Intelligent building energy conservation and emission reduction digital twin management system and method

Abstract

The invention provides an intelligent building energy conservation and emission reduction digital twin management system and method, and relates to the field of building automation and energy conservation control. The intelligent building energy conservation and emission reduction digital twinborn management system comprises a digital twinborn model construction module, an energy consumption prediction and regulation module, a multi-device cooperative control module, a real-time feedback and self-learning module and an edge-cloud cooperative architecture. By integrating a building information model, real-time sensor data, meteorological data and the like, a three-dimensional visual dynamic twinborn model is constructed, historical data playback and thermodynamic diagram analysis are supported, omnibearing virtual mapping of a physical building is realized, a physical system and virtual model data are synchronized in real time through an Internet of Things sensor network, and the real-time real-time dynamic twinborn model is established. The consistency of a virtual model and an actual system is ensured, the precision and reliability of the model are remarkably improved, artificial illumination is dynamically compensated according to natural illumination, lamps are automatically turned off in an unmanned area through infrared induction, and equipment-level energy saving is achieved in combination with air conditioning system optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of building automation and energy-saving control, and specifically to an intelligent building energy-saving and emission-reduction digital twin management system and method. Background Art

[0002] Building management refers to the systematic management and maintenance of buildings and their affiliated facilities, equipment, environment, and operations to ensure the safe, efficient, and energy-saving operation of buildings and improve the user experience. It covers multiple aspects from daily maintenance to long-term planning and is an essential part of the operation of modern urban buildings.

[0003] Existing building energy management systems rely on static models and cannot synchronize physical environment changes in real time, resulting in a large deviation between prediction and actual operation. Energy consumption prediction relies on single historical data and lacks multi-source data fusion. Equipment such as air conditioners and lighting is independently regulated, lacking cross-system linkage, resulting in low energy efficiency. It relies on manual parameter adjustment and cannot self-correct according to real-time data, and prediction errors accumulate after long-term operation. Summary of the Invention

[0004] (I) Technical Problems to be Solved

[0005] Aiming at the deficiencies of the existing technology, the present invention provides an intelligent building energy-saving and emission-reduction digital twin management system and method, which solves the problem that traditional building energy management systems rely on static models and cannot synchronize physical environment changes in real time, resulting in a large deviation between prediction and actual operation.

[0006] (II) Technical Solutions

[0007] To achieve the above objectives, the present invention is realized through the following technical solutions: An intelligent building energy conservation and emission reduction digital twin management system and method, including a digital twin model construction module, an energy consumption prediction and regulation module, a multi-device collaborative control module, a real-time feedback and self-learning module, and an edge-cloud collaborative architecture. The digital twin model construction module establishes a virtual mapping model of the building physical system based on building structure, equipment parameters, and historical operation data, and realizes real-time synchronization of environmental data and equipment status and dynamic update of model parameters through the Internet of Things sensor network. The energy consumption prediction and regulation module predicts the air conditioning and lighting loads through AI driving combined with meteorological parameters, historical load data, and personnel flow characteristics, and dynamically adjusts the target temperature and fresh air optimization control strategy. The multi-device collaborative control module optimizes the air conditioning start-stop strategy based on the load prediction result, dynamically adjusts the lighting brightness, and realizes cross-device linkage control of the air conditioning and lighting systems. The real-time feedback and self-learning module corrects the prediction error by comparing the digital twin model with the actual operation data, and uses the reinforcement learning algorithm to iteratively optimize the control strategy. The edge-cloud collaborative architecture issues real-time control commands to air conditioning and lighting equipment, and the cloud performs big data analysis and model training to support short, medium, and long-term energy consumption prediction.

[0008] Preferably, the digital twin model construction module includes an Internet of Things sensor network, a three-dimensional visualization dynamic twin model, and a dynamic update mechanism. The Internet of Things sensor network collects environmental data in real time. The three-dimensional visualization dynamic twin model integrates structured data of the building information model, equipment parameters, real-time environmental data collected by the Internet of Things sensors, and outdoor climate data provided by the meteorological API, and provides a virtual mapping and visualization interface. The dynamic update mechanism uses the differential evolution algorithm to compare the actual data with the twin data every 30 minutes to correct the model parameter deviation.

[0009] Preferably, the energy consumption prediction and regulation module includes a load prediction unit, a dynamic temperature regulation unit, and a fresh air intelligent control unit. The load prediction unit uses a hybrid neural network of LSTM-Transformer to predict the terminal air conditioning load and optimize the air volume and fresh air volume. The dynamic temperature regulation unit controls the temperature within a dynamic range of 22-28°C according to outdoor climate conditions and the human thermal comfort evaluation standard. The human thermal comfort evaluation standard is based on the PMV index. The fresh air intelligent control unit is used to monitor the indoor carbon dioxide concentration in real time and optimize the opening strategy of the mechanical fresh air valve.

[0010] Preferably, the multi-device collaborative control module includes an air-conditioning system optimization unit, a lighting adaptive adjustment unit, and a cross-system linkage unit. The air-conditioning system optimization unit is used to delay the start-up time of the large system and formulate the start-stop strategies of the chiller and the water pump. The lighting adaptive adjustment unit dynamically adjusts the regional brightness according to the light intensity and personnel distribution, and automatically turns off the lighting in unoccupied areas. The cross-system linkage unit is used to synchronously increase the lighting brightness and the air-conditioning cooling capacity during peak passenger flow periods.

[0011] Preferably, the real-time feedback and self-learning module includes an error correction unit and a reinforcement learning unit. The error correction unit dynamically optimizes the algorithm parameters by comparing the digital twin model with the actual operation data. The reinforcement learning unit uses the Q-learning algorithm with the joint scoring model of energy consumption and comfort as the target to dynamically adjust the weight coefficients α and β. The weight coefficients α and β of the joint scoring model of energy consumption and comfort are dynamically adjusted according to the season type and usage scenario. In summer, α:β = 6:4; in winter, α:β = 5:5; in the transitional season, α:β = 7:3.

[0012] Preferably, the edge node of the edge-cloud collaborative architecture issues real-time control instructions to air-conditioning and lighting devices through the Modbus / OPC protocol, and the cloud performs big data analysis and model training to support short-, medium-, and long-term energy consumption prediction.

[0013] Preferably, the method steps are as follows: Phase 1: System initialization S1. Import the building information model and equipment parameters for building modeling, construct a three-dimensional visualization digital twin base, establish a topology map of sensor deployment, and plan temperature and humidity monitoring nodes, carbon dioxide concentration monitoring nodes, light intensity monitoring nodes, and energy consumption monitoring nodes. S2. Deploy the Internet of Things sensor network for infrastructure installation, connect to the building automation control system, and complete the communication debugging of the Modbus protocol and the OPC protocol. Phase 2: Data synchronization S3. The real-time data flow sensor collects environmental parameters at a frequency of once per second, and the edge node performs data cleaning to filter out noise data and compress redundant information. S4. Access historical data, load the energy consumption data of the past three years, integrate the application program interface of the meteorological bureau, and import the weather prediction data for the next three days. Phase 3: Dynamic prediction S5. The load prediction engine uses the long short-term memory neural network module to process time series data, combines the transformer module to capture spatial correlation, and outputs the load curves of the air-conditioning system and the lighting system within the next hour, with a prediction confidence exceeding 85%. Phase 4: Real-time regulation S6, Temperature Dynamic Equilibrium S6.1, Energy Consumption Threshold Trigger Mechanism: The operating power of the air conditioning system is monitored in real time. When the cooling power in summer exceeds 90% of the rated value and the heating power in winter exceeds 85% of the rated value, the dynamic temperature regulation program is activated. When not exceeding the limit, the indoor environmental comfort standard defined by the International Organization for Standardization is maintained to ensure that the predicted mean vote index is in the comfort range of -0.5 to +0.5; S6.2, Temperature Dynamic Regulation Process Summer Operating Mode: The initial set air conditioning target temperature is 26 degrees Celsius. If the energy consumption continues to exceed the standard, the upper limit of the temperature is increased by 0.5 degrees Celsius every 30 minutes, and the maximum is relaxed to 28 degrees Celsius. During the temperature adjustment period, the predicted mean vote index is continuously monitored. When it is detected that the index in a local area exceeds plus or minus 0.5, the air supply speed is automatically increased to compensate for the thermal comfort; Winter Operating Mode: The initial set target temperature is 22 degrees Celsius. If the energy consumption exceeds the standard, the lower limit of the temperature is reduced by 0.5 degrees Celsius every 30 minutes, and the minimum is adjusted to 20 degrees Celsius. At the same time, by adjusting the air supply outlet angle, the hot air preferentially covers the personnel activity area to avoid the comfort loss caused by the overall temperature drop; S6.3, Cross-System Linkage Control: After the temperature regulation instruction is generated, it is sent to the air conditioning control system in real time through the Modbus protocol, and at the same time, the lighting brightness in the corresponding area is reduced to 70% of the reference value to reduce the interference of equipment heat generation on temperature regulation; S7, Fresh Air Optimization Control S7.1, Carbon Dioxide Concentration Hierarchical Control Strategy: When the indoor carbon dioxide concentration is lower than 800 ppm, the minimum opening of the fresh air valve is maintained at 30% to reduce the fresh air treatment load of the air conditioning system. When the carbon dioxide concentration rises to the range of 800 to 1200 ppm, the system linearly adjusts the opening of the fresh air valve to 80% within 15 minutes, and at the same time increases the air conditioning cooling capacity to offset the heat load brought by the newly added fresh air. When the carbon dioxide concentration exceeds 1200 ppm, the fresh air valve is immediately fully opened and the emergency ventilation mode is started, and the lighting equipment in non-critical areas is turned off synchronously to prevent instantaneous grid overload; S7.2, Energy Efficiency Optimization Mechanism: During the low personnel density period from 23:00 to 6:00 the next day, the minimum fresh air volume mode is forcibly enabled to limit the opening of the fresh air valve and reduce the energy consumption of the air conditioning system; Phase 5: Equipment Collaboration S8, Air Conditioning System Optimization S8.1, Air Conditioning Start-Stop Strategy: The chiller is pre-started 30 minutes in advance according to the load prediction result; S8.2, Water Pump Variable Frequency Control: The speed of the water pump is dynamically adjusted according to the end differential pressure to match the actual cooling demand; S9. The lighting system is linked to adapt the area lighting brightness from personnel presence detection. The curtain opening degree and the color temperature of the light-emitting diodes are adjusted every five minutes, and natural light compensation is used to reduce the energy consumption of artificial lighting. Phase 6: Closed-loop optimization S10. Error correction compares the predicted value with the actual value every hour and calculates the mean absolute percentage error. When the mean absolute percentage error exceeds 10%, model retraining is triggered, and the weight parameters of the long short-term memory neural network are updated using incremental learning. S11. Policy evolution mechanism S11.1. The reinforcement learning-driven system starts a policy optimization cycle every 24 hours to explore the best balance point between energy consumption and comfort through trial-and-error learning. S11.2. Dynamic weight adjustment automatically adjusts the proportion relationship between the energy consumption optimization weight and the human comfort weight according to real-time operation data. When the energy consumption exceeds the standard for three consecutive days, the energy consumption optimization weight is increased to a maximum of 0.8 to prioritize the energy-saving goal. When the user complaint rate exceeds 5%, the human comfort weight is increased to a maximum of 0.7 to focus on comfort guarantee. S11.3. Policy iteration process 1. Exploration: Simulate and generate 300 groups of candidate control schemes in the virtual twin, including air conditioner start / stop strategies and lighting brightness gradients. 2. Verification and evaluation: Based on the joint scoring model of energy consumption and comfort, screen the top 10% scoring strategies among air conditioner start / stop strategies, lighting brightness gradients, cross-system linkage strategies, and fresh air optimization control strategies. 3. Practical application: Inject the screened strategies into the real building control system for operation and continuously monitor the actual effect for 8 hours. 4. Policy solidification: Standardize and package the strategies with a stable increase in the comprehensive score exceeding 2% and update them to the system policy library. Phase 7: Cross-layer collaboration S12. Edge and cloud interaction Edge side: Execute emergency regulation with a delay of less than 50 milliseconds. In case of a sudden temperature rise, perform a rapid cooling response. Cloud side: Perform global policy optimization every day at midnight to generate a weekly energy efficiency improvement plan, equipment maintenance plan, and load distribution strategy. S13. Digital twin synchronization updates the virtual model status every ten minutes, supports virtual reality and augmented reality visual monitoring, and historical operations support full traceability. The system regulation process at any time point can be replayed. Phase 8: Long-term evolution S14. System upgrade imports new device parameters monthly to expand the coverage of the digital twin model; updates the artificial intelligence prediction model quarterly and incorporates the latest climate pattern data and user behavior characteristics. S15. The energy efficiency report automatically generates multi-dimensional analysis reports, including the proportion of sub-item energy consumption, the contribution degree of energy-saving measures, the comfort compliance rate, and the evaluation of equipment operation efficiency.

[0014] (3) Beneficial effects

[0015] The present invention provides an intelligent building energy-saving and emission-reduction digital twin management system and method, having the following beneficial effects: 1. By integrating building information models, real-time sensor data, meteorological data, etc., the present invention constructs a three-dimensional visual dynamic twin model, supports historical data playback and heat map analysis, realizes the full-range virtual mapping of the physical building, synchronizes the physical system and virtual model data in real time through the Internet of Things sensor network, and combines the differential evolution algorithm to dynamically optimize the model parameters every 30 minutes to ensure the consistency between the virtual model and the actual system, significantly improving the model accuracy and reliability.

[0016] 2. By fusing time series features and spatial features, the present invention realizes short-term and medium- to long-term energy consumption prediction, updates the prediction results every 15 minutes and generates a load curve, dynamically adjusts the temperature based on the PMV index, relaxes the temperature control range elastically in summer / winter scenarios, combines fresh air hierarchical control and lighting collaborative dimming, reduces redundant energy consumption while ensuring thermal comfort, dynamically compensates artificial lighting according to natural light, automatically turns off the lights in unoccupied areas with infrared induction, combines air-conditioning system optimization to achieve equipment-level energy saving, pre-cools in advance and reduces the lighting brightness in high-temperature weather, and reduces the load of the lighting system to balance the grid load when the air conditioner fails, and responds to extreme scenarios through equipment linkage strategies to improve the system stability.

[0017] 3. By comparing the model prediction with the actual data every 30 minutes, dynamically adjusting the parameters and recording the error improvement rate, the present invention forms the ability of fault traceability and optimization. Taking the energy consumption and comfort joint scoring model as the goal, adapting to the seasonal scenarios, generating the optimal control instructions through Q-learning, realizing the closed-loop of "prediction - execution - feedback - optimization", continuously improving the system adaptability. The edge nodes are responsible for millisecond-level real-time control, the cloud centrally processes big data analysis and AI model training, supports short-, medium- and long-term energy consumption prediction, uses AES-256 encryption and RBAC permission management to ensure data security, and the edge cache maintains the basic operation when the network is disconnected and synchronizes with the cloud after the network is restored to ensure the system robustness. Description of the drawings

[0018] Figure 1 It is the system flow chart of the present invention. Specific implementation manners

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention. Embodiment

[0020] As Figure 1 shown, the embodiment of the present invention provides an intelligent building energy conservation and emission reduction digital twin management system and method, including a digital twin model construction module, an energy consumption prediction and regulation module, a multi-device collaborative control module, a real-time feedback and self-learning module, and an edge-cloud collaborative architecture. The digital twin model construction module establishes a virtual mapping model of the building physical system based on the building structure, equipment parameters, and historical operation data, and realizes the real-time synchronization of environmental data and equipment status and the dynamic update of model parameters through the Internet of Things sensor network. The energy consumption prediction and regulation module predicts the air conditioning and lighting loads by combining AI with meteorological parameters, historical load data, and personnel flow characteristics, and dynamically adjusts the target temperature and fresh air optimization control strategy. The multi-device collaborative control module optimizes the air conditioning start-stop strategy based on the load prediction result, dynamically adjusts the lighting brightness, and realizes the cross-device linkage control of the air conditioning and lighting systems. The real-time feedback and self-learning module corrects the prediction error by comparing the digital twin model with the actual operation data, and uses the reinforcement learning algorithm to iteratively optimize the control strategy. The edge-cloud collaborative architecture issues real-time control instructions to the air conditioning and lighting equipment, and the cloud performs big data analysis and model training to support short, medium, and long-term energy consumption prediction.

[0021] The digital twin model construction module includes an Internet of Things sensor network, a three-dimensional visualization dynamic twin model, and a dynamic update mechanism. Through multi-source data fusion and real-time synchronization technology, it integrates the geometric layout and device parameter structured data of the building information model, historical operation data, temperature and humidity, CO2 concentration, device status data collected by real-time environmental sensors, and data on outdoor climate temperature, wind speed, and rainfall provided by the meteorological API to generate an initial three-dimensional visualization digital twin body. An embedded real-time data interface is used to render a three-dimensional dynamic interface through a three-dimensional engine, display the device status in real time, and support historical data playback and heat map analysis, thereby constructing a high-precision building virtual mapping system. The Internet of Things sensor network deploys distributed sensors to cover each functional area of the building, including the office area, corridor, and computer room, to collect environmental data and device operation status in real time. Wireless transmission technology uploads the data to the edge node, and after preprocessing such as filtering and compression, it is transmitted to the digital twin model to achieve millisecond-level data synchronization between the physical system and the virtual model. The dynamic update and optimization mechanism compares the sensor data with the model prediction value every 30 minutes through the differential evolution algorithm, calculates the parameter deviation, and dynamically adjusts the model parameters to ensure that the temperature prediction error between the virtual model and the actual system is ≤0.5°C and the energy consumption deviation is ≤3%, ensuring the consistency between the virtual model and the actual system. After updating the model, a log is generated, recording the parameter changes and error improvement rate before and after correction in the log, and fusing multi-source heterogeneous data to form a unified data pool, thereby supporting fault tracing and model optimization and reducing the prediction error.

[0022] The energy consumption prediction and regulation module includes a load prediction unit, a dynamic temperature regulation unit, and a fresh air intelligent control unit. Through artificial intelligence technology, it integrates meteorological parameters, historical load data, and personnel flow characteristics to achieve the collaborative optimization of energy consumption prediction and dynamic regulation. Through multi-unit linkage and closed-loop data feedback, the system reduces redundant energy consumption while ensuring a comfortable range of PMV index from -0.5 to +0.5, achieving a dynamic balance between energy efficiency and user experience. The load prediction unit uses the LSTM-Transformer model to fuse the spatial characteristics of time series data, historical load, meteorological parameters, and personnel distribution, and inputs the historical load data , matching W, and P at the same moment one by one according to the time stamp. For each time step t, the "historical load characteristics, meteorological characteristics, and personnel characteristics" are concatenated into a new feature vector, that is: enhanced input , where is the feature at time step t, , Similarly, is the time step length, is the feature dimension, is the meteorological parameter, is the personnel flow characteristic. Therefore, in terms of data structure is the spatio-temporal feature matrix of historical load. Each row corresponds to "all load features at a certain moment". The matrix has T rows, corresponding to T historical moments. Each column corresponds to "the changes of a certain type of load feature at different moments". The matrix has d columns, corresponding to d types of load features; LSTM layer: ; Transformer layer: ; Simultaneously undertakes three core roles. Query uses historical load data as the "query vector" to actively "query" which moment's load is more important for the current prediction. Key uses historical load data as the "key vector" as the "matching basis" to calculate the similarity with Query, thereby judging the importance weights of different moments. Value uses historical load data as the "value vector" as the "information to be weighted". According to the attention weights matched by Query and Key, a weighted sum is performed on Value to obtain the final attention output , when Query, Key, and Value come from the same set of data, this kind of attention is called self-attention. Therefore, in terms of functional logic, is the input of the Transformer self-attention mechanism, mixed output: , predict the short-term energy consumption demand of air conditioning and lighting systems within 1 hour and the medium- and long-term energy consumption demand of 24 hours to 1 week, update the prediction results every 15 minutes, output the load curve of cooling capacity and lighting power in the future period, and recommend air volume adjustment strategies, and feed back real-time operation data to the digital twin model to drive the update of model parameters and improve long-term prediction accuracy. The dynamic temperature adjustment unit inputs indoor temperature and humidity, personnel activity intensity and clothing thermal resistance based on the PMV index, and dynamically calculates the thermal comfort score to ensure that the score is in the comfort range of -0.5 to +0.5. The initial air conditioning target temperature is set to 26℃ in the summer scene. If the energy consumption exceeds the standard, the temperature upper limit is relaxed by 0.5℃ every 30 minutes, up to 28℃. At the same time, the air supply wind speed is increased to compensate for local overheating. The initial temperature lower limit of the winter scene is set to 22℃. When the energy consumption exceeds the standard, The temperature is gradually reduced to 20℃, and the angle of the air supply outlet is adjusted to give priority to covering the area where people are active. The temperature command is issued through the Modbus protocol, and the lighting brightness of the corresponding area is simultaneously reduced to 70% to reduce the interference of equipment heating. The fresh air intelligent control unit adopts a graded air supply strategy and is graded according to the CO2 concentration. When the CO2 concentration is lower than 800ppm, the minimum opening of the fresh air valve is maintained at 30%. When the concentration rises to 800-1200ppm, the opening is linearly adjusted to 80% within 15 minutes, and the air conditioning cooling capacity is increased to offset the heat load. When the concentration exceeds 1200ppm, the fresh air valve is fully opened and emergency ventilation is started, and the lighting in non-critical areas is turned off. The minimum fresh air volume mode is forced to be enabled during the low-load period at night, and the opening is ≤30% to reduce air conditioning energy consumption and coordinate with the air conditioning system to avoid a sudden increase in fresh air volume and a surge in load.

[0023] The multi-device collaborative control module consists of an air-conditioning system optimization unit, a lighting adaptive adjustment unit, and a cross-system linkage unit. Based on energy consumption prediction results and real-time environmental data, it optimizes through intelligent algorithms to coordinate the operation strategies of multiple devices and realize scenario-based energy efficiency management. The air-conditioning system optimization unit dynamically plans the start and stop time of chillers and water pumps, delays the start of high-power equipment during non-peak hours, and adjusts the air supply volume and target temperature based on historical load curves to match actual cooling needs and avoid overcooling or overheating, thereby reducing redundant energy consumption. The lighting adaptive adjustment unit dynamically compensates for artificial lighting brightness based on natural light intensity, reducing brightness on sunny days. The temperature is reduced to half of the baseline value, and the distribution of people is detected by infrared sensors. The lighting is automatically turned off in unmanned areas and the brightness is adjusted as needed in occupied areas. The office mode has uniform lighting and the conference mode has focused lighting. It supports one-click switching of color temperature configuration to reduce lighting energy consumption. The cross-system linkage unit simultaneously increases the air conditioning cooling capacity and lighting brightness during high passenger flow periods. In hot weather, the air conditioning pre-cooling is started in advance and the lighting brightness is lowered. When the air conditioner fails, the lighting system is automatically unloaded to balance the grid load, shorten the fault recovery time, and ensure the stability of the system in extreme scenarios. Through the coordination and dynamic response mechanism of the strategies of multiple devices, a global balance between energy efficiency optimization and user experience is achieved.

[0024] The real-time feedback and self-learning module includes an error correction unit and a reinforcement learning unit. By comparing the predicted data of the digital twin model with the actual operation data, it dynamically corrects the model error and optimizes the control strategy to continuously improve energy efficiency and comfort. The error correction unit extracts the model prediction value and actual sensor data every 30 minutes, calculates the error rate, identifies systematic errors, and dynamically adjusts the model parameters using the differential evolution algorithm; Minimize the model prediction error: ; Among them, is the model device efficiency parameter, is the sensor data, is the output of the digital twin model. The parameter optimization is executed every 30 minutes, and the parameter change rate , the error improvement rate , reduce subsequent prediction errors, record the numerical changes before and after parameter correction and the error improvement effect, support fault tracing and model tuning, ensure the accuracy and reliability of the long-term operation of the system. The reinforcement learning unit constructs a joint energy consumption and comfort score model using the Q-learning algorithm. The joint energy consumption and comfort score model is used to quantify the comprehensive performance of energy efficiency and comfort of the strategy, including energy consumption score and comfort score. According to the difference between the actual energy consumption and the predicted energy consumption when the equipment strategy is running, the electricity saved by the chilled water unit of the air conditioner by delaying startup / advancing shutdown, the energy consumption reduced by the lighting system through natural light compensation and turning off lights when no one is present, the air-conditioning load reduced by optimizing the opening degree of the fresh air valve, and also including cross-system linkage energy saving, such as the synergistic effect of reducing the lighting brightness during high-temperature periods to relieve the air-conditioning burden. Judge the energy consumption score α. The scoring rule is that the lower the energy consumption and the smaller the prediction deviation, the closer it is to the predetermined energy-saving effect of the strategy, and the higher the score. The comfort score β, the proportion of the duration when the PMV index in the monitored area is in the comfortable range (-0.5 to +0.5), and count the area proportion of the local area where PMV is not in the comfortable range due to temperature adjustment. The higher the duration proportion and the smaller the area proportion, the stronger the PMV stability. Count the cumulative duration when the CO2 concentration exceeds 1200 ppm. The shorter the duration, the higher the score. Record the user complaint rate, such as the proportion of the number of complaints about uncomfortable temperature and too dim lighting in the total operation duration. The stronger the PMV stability, the higher the CO2 compliance rate, and the fewer user complaints, the higher the score. The dynamic weight mechanism α:β, in summer, the weight α:β = 6:4, emphasizing energy saving, allowing the upper temperature limit to be relaxed to 28°C. In winter, the weight α:β = 5:5, balancing energy consumption and comfort, and the lower temperature limit is not lower than 20°C. In the transitional season, the weight α:β = 7:3, with a strong energy-saving orientation, allowing greater temperature fluctuations. If the energy consumption exceeds the standard for 3 consecutive days, increase α to 0.8 to prioritize energy saving. If the user complaint rate > 5%, increase β to 0.7 to prioritize comfort. Collect real-time energy consumption data and comfort data during the operation of the strategy, calculate the energy consumption score and comfort score according to the energy consumption data and comfort data. Energy consumption score = α × (energy-saving effect coefficient + prediction accuracy coefficient), comfort score = β × (PMV compliance rate × 60% + CO2 compliance rate × 30% + user satisfaction × 10%). User satisfaction = 1 - complaint rate. The final score = energy consumption score + comfort score, with a full score of 100 points. It is necessary to meet both energy consumption reduction and comfort compliance. Generate air conditioner start / stop strategies, lighting brightness gradients, cross-system linkage strategies, and fresh air optimization control strategies according to real-time environmental data. Verify the strategy effect through simulation and actual operation, screen the optimal solution and update it to the control system. The optimized strategy is sent to the device for execution through the edge-cloud collaborative architecture, and the real-time operation data is synchronously fed back to the digital twin model and the energy consumption prediction module to form a "prediction-execution-feedback-optimization" closed loop, continuously improving the system's adaptive ability and overall energy efficiency.

[0025] The edge nodes of the edge-cloud collaborative architecture send real-time control instructions to air conditioners and lighting equipment through the Modbus / OPC protocol. The cloud performs big data analysis and model training to support short-, medium- and long-term energy consumption prediction. The edge-cloud collaborative architecture rationally distributes real-time control and big data analysis tasks to edge nodes and cloud platforms through a layered processing mechanism to achieve high-efficiency and low-latency building energy management. Edge nodes are deployed locally in the building and communicate directly with air conditioners and lighting equipment through the Modbus / OPC protocol to perform real-time data collection, preprocessing and control instruction issuance to ensure millisecond-level response. They perform low-latency operations such as equipment start and stop, temperature adjustment, and lighting switches to ensure rapid response of the building environment. In scenarios with network disconnection or high latency, basic operations are maintained based on cache strategies, key data is temporarily stored, and synchronized to the cloud after the network is restored. The cloud platform is responsible for centrally storing massive historical and real-time data and running A The I model is used for big data analysis, energy consumption prediction and strategy optimization. The air conditioning and lighting loads in the next hour are predicted based on the LSTM model. The energy consumption demand in the next day is predicted by combining weather trends and personnel flow patterns. The seasonal and periodic laws are analyzed to generate energy efficiency optimization plans for the next week to one month. The energy consumption and comfort weight coefficients are dynamically adjusted through reinforcement learning algorithms to generate cross-system collaboration strategies. In terms of collaboration and security mechanisms, the edge and cloud realize data synchronization and command interaction through secure communication protocols, and use AES-256 to encrypt data transmission to prevent information leakage. The device access rights are managed based on the RBAC model to ensure the legitimacy of the command issuance and the security of the system. Through division of labor and collaboration, the edge nodes ensure real-time response and local reliability. The cloud platform provides global analysis and strategy optimization capabilities to form an efficient closed loop, significantly improving the intelligence level of building energy management, and ensuring user experience while saving energy and reducing emissions.

[0026] Generate 300 sets of candidate control schemes in the virtual twin model, including air conditioner start-stop strategies, lighting brightness gradients, cross-system linkage strategies, and fresh air optimization control strategies. Based on the "energy consumption - comfort joint scoring model", screen the top 10% of the strategies with the highest scores as "optimal strategies", inject the optimal strategies into the real building control system, continuously monitor the actual effects for 8 hours, standardize and package the strategies that can stably improve the comprehensive score by more than 2%, and update them to the system strategy library. The air conditioner start-stop strategy dynamically adjusts the start and stop times of the air conditioning system according to the load prediction results, delays the start of high-power equipment such as chillers during off-peak hours, and pre-starts the chiller in advance, such as starting 30 minutes in advance, and dynamically adjusts the pump speed in combination with the end differential pressure to match the actual cooling demand. The lighting brightness gradient adjustment strategy dynamically adjusts the brightness gradient and color temperature of the lighting system according to the natural light intensity, personnel distribution, and scene requirements. On sunny days, reduce the artificial lighting brightness to half of the reference value, use natural light compensation to reduce artificial lighting energy consumption, detect personnel distribution through infrared sensors, automatically turn off the lighting in unoccupied areas, and automatically adapt the lighting state according to the scene mode. Uniform lighting is required in the office mode, and focused lighting is required in the meeting mode. The color temperature configuration can be switched according to different scenes through the system. The cross-system linkage strategy coordinates the operation of the air conditioner and the lighting system to achieve equipment-level energy saving. When the temperature adjustment command is issued, simultaneously reduce the lighting brightness in the corresponding area to 70% of the reference value to reduce the interference of equipment heat generation on temperature adjustment. In hot weather, pre-cool the air conditioner in advance and lower the lighting brightness. When the air conditioner fails, automatically reduce the load of the lighting system to balance the grid load. During high passenger flow periods, it is necessary to simultaneously increase the air conditioning cooling capacity and lighting brightness. The fresh air optimization control strategy adjusts the opening degree of the fresh air valve in stages according to the indoor carbon dioxide concentration. When the carbon dioxide concentration < 800 ppm, the minimum opening degree of the fresh air valve is 30%. When the carbon dioxide concentration is between 800 - 1200 ppm, linearly adjust the opening degree to 80% within 15 minutes and increase the air conditioning cooling capacity to offset the new heat load. When the carbon dioxide concentration > 1200 ppm, fully open the fresh air valve and start emergency ventilation, and simultaneously turn off the lighting in non-critical areas. There is a night energy-saving mode. During the low personnel density period from 23:00 to 6:00 the next day, the minimum fresh air volume mode is forcibly enabled to reduce air conditioning energy consumption, dynamically balance energy consumption and comfort, and dynamically adjust the weight coefficients through the reinforcement learning algorithm. α is the energy consumption optimization weight, and β is the comfort weight, adapting to different seasonal scenarios. In summer: α:β = 6:4, focusing on energy conservation. In winter: α:β = 5:5, balancing energy consumption and comfort. In the transitional season: α:β = 7:3, focusing on energy conservation. Through equipment collaboration and closed-loop feedback mechanisms, energy consumption is reduced while ensuring that the thermal comfort index is maintained within the comfortable range of -0.5 to +0.5.

[0027] The method steps are as follows: 1: System initialization S1. Import the building information model and equipment parameters for building modeling, construct a three-dimensional visual digital twin base, establish a topology map for sensor deployment, and plan temperature and humidity monitoring nodes, carbon dioxide concentration monitoring nodes, light intensity monitoring nodes, and energy consumption monitoring nodes; S2. Deploy and install the Internet of Things sensor network for infrastructure, connect to the building automation control system, and complete the communication debugging of the Modbus protocol and the OPC protocol; Phase 2: Data Synchronization S3. The real-time data stream sensor collects environmental parameters at a frequency of once per second, and the edge node performs data cleaning to filter out noise data and compress redundant information; S4. Access historical data to load the energy consumption data for the past three years, integrate the application program interface of the meteorological bureau, and import the weather prediction data for the next three days; Phase 3: Dynamic Prediction S5. The load prediction engine uses a long short-term memory neural network module to process time series data, combines a transformer module to capture spatial correlation, outputs the load curves of the air conditioning system and the lighting system within the next hour, and the prediction confidence exceeds 85%; Phase 4: Real-time Regulation S6. Temperature Dynamic Balance S6.1. The energy consumption threshold trigger mechanism monitors the operating power of the air conditioning system in real time. When the cooling power in summer exceeds 90% of the rated value and the heating power in winter exceeds 85% of the rated value, the dynamic temperature regulation program is activated. When not exceeding the limit, it maintains the indoor environmental comfort standard defined by the International Organization for Standardization to ensure that the predicted mean vote index is in the comfort interval of -0.5 to +0.5; S6.2. Temperature Dynamic Regulation Process Summer operating mode: Initially set the air conditioning target temperature to 26 degrees Celsius. If the energy consumption continues to exceed the standard, the upper limit of the temperature allowed will be increased by 0.5 degrees Celsius every 30 minutes, and the maximum will be relaxed to 28 degrees Celsius. During the temperature adjustment period, continuously monitor the predicted mean vote index. When it is detected that the index in a local area exceeds plus or minus 0.5, automatically increase the air supply speed to compensate for thermal comfort; Winter operating mode: Initially set the target temperature to 22 degrees Celsius. If the energy consumption exceeds the standard, the lower limit of the temperature allowed will be reduced by 0.5 degrees Celsius every 30 minutes, and the minimum will be adjusted to 20 degrees Celsius. At the same time, by adjusting the air outlet angle, the hot air preferentially covers the area where people are active to avoid comfort loss caused by the overall temperature drop; S6.3. After the temperature regulation instruction is generated for cross-system linkage control, it is sent to the air conditioning control system in real time through the Modbus protocol, and at the same time, the lighting brightness in the corresponding area is reduced to 70% of the reference value to reduce the interference of equipment heat generation on temperature regulation; S7. Fresh Air Optimization Control S7.1. The fresh air optimization control strategy is based on hierarchical control according to the carbon dioxide concentration. When the indoor carbon dioxide concentration is below 800 ppm, the minimum opening of the fresh air valve is maintained at 30%, reducing the fresh air treatment load of the air conditioning system. When the carbon dioxide concentration rises to the range of 800 - 1200 ppm, the system linearly adjusts the opening of the fresh air valve to 80% within 15 minutes, and at the same time increases the cooling capacity of the air conditioner to offset the heat load brought by the newly added fresh air. When the carbon dioxide concentration exceeds 1200 ppm, the fresh air valve is immediately fully opened and the emergency ventilation mode is started, and the lighting equipment in non-critical areas is synchronously turned off to prevent instantaneous overload of the power grid; S7.2. The energy efficiency optimization mechanism forcibly enables the minimum fresh air volume mode during the low personnel density period from 23:00 to 6:00 the next day, restricting the opening of the fresh air valve to reduce the energy consumption of the air conditioning system; Phase 5: Equipment coordination S8. Air conditioning system optimization S8.1. The start-stop strategy of the chiller pre-starts the chiller 30 minutes in advance according to the predicted load; S8.2. The variable frequency control of the water pump dynamically adjusts the speed of the water pump according to the end differential pressure to match the actual cooling demand; S9. The lighting brightness gradient adjustment strategy dynamically adjusts the brightness gradient and color temperature of the lighting system according to the natural light intensity, personnel distribution and scene requirements. On sunny days, the artificial lighting brightness is reduced to half of the reference value, using natural light compensation to reduce the energy consumption of artificial lighting. The cross-system linkage strategy coordinates the operation of the air conditioning and lighting systems to achieve equipment-level energy saving; Phase 6: Closed-loop optimization S10. Error correction compares the predicted value with the actual value every hour and calculates the mean absolute percentage error; when the mean absolute percentage error exceeds 10%, trigger model retraining and update the weight parameters of the long short-term memory neural network using incremental learning; S11. Strategy evolution mechanism S11.1. The reinforcement learning-driven system starts a strategy optimization cycle every 24 hours, exploring the optimal balance point between energy consumption and comfort through trial-and-error learning; S11.2. Dynamic weight adjustment automatically adjusts the ratio relationship between the energy consumption optimization weight and the human comfort weight according to the real-time operation data. When the energy consumption exceeds the standard for three consecutive days, the energy consumption optimization weight is increased to a maximum of 0.8 to give priority to ensuring the energy saving goal. When the user complaint rate exceeds 5%, the human comfort weight is increased to a maximum of 0.7 to focus on comfort guarantee; S11.3. Strategy iteration process 1. Exploration: Simulate and generate 300 groups of candidate control schemes in the virtual twin, including the air conditioning start-stop strategy and the lighting brightness gradient; 2. Verification and evaluation: Based on the combined scoring model of energy consumption and comfort, select the top 10% of the strategies in the air conditioner start-stop strategy, lighting brightness gradient, cross-system linkage strategy, and fresh air optimization control strategy; 3. Practical application: Inject the selected strategies into the real building control system for operation, and continuously monitor the actual effects for eight hours; 4. Strategy solidification: Standardize and package the strategies that can stably improve the comprehensive score by more than 2%, and update them to the system strategy library; Phase 7: Cross-layer collaboration S12. Edge-cloud interaction Edge side: Execute emergency regulation with a delay of less than 50 milliseconds. In case of a sudden temperature rise, perform a rapid cooling response; Cloud side: Perform global strategy optimization every day at midnight, generate a weekly energy efficiency improvement plan, equipment maintenance plan, and load distribution strategy; S13. Digital twin synchronization updates the virtual model status every ten minutes, supports virtual reality and augmented reality visual monitoring, and historical operations support full traceability. The system regulation process at any time point can be replayed; Phase 8: Long-term evolution S14. System upgrade Import new device parameters monthly, expand the coverage of the digital twin model, and update the artificial intelligence prediction model quarterly, incorporating the latest climate pattern data and user behavior characteristics; S15. Energy efficiency report Automatically generate multi-dimensional analysis reports, including the proportion of sub-item energy consumption, contribution degree of energy-saving measures, comfort compliance rate, and equipment operation efficiency evaluation.

[0028] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent building energy conservation and emission reduction digital twin management system, characterized in that: It includes a digital twin model construction module, an energy consumption prediction and regulation module, a multi-device collaborative control module, a real-time feedback and self-learning module, and an edge-cloud collaborative architecture. The digital twin model construction module establishes a virtual mapping model of the building physical system based on building structure, equipment parameters, and historical operation data, and realizes the real-time synchronization of environmental data and equipment status and the dynamic update of model parameters through the Internet of Things sensor network. The energy consumption prediction and regulation module predicts the air conditioning and lighting loads through AI-driven combined with meteorological parameters, historical load data, and personnel flow characteristics. The multi-device collaborative control module optimizes the air conditioning start-stop strategy and dynamically adjusts the lighting brightness based on the load prediction results. The real-time feedback and self-learning module corrects the prediction error by comparing the digital twin model with the actual operation data, and uses the reinforcement learning algorithm to iteratively optimize the control strategy. The edge-cloud collaborative architecture issues real-time control commands to air conditioning and lighting equipment, and the cloud performs big data analysis and model training to support short-, medium-, and long-term energy consumption prediction.

2. The intelligent building energy conservation and emission reduction digital twin management system according to claim 1, characterized in that: The digital twin model construction module includes an Internet of Things sensor network, a three-dimensional visualization dynamic twin model, and a dynamic update mechanism. The Internet of Things sensor network collects environmental data in real time. The three-dimensional visualization dynamic twin model integrates the structured data of the building information model, equipment parameters, real-time environmental data collected by the Internet of Things sensors, and outdoor climate data provided by the meteorological API, and provides a virtual mapping and visualization interface. The dynamic update mechanism uses the differential evolution algorithm to compare the actual data with the twin data every 30 minutes to correct the model parameter deviation.

3. An intelligent building energy conservation and emission reduction digital twin management system according to claim 1, characterized in that: The energy consumption prediction and regulation module includes a load prediction unit, a dynamic temperature regulation unit, and a fresh air intelligent control unit. The load prediction unit uses a hybrid neural network of LSTM-Transformer to predict the terminal air conditioning load and optimize the air volume and fresh air volume. The dynamic temperature regulation unit controls the temperature within the dynamic range of 22-28°C according to outdoor climate conditions and the human thermal comfort evaluation standard, and the human thermal comfort evaluation standard is based on the PMV index. The fresh air intelligent control unit is used to monitor the indoor CO2 concentration in real time and optimize the opening strategy of the mechanical fresh air valve.

4. An intelligent building energy conservation and emission reduction digital twin management system according to claim 1, characterized in that: The multi-device collaborative control module includes an air conditioning system optimization unit, a lighting adaptive adjustment unit, and a cross-system linkage unit. The air conditioning system optimization unit is used to delay the start time of the large system and formulate the start-stop strategies of the chiller and the pump. The lighting adaptive adjustment unit dynamically adjusts the regional brightness according to the light intensity and personnel distribution, and automatically turns off the lighting in unoccupied areas. The cross-system linkage unit is used to synchronously increase the lighting brightness and the air conditioning cooling capacity during peak passenger flow periods.

5. An intelligent building energy conservation and emission reduction digital twin management system according to claim 1, characterized in that: The real-time feedback and self-learning module includes an error correction unit and a reinforcement learning unit. The error correction unit dynamically optimizes the algorithm parameters by comparing with the actual operation data through the digital twin model. The reinforcement learning unit adopts the Q-learning algorithm with the joint scoring model of energy consumption and comfort as the goal, and dynamically adjusts the weight coefficients α and β. The weight coefficients α and β of the joint scoring model of energy consumption and comfort are dynamically adjusted according to the season type and usage scenario. In the summer scenario, α:β = 6:4; in the winter scenario, α:β = 5:5; in the transitional season, α:β = 7:

3.

6. The intelligent building energy conservation and emission reduction digital twin management system according to claim 1, characterized in that: The edge node of the edge-cloud collaborative architecture issues real-time control instructions to air conditioners and lighting devices through the Modbus / OPC protocol. The cloud performs big data analysis and model training, and supports short, medium, and long-term energy consumption prediction.

7. A method applied to an intelligent building energy conservation and emission reduction digital twin management system as described in any one of claims 1-6, characterized in that: The method steps are as follows: 1: System initialization S1. Import the building information model and equipment parameters for building modeling, construct a three-dimensional visual digital twin base, establish a topology map for sensor deployment, and plan temperature and humidity monitoring nodes, carbon dioxide concentration monitoring nodes, light intensity monitoring nodes, and energy consumption monitoring nodes. S2. Deploy the Internet of Things sensor network for infrastructure installation, connect to the building automation control system, and complete the communication debugging of the Modbus protocol and the OPC protocol. Phase 2: Data synchronization S3. The real-time data flow sensor collects environmental parameters at a frequency of once per second. The edge node performs data cleaning, filters out noise data, and compresses redundant information. S4. Access historical data, load the energy consumption data of the past three years, integrate the application program interface of the meteorological bureau, and import the weather prediction data for the next three days. Phase 3: Dynamic prediction S5. The load prediction engine uses the long short-term memory neural network module to process time series data, combines the transformer module to capture spatial correlation, and outputs the load curves of the air conditioning system and the lighting system within the next hour, with a prediction confidence level exceeding 85%. Phase 4: Real-time regulation S6. Temperature dynamic balance S6.

1. The energy consumption threshold trigger mechanism monitors the operating power of the air conditioning system in real time. When the cooling power in summer exceeds 90% of the rated value and the heating power in winter exceeds 85% of the rated value, the dynamic temperature adjustment program is activated. When not exceeding the limit, it maintains the indoor environmental comfort standard defined by the International Organization for Standardization, ensuring that the predicted mean vote index is in the comfort interval of -0.5 to +0.

5. S6.

2. Temperature dynamic adjustment process Summer operation mode: Initially set the target temperature of the air conditioner to 26 degrees Celsius. If the energy consumption continues to exceed the standard, the upper limit of the temperature allowance will be increased by 0.5 degrees Celsius every 30 minutes, up to a maximum of 28 degrees Celsius. During the temperature adjustment period, continuously monitor the predicted mean vote index. When it is detected that the index in a local area exceeds ±0.5, automatically increase the air supply speed to compensate for thermal comfort. Winter operation mode: Initially set the target temperature to 22 degrees Celsius. If the energy consumption exceeds the standard, the lower limit of the temperature allowance will be reduced by 0.5 degrees Celsius every 30 minutes, down to a minimum of 20 degrees Celsius. At the same time, by adjusting the air outlet angle, the hot air preferentially covers the personnel activity area to avoid comfort loss caused by the overall temperature drop. S6.

3. After the cross-system linkage control temperature adjustment instruction is generated, it is sent to the air-conditioning control system in real time through the Modbus protocol, and the lighting brightness in the corresponding area is synchronously reduced to 70% of the reference value to reduce the interference of equipment heat generation on temperature adjustment; S7. Fresh air optimization control S7.

1. The fresh air optimization control strategy is hierarchically controlled according to the carbon dioxide concentration. When the indoor carbon dioxide concentration is lower than 800 ppm, the minimum opening of the fresh air valve is maintained at 30% to reduce the fresh air treatment load of the air-conditioning system. When the carbon dioxide concentration rises to the range of 800 to 1200 ppm, the system linearly adjusts the opening of the fresh air valve to 80% within 15 minutes, and at the same time increases the air-conditioning cooling capacity to offset the heat load brought by the newly added fresh air. When the carbon dioxide concentration exceeds 1200 ppm, the fresh air valve is immediately fully opened and the emergency ventilation mode is started, and the lighting equipment in non-critical areas is synchronously turned off to prevent instantaneous grid overload; S7.

2. The energy efficiency optimization mechanism forcibly enables the minimum fresh air volume mode during the low occupancy period from 23:00 to 6:00 the next day, which will limit the opening of the fresh air valve and reduce the energy consumption of the air-conditioning system; Phase 5: Equipment collaboration S8. Air-conditioning system optimization S8.

1. The start-stop strategy of the chiller pre-starts the chiller 30 minutes in advance according to the predicted load; S8.

2. The variable frequency control of the water pump dynamically adjusts the water pump speed according to the end differential pressure to match the actual cooling demand; S9. The lighting brightness gradient adjustment strategy dynamically adjusts the brightness gradient and color temperature of the lighting system according to the natural light intensity, personnel distribution and scene requirements. On sunny days, the artificial lighting brightness is reduced to half of the reference value, and the natural light compensation is used to reduce the artificial lighting energy consumption. The cross-system linkage strategy coordinates the operation of the air-conditioning and lighting systems to achieve equipment-level energy saving; Phase 6: Closed-loop optimization S10. Error correction compares the predicted value with the actual value every hour and calculates the mean absolute percentage error; when the mean absolute percentage error exceeds 10%, trigger model retraining and update the weight parameters of the long short-term memory neural network using incremental learning; S11. Policy evolution mechanism S11.

1. The reinforcement learning-driven system starts the policy optimization cycle every 24 hours to explore the best balance point between energy consumption and comfort through trial-and-error learning; S11.

2. Dynamic weight adjustment automatically adjusts the proportion relationship between the energy consumption optimization weight and the human comfort weight according to the real-time operation data. When the energy consumption exceeds the standard for three consecutive days, the energy consumption optimization weight is increased to a maximum of 0.8 to prioritize the energy-saving goal. When the user complaint rate exceeds 5%, the human comfort weight is increased to a maximum of 0.7 to focus on comfort guarantee; S11.

3. Policy iteration process 1. Exploration: Simulate and generate 300 groups of candidate control schemes in the virtual twin, including the air-conditioning start-stop strategy and the lighting brightness gradient; 2. Verification and evaluation: Based on the joint scoring model of energy consumption and comfort, screen the top 10% of the strategies; 3. Practical application: Inject the selected strategies into the real building control system for operation and continuously monitor the actual effect for 8 hours; 4. Policy solidification: Standardize and package the strategies with a stable increase in the comprehensive score exceeding 2% and update them to the system policy library; Phase 7: Cross-layer collaboration S12. Edge interacts with the cloud Edge side: Execute emergency regulation with a delay of less than 50 milliseconds. When the temperature rises suddenly, perform a rapid refrigeration response; Cloud side: Conduct global policy optimization every day at dawn, generate weekly energy efficiency improvement plans, equipment maintenance plans, and load distribution strategies; S13. Digital twin synchronously updates the virtual model status every ten minutes, supports virtual reality and augmented reality visual monitoring, and historical operations support full traceability. The system regulation process at any time point can be replayed; Phase 8: Long-term evolution S14. The system upgrades by importing new device parameters every month, expanding the coverage of the digital twin model, and updating the artificial intelligence prediction model every quarter, incorporating the latest climate pattern data and user behavior characteristics; S15. The energy efficiency report automatically generates multi-dimensional analysis reports, including the proportion of sub-item energy consumption, the contribution degree of energy-saving measures, the comfort compliance rate, and the evaluation of equipment operation efficiency.

Images