avoiding adjustment failures caused by environmental changes or control values ​​exceeding the equipment range. The brightness adjustment encoding processing and signal transmission steps ensure that the adjustment signals can be correctly identified and executed by the lighting equipment drive module, realizing a closed loop from control decision to equipment action. Through the coordination of each sub-step, the overall process converts illuminance deviation values ​​into actual brightness adjustments of lighting equipment, accurately matching the target lighting needs of each functional area, and ultimately achieving adaptive control of lighting brightness in office building areas, ensuring the lighting comfort of office workers.

[0132] The following is a complete example for steps S101 to S104, such as Figure 2 As shown, the implementation scenarios are the open office area A1, corridor A2, and the front area of ​​the meeting room A3 in the building. A1 has an area of ​​200㎡ and detected 8 people in real time. A2 has a length of 50m, an average width of 2m (area of ​​100㎡), and detected 1.5 people in real time. A3 has an area of ​​40㎡ and detected 2.4 people in real time.

[0133] First, data collection is performed, and personnel density is calculated using the formula D=P / A, where D represents personnel density (persons / m²), P represents the real-time number of people (persons), and A represents the area of ​​the functional area (m²). The personnel density for A1 is calculated as 8 people ÷ 200m² = 0.04 people / m², for A2 it's 1.5 people ÷ 100m² = 0.015 people / m², and for A3 it's 2.4 people ÷ 40m² = 0.06 people / m². Simultaneously, real-time illuminance is collected for each area: 280 lux for A1, 220 lux for A2, and 250 lux for A3. This personnel density and real-time illuminance data will be used for subsequent lighting demand prediction.

[0134] Next, the light demand prediction step is performed, using three BP neural networks optimized by particle swarm optimization in parallel prediction. The historical prediction accuracies of the three networks are 92%, 88%, and 90%, respectively. Then, the confidence weights of each initial illuminance value are calculated using the formula... ,in, This represents the confidence weight of the i-th initial illuminance value. This represents the historical prediction accuracy of the BP neural network that outputs the i-th initial value. From k=1 to n Let represent the sum of the historical prediction accuracies of the three networks, and n represent the number of initial illuminance values ​​(3 in this case). The sum of the accuracies for A1 is 92 + 88 + 90 = 270. Therefore, the weights corresponding to the three initial illuminance values ​​of 335 lux, 350 lux, and 342 lux are approximately 0.34 (92 ÷ 270 ≈ 0.34), 0.32 (88 ÷ 270 ≈ 0.32), and 0.34 (90 ÷ 270 ≈ 0.34), respectively. Then, the weighted average illuminance value is calculated using the formula... ,in, This represents the weighted average illuminance value, in lux. Let A1 represent the i-th initial illuminance value in lux. The weighted average of A1 is 335×0.34+350×0.32+342×0.34=113.9+112+116.28=342.18lux. Then, the stability coefficient is calculated using the formula... ,in, As the stability coefficient, substituting the data, the three coefficients of A1 are 0.98, 0.97, and 0.99, all higher than the threshold of 0.9, with no outliers. The target illuminance value is then calculated using the arithmetic mean formula. ,in, The target illuminance value is represented as follows: A1 is (335 + 350 + 342) ÷ 3 = 342.33 lux; similarly, the target value of A2 is calculated from the three initial values ​​of 225 lux, 238 lux, and 230 lux to be 231 lux, and the target value of A3 is calculated from 265 lux, 278 lux, and 270 lux to be 271 lux; then the illuminance deviation value is calculated using the formula... ,in, This represents the illuminance deviation value, in lux. The values ​​represent real-time illuminance in lux. The deviation values ​​for A1 are 342.33-280=62.33 lux, for A2 they are 231-220=11 lux, and for A3 they are 271-250=21 lux. These deviation values ​​will be used for subsequent fuzzy PID control.

[0135] Finally, fuzzy PID control is executed. The preset lighting level classification rule is that a deviation of 0-10 lux is a small deviation, 11-30 lux is a medium deviation, and 31 lux and above is a large deviation. Therefore, A1's 62.33 lux is a large deviation, and A2 and A3's 11 lux and 21 lux are medium deviations. The preset lighting adjustment rule is that a large deviation corresponds to a proportional coefficient adjustment of 0.6-1.0, an integral time adjustment of 5-8s, and a derivative coefficient adjustment of 0.3-0.5, while a medium deviation corresponds to 0.3-0.6, 3-5s, and 0.15-0.3. Accordingly, A1 is matched with a proportional coefficient adjustment of 0.8, an integral time adjustment of 6s, and a derivative coefficient adjustment of 0.4; A2 is matched with 0.3, 3s, and 0.15; and A3 is matched with 0.5, 4s, and 0.25. Subsequently, the centroid method is used to defuzzify and calculate the precise adjustment value. The formula is as follows: Where u represents the precise adjustment value, Represents the elements in a fuzzy set. Let A1 represent the membership degree of the element. The fuzzy set of the proportional coefficient adjustment is 0.6, 0.7, 0.8, 0.9, 1.0, and the membership degree is 0.1, 0.2, 0.4, 0.2, 0.1. The calculated value is u = (0.6 × 0.1 + 0.7 × 0.2 + 0.8 × 0.4 + 0.9 × 0.2 + 1.0 × 0.1) ÷ (0.1 + 0.2 + 0.4 + 0.2 + 0.1) = 0.8, which is the proportional coefficient. The precise adjustment value of the coefficient is 0.8. Similarly, the integral time is 6s and the derivative coefficient is 0.4. The preset baseline PID parameters are a proportional coefficient of 2.0, an integral time of 10s, and a derivative coefficient of 0.5. The fusion method is dynamic parameter = baseline parameter + precise adjustment value. Therefore, the dynamic proportional coefficient of A1 is 2.0 + 0.8 = 2.8, the dynamic integral time is 10s + 6s = 16s, and the dynamic derivative coefficient is 0.5 + 0.4 = 0.9. A dynamic compensation algorithm is adopted, and it is identified that the current natural light in A1 is 5 lux weaker than 5 minutes ago. The compensation coefficient is preset to 0.02, and the compensation amount is 5 × 0.02 = 0.1. The initial control amount calculated based on the dynamic parameters is 65, and the initial control amount after compensation is 65 + 0.1 = 65.1. Then, the maximum control amount of the lighting equipment in A1 is preset to 70 and the minimum control amount is 5. After output limiting of 65.1, the lighting equipment control signal 65.1 is obtained. The control signal is converted into binary code 01100101 using the RS485 encoding protocol to obtain the lighting brightness adjustment signal and transmit it to the LED light driver module of A1. The driver module adjusts the brightness of the LED light in A1 from 280 lux to about 342 lux. Similarly, the control signal of A2 is calculated and encoded and transmitted, and the brightness is adjusted from 220 lux to about 231 lux. The brightness of A3 is adjusted from 250 lux to about 271 lux, so as to complete the entire adaptive lighting control process.

[0136] Figure 3 This is a schematic diagram illustrating a specific implementation of an adaptive intelligent control system for lighting in a building office area, as provided in this application. (Refer to...) Figure 3 The system may include:

[0137] The data acquisition module 31 is used to collect personnel density data and illuminance data in various functional areas of the building's office space.

[0138] The prediction module 32 is used to predict the lighting demand of each functional area of ​​the building office space based on the personnel density data and through multiple BP neural networks optimized by the particle swarm optimization algorithm, so as to obtain multiple initial illuminance values ​​for each functional area.

[0139] The voting module 33 is used to perform predictive average voting processing on all the initial illuminance values ​​of the same functional area to obtain the target illuminance value of the functional area, and calculate the difference between the target illuminance value and the illuminance data to obtain the illuminance deviation value.

[0140] The adjustment module 34 is used to perform fuzzy PID adjustment processing on the illuminance deviation value, generate the lighting brightness adjustment signal of the corresponding functional area, and transmit the lighting brightness adjustment signal to the lighting equipment drive module of the corresponding functional area to complete the adaptive control of lighting brightness.

[0141] The adaptive intelligent control system for building office lighting in this application is used to implement the aforementioned adaptive intelligent control method for building office lighting. Therefore, the specific implementation of the adaptive intelligent control system for building office lighting can be found in the embodiment section of the adaptive intelligent control method for building office lighting mentioned above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0142] like Figure 4 As shown, this application also provides an electronic device, including: a memory 41 for storing a computer program; and a processor 42 for executing the computer program to implement the steps of any of the above-described adaptive intelligent control methods for building office space lighting.

[0143] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described adaptive intelligent control methods for lighting in building office areas.

[0144] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0145] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above embodiments of the adaptive intelligent control method for building office lighting.

[0146] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0147] The above provides a detailed description of the adaptive intelligent control method and system for office lighting in buildings provided by this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for adaptive intelligent control of lighting in office areas of a building, characterized in that, include: Collect personnel density and illuminance data for each functional area of ​​the building's office space; Based on the personnel density data, multiple BP neural networks optimized by particle swarm optimization algorithm are used to predict the lighting demand of each functional area in the building office area, and multiple initial illuminance values ​​for each functional area are obtained. A predictive average voting process is performed on all the initial illuminance values ​​of the same functional area to obtain the target illuminance value of the functional area. The difference between the target illuminance value and the illuminance data is calculated to obtain the illuminance deviation value. The illuminance deviation value is processed by fuzzy PID control to generate a lighting brightness adjustment signal for the corresponding functional area. The lighting brightness adjustment signal is then transmitted to the lighting equipment driver module of the corresponding functional area to complete the adaptive control of the lighting brightness.

2. The method according to claim 1, characterized in that, Based on the personnel density data, multiple backpropagation (BP) neural networks optimized by particle swarm optimization are used to predict the lighting demand of each functional area in the building's office space, resulting in multiple initial illuminance values ​​for each functional area, including: Obtain the lighting requirements baseline for each functional area; A lighting equipment deployment model is constructed. Based on the lighting equipment deployment model and light intensity distribution characteristics, the light contribution weight of each lighting equipment to each location in the functional area is calculated to obtain the equipment light influence model. Based on the personnel density data and lighting demand benchmarks of each functional area, and combined with the pre-built equipment lighting impact model, multiple BP neural networks optimized by the particle swarm optimization algorithm are used to predict the lighting demand of each functional area in the building office area, and obtain multiple initial illuminance values ​​for each functional area.

3. The method according to claim 2, characterized in that, Based on the personnel density data and lighting demand benchmarks of each functional area, and combined with a pre-built equipment lighting impact model, multiple backpropagation (BP) neural networks optimized by particle swarm optimization are used to predict the lighting demand of each functional area in the building's office space, resulting in multiple initial illuminance values ​​for each functional area, including: The corresponding features are extracted from the personnel density data, the lighting demand benchmark, and the equipment lighting impact model, respectively. According to the preset combination rules, all features are combined to form the neural network input feature vector. The input feature vector of the neural network is input into multiple parallel BP neural networks for forward propagation processing to predict the lighting demand of each functional area in the building office space. The initial illuminance value of each functional area is obtained from the output of each BP neural network. The network parameters of each BP neural network are pre-optimized by the particle swarm optimization algorithm. During the optimization process, the comfort of the people in the functional area is taken as the optimization goal. The fitness function designed based on the optimization goal includes a prediction accuracy term and an energy consumption index term, and the parameters of different BP neural networks are different.

4. The method according to claim 1, characterized in that, The step of performing a predicted average voting process on all the initial illuminance values ​​of the same functional area to obtain the target illuminance value of the functional area includes: A confidence assessment is performed on all initial illuminance values ​​of the same functional area to obtain a confidence weight corresponding to each initial illuminance value. The confidence weight is dynamically adjusted based on the historical prediction accuracy of the corresponding BP neural network during the confidence assessment process. Based on the confidence weights, calculate the weighted average illuminance value of the corresponding functional area; Based on the weighted average illuminance value, the dispersion evaluation process is performed on all the initial illuminance values ​​corresponding to the functional area to identify the abnormal illuminance values ​​of the corresponding functional area. Calculate the arithmetic mean of all normal initial illuminance values ​​in the functional area after removing the abnormal illuminance values, and use the arithmetic mean as the target illuminance value for the corresponding functional area.

5. The method according to claim 4, characterized in that, The step of performing a dispersion assessment on all the initial illuminance values ​​corresponding to the functional area based on the weighted average illuminance value to identify abnormal illuminance values ​​in the corresponding functional area includes: The window size is adaptively adjusted based on the rate of change in personnel density in functional areas to construct a dynamic evaluation window; Based on the dynamic evaluation window, a relative stability analysis method is used to fuse the weighted average illuminance value and the initial illuminance value to obtain the stability coefficient corresponding to each initial illuminance value. Based on the lighting requirements benchmark of the functional area and the personnel density data, the boundary conditions for anomaly detection of the functional area are determined. Based on the stability coefficient and the anomaly determination boundary conditions, an anomaly marking algorithm is used to filter the initial illuminance values ​​of the anomalies to obtain the abnormal illuminance values ​​of the functional areas.

6. The method according to claim 1, characterized in that, The step of performing fuzzy PID adjustment processing on the illuminance deviation value to generate a lighting brightness adjustment signal for the corresponding functional area, and transmitting the lighting brightness adjustment signal to the lighting equipment driver module of the corresponding functional area to complete the adaptive control of lighting brightness, includes: Based on the preset lighting level classification rules, the illuminance deviation value is fuzzified to obtain the level distribution. The level distribution is then matched with the preset lighting adjustment rules to obtain the adjustment amount of each lighting parameter. The adjustment amounts of each lighting parameter are precisely converted to obtain precise adjustment values ​​for each parameter. Based on these precise adjustment values, a dynamic compensation algorithm is used to perform control quantity compensation calculations to generate control signals for the lighting equipment. The control signal of the lighting equipment is processed by brightness adjustment encoding to obtain the lighting brightness adjustment signal for each functional area; The lighting brightness adjustment signal is transmitted to the lighting device driver module corresponding to the functional area to complete the adaptive control of lighting brightness.

7. The method according to claim 6, characterized in that, The adjustment amounts of the lighting parameters are precisely converted to obtain precise adjustment values ​​for each parameter. Based on these precise adjustment values, a dynamic compensation algorithm is used to perform control quantity compensation calculations to generate a lighting equipment control signal, including: The centroid method is used to defuzzify the lighting parameters and perform precise conversion to obtain the precise adjustment values ​​corresponding to the proportional coefficient, integral time and derivative coefficient. The precise adjustment value is fused with the preset benchmark PID parameters to obtain a set of dynamic control parameters; Based on the aforementioned set of dynamic control parameters, a dynamic compensation algorithm is used to perform numerical compensation of the control quantity to obtain the initial control quantity; The initial control quantity is subjected to output limiting processing to obtain the lighting equipment control signal.

8. A building office area lighting adaptive intelligent control system, characterized in that, include: The data acquisition module is used to collect personnel density data and illuminance data in various functional areas of the building's office space; The prediction module is used to predict the lighting demand of each functional area of ​​the building office space based on the personnel density data and through multiple BP neural networks optimized by the particle swarm optimization algorithm, so as to obtain multiple initial illuminance values ​​for each functional area. The voting module is used to perform predictive averaging voting on all the initial illuminance values ​​of the same functional area to obtain the target illuminance value of the functional area, and to calculate the difference between the target illuminance value and the illuminance data to obtain the illuminance deviation value. The adjustment module is used to perform fuzzy PID adjustment processing on the illuminance deviation value, generate the lighting brightness adjustment signal of the corresponding functional area, and transmit the lighting brightness adjustment signal to the lighting equipment drive module of the corresponding functional area to complete the adaptive control of lighting brightness.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the adaptive intelligent control method for building office space lighting as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the adaptive intelligent control method for building office space lighting as described in any one of claims 1 to 7.

Images