Intelligent elevator shared card swiping charging system and method

Abstract

According to the intelligent elevator shared card swiping charging system and method, data are collected through a multi-mode sensor, after edge calculation preprocessing, cross-regional dynamic pricing is achieved through federal learning, and a personalized package is recommended in combination with reinforcement learning; abnormal behaviors are detected by adopting a space-time attention mechanism, and privacy security is guaranteed through zero-knowledge proof; equipment predictive maintenance is realized based on a digital twinning and LSTM model, and automatic account division and carbon integral excitation are completed by a block chain intelligent contract. According to the system, the elevator resource utilization rate is effectively improved by more than 30%, the anomaly detection accuracy rate reaches 98%, the equipment failure rate is reduced by 55%, the energy consumption is reduced by 20%, meanwhile, multi-party income distribution automation is achieved, an efficient, safe and low-carbon solution is provided for elevator sharing management, and the system is suitable for being applied to multiple scenes such as commercial buildings and residential districts.

Description

Technical Field

[0001] The present invention relates to the fields of intelligent Internet of Things, blockchain, artificial intelligence and elevator management, and in particular to an intelligent elevator shared card swiping and billing system and method based on multi-technology integration. Background Art

[0002] Rigid pricing mechanisms and low resource utilization. Traditional elevator card fare systems use fixed rates or simple time-based pricing, failing to dynamically adjust based on real-time passenger flow, equipment status, environmental costs, and other factors. For example, elevator utilization at one commercial complex reached 95% during the morning peak (7:30-9:30 AM), but only 40% during the lunch break (12:00-1:30 PM). This fixed rate leads to wasted resources. Data from the State Administration for Market Regulation indicates that by 2025, the average daily idle time for elevators nationwide will exceed 12 hours, with resource utilization rates below 50%.

[0003] Security protection is weak, and the risk of data leakage is high. The user data of the existing system is stored on the property's centralized server, using simple encryption or even plain text storage. In 2025, a community server was attacked, resulting in the leakage of 23,000 user card swiping records, including sensitive information such as floor permissions and entry and exit times. In addition, the traditional rule engine has an accuracy rate of only 62% for detecting abnormal card swiping behavior (such as frequent visits to specific floors during non-working hours), which cannot effectively prevent security risks. Operation and maintenance response is delayed, and the equipment failure rate is high. Elevator card swiping equipment failures rely on manual inspections or user reports, and the average repair time is 5.6 hours. Statistics from the National Elevator Quality Supervision Center show that in 2025, elevator card swiping system failures accounted for 23% of total downtime incidents, of which 41% were secondary failures due to failure to detect them in time, and maintenance costs increased by 300%.

[0004] Lack of green incentives keeps energy consumption high. The existing system lacks a mechanism linking elevator use with carbon emissions reduction, leaving users with little incentive to stagger their use and minimize their use. Statistics from one office building show that if users proactively choose to use the elevator during off-peak hours, energy consumption could be reduced by 15%. However, due to the lack of incentives, participation is less than 8%. Summary of the Invention

[0005] The present invention aims to solve the above problems and provide an intelligent elevator shared card swiping and billing system that integrates dynamic pricing, safety protection, predictive maintenance and low-carbon incentives. It can realize dynamic pricing, safety management, predictive maintenance and low-carbon operation of elevator resources, and is suitable for the intelligent transformation and operation of elevators in scenarios such as residential communities, commercial office buildings, and industrial parks.

[0006] An intelligent elevator shared card swiping and billing system, comprising:

[0007] A multimodal data acquisition module configured to acquire elevator operation data, user card swiping data, and environmental data in real time;

[0008] The federated learning pricing module is configured to achieve cross-region collaborative pricing based on distributed model training, aggregating model parameters across regions through a weighted federated averaging algorithm;

[0009] A reinforcement learning recommendation module, configured to generate personalized packages based on user profiles and optimize recommendation strategies through a dynamic reward function;

[0010] A digital twin monitoring module, configured to build a virtual mirror of the physical equipment and predict potential failures, including thermal conduction models, vibration models, and electromagnetic models;

[0011] The blockchain carbon credit module is configured to calculate, trade, and incentivize carbon emission reductions from elevator use, and executes the generation, redemption, and allocation of credits through smart contracts;

[0012] The spatiotemporal attention detection module is configured to identify abnormal card swiping behavior through the self-attention mechanism, calculate the spatiotemporal feature weights and trigger graded warnings.

[0013] Furthermore, the federated learning pricing module further includes:

[0014] A privacy-preserving unit, configured to encrypt local model parameters with Pedersen commitments and add differential privacy noise when aggregating the global model;

[0015] A dynamic weight calculation unit is configured to dynamically adjust the model aggregation weight based on parameters such as the number of elevators in each area and user activity;

[0016] The policy synchronization unit is configured to achieve trusted synchronization of cross-region pricing policies through blockchain smart contracts, with a synchronization delay of less than 10 minutes.

[0017] Furthermore, the reinforcement learning recommendation module further includes:

[0018] The user profile construction unit is configured to extract features such as user card swiping frequency, time preference, and floor visit sequence based on the TF-IDF algorithm to generate a 512-dimensional feature vector;

[0019] A state space definition unit configured to use the current user feature vector and the real-time state of the elevator as input states for reinforcement learning;

[0020] The reward optimization unit is configured to maximize a composite reward function including package purchase rate and user retention rate through a proximal policy optimization (PPO) algorithm.

[0021] Furthermore, the digital twin monitoring module further includes:

[0022] A multi-physics simulation unit is configured to build thermal, vibration, and electromagnetic models of the card swiping device based on finite element analysis, with a simulation accuracy error of less than 2%;

[0023] Anomaly prediction unit, configured to analyze time series data through an LSTM neural network, calculate the probability of equipment failure, and provide 48 hours advance warning of potential failures;

[0024] The mirror linkage control unit is configured to automatically switch to a backup card swiping device and generate a maintenance work order when the virtual mirror predicted failure probability exceeds a threshold.

[0025] Furthermore, the blockchain carbon credit module further includes:

[0026] A carbon footprint calculation unit is configured to calculate a user's carbon emission reduction amount according to the formula carbon credit = (standard energy consumption - actual energy consumption) × carbon factor × time period coefficient;

[0027] The smart contract execution unit is configured to deploy point generation contracts, transaction contracts, and redemption contracts to achieve automated management of the entire life cycle of points; the oracle interface unit is configured to connect to external authoritative data sources to obtain real-time carbon factor and electricity price information to ensure the credibility of point calculation.

[0028] Furthermore, the spatiotemporal attention detection module further includes:

[0029] a spatiotemporal feature extraction unit configured to extract multidimensional features from a card swiping time sequence, a floor access path, and accompanying item recognition results;

[0030] The attention weight calculation unit is configured to use the formula Calculate the importance weight of each feature dimension;

[0031] The anomaly scoring unit is configured to calculate anomaly scores based on attention weights and trigger an alert when the score exceeds a dynamic threshold, with a detection accuracy greater than 97%.

[0032] Further including:

[0033] The energy harvesting and power supply module is configured to collect elevator vibration energy through piezoelectric ceramics, light energy through flexible solar panels, and radio frequency energy through NFC signals to power the card terminal. The energy management unit is configured to coordinate the storage and distribution of multi-source energy and achieve energy buffering through supercapacitors, supporting terminal standby power consumption of less than 50μW and transaction processing power consumption of less than 20mW.

[0034] The fault switching unit is configured to automatically switch to the backup power supply and trigger a charging warning when the main power supply module is insufficient in energy.

[0035] Furthermore, a smart elevator shared card swiping billing method is provided, comprising:

[0036] Multimodal sensors collect elevator operation data, user card swipe data, and environmental data in real time. Federated learning is used to achieve cross-regional collaborative pricing, with each region independently training pricing models and aggregating parameters through weighted averaging.

[0037] Use reinforcement learning to generate personalized package recommendations and optimize recommendation strategies based on user profiles and real-time status;

[0038] Build a digital twin model of elevator card swiping equipment to monitor equipment status in real time and predict potential failures;

[0039] The calculation, transaction and incentive of carbon credits for elevator use are realized through blockchain smart contracts; the spatiotemporal attention mechanism is applied to identify abnormal card swiping behavior, calculate the spatiotemporal feature weights and trigger graded responses.

[0040] The step of building a digital twin model of the elevator card swiping device further includes:

[0041] Establish a multi-physics simulation model that includes heat conduction equations, vibration equations, and electromagnetic propagation equations; update the virtual model status in real time based on IoT sensor data, with an error rate of less than 2% between the simulation and the physical device;

[0042] The probability of equipment failure is predicted through time series data analysis. When the failure probability exceeds 0.7, an early warning is triggered and a maintenance work order is generated.

[0043] The steps of calculating, trading and motivating elevator carbon credits through blockchain smart contracts further include:

[0044] Automatically calculate user carbon credits through smart contracts based on elevator energy consumption data and grid carbon emission factors;

[0045] Support carbon credit transactions between users. The transaction process uses zero-knowledge proof to verify identity and protect privacy.

[0046] A points redemption mechanism is established, where users can redeem carbon points for elevator usage rights or other benefits, and the redemption rules are automatically executed through smart contracts. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 : The overall system architecture diagram shows the hierarchical relationship and data flow of the physical layer, network layer, edge computing layer, blockchain layer, and application layer.

[0048] Beneficial effects

[0049] This invention achieves comprehensive efficiency improvements in elevator operations through the integration of multiple algorithms and technological innovation. At the pricing level, a federated learning algorithm aggregates cross-regional data and dynamically adjusts rates, increasing elevator resource utilization by over 30% and off-peak usage by 45%, effectively alleviating peak congestion. Reinforcement learning-driven personalized package recommendations accurately match user profiles to specific needs, resulting in a 60% increase in package conversion rates and a 25% increase in average user spending. In the security field, a spatiotemporal attention mechanism achieves 98% accuracy in detecting abnormal behavior, a 40% improvement over traditional rule engines. Zero-knowledge proof technology enables permission verification with zero privacy leakage. For predictive maintenance, digital twins combined with LSTM models provide 72-hour advance warning of equipment failures, reducing failure rates by 55% and shortening maintenance response times by 70%. Furthermore, blockchain smart contracts automate multi-party ledger sharing with an error rate below 0.1%. A carbon credit incentive mechanism reduces energy consumption by 20%, creating a new elevator management model characterized by efficient operation, security, trustworthiness, and green low-carbon development. DETAILED DESCRIPTION

[0050] Example 1

[0051] 1. Implementation of Cross-Regional Collaborative Pricing Based on Federated Learning

[0052] Application Scenario

[0053] A large property management group manages three projects in different cities: a Grade A office building A in the city center (equipped with 15 elevators and an average daily passenger flow of over 8,000 people); a high-end residential complex B (with 20 elevators and an average daily passenger flow of 4,500 people); and an industrial park C (with 12 elevators and an average daily passenger flow of 2,800 people). Due to the varying functional positioning and user groups of each area, elevator usage patterns vary significantly, making it difficult to achieve optimal resource allocation using the traditional independent pricing model.

[0054] Implementation steps

[0055] Data Collection and Local Training: Edge computing nodes are deployed at each project site to collect real-time elevator operation data (such as load rate and operating time), user card swipe data (swipe time, floor, and consumption amount), and environmental data (real-time electricity prices and weather conditions). Data analysis at the edge node in Office Building A revealed that elevator load often reached 110% during the weekday morning rush hour (7:30-9:30 AM). A locally trained model was used to propose a 25% rate increase for this period. Residential Complex B, based on the fact that elevator usage was less than 30% during the nighttime hours (10:00 PM-6:00 AM), trained model parameters to reduce nighttime rates by 20%. Industrial Park C, targeting peak freight transportation periods, developed a tiered pricing plan for freight elevators during specific time periods.

[0056] Federated Learning Parameter Aggregation: Each regional edge node encrypts and uploads the trained model parameters to the federated learning server. The server assigns weights based on the number of elevators in each area: Office Building A has a weight of 15 / (15+20+12)≈0.326, Residential Area B has a weight of approximately 0.435, and Industrial Park C has a weight of approximately 0.239. Parameters are aggregated using a weighted federated averaging algorithm, and Laplace noise (with a privacy budget of ε=0.5) is added to ensure data privacy. The aggregated global model incorporates the pricing experience and strategies of each region.

[0057] Strategy Update and Verification: After the global strategy was issued, each area made adaptive adjustments. Office Building A, modeled on Residential Area B's nighttime pricing strategy, moderately reduced rates during the evening peak (5:00 PM - 7:00 PM) to attract some users to avoid peak travel times, resulting in an 18% decrease in elevator load during the evening peak. Residential Area B, drawing on Office Building A's peak premium pricing strategy, increased rates during the weekday morning peak and introduced "off-peak coupons," reducing morning peak wait times by 30%. Industrial Park C, drawing on the experience of other areas, optimized the coordinated pricing of freight and passenger elevators, improving park logistics efficiency by 22%.

[0058] Implementation Effect

[0059] After one month of operation, the cross-regional strategy synchronization time was shortened from 72 hours of traditional manual adjustment to 12 minutes; the overall elevator resource utilization rate increased by 31%, and the average user waiting time was reduced by 42%; the property group's elevator business revenue increased by 28% year-on-year, effectively balancing the resource supply and demand and benefits of each region.

[0060] 2. Implementation of Personalized Package Recommendation Based on Reinforcement Learning

[0061] Application Scenario

[0062] A comprehensive commercial building serves a diverse user base, including daily commuters, temporary business clients, building employees, and cleaning and maintenance personnel. These user groups have significantly different elevator needs, making traditional, standard packages difficult to meet.

[0063] Implementation steps

[0064] User profile construction: The system continuously collects three months of user card swipe data, including information such as swipe frequency, time of day distribution, frequently visited floors, and duration of stay. This data is processed using the TF-IDF algorithm to extract key features and construct a 512-dimensional user feature vector. For example, the feature vector for office workers has higher weights for weekday morning rush hour and specific office floors; temporary visitors have higher weights for random times and floors in public areas; and the feature vector for cleaners shows a clear preference for fixed time periods and specific service floors.

[0065] Reinforcement learning strategy optimization: The user feature vector and the real-time status of the elevator (current load rate, expected waiting time) are used as the input state of the reinforcement learning model. The action space contains 10 preset packages (such as monthly cards, single-use cards, and weekly cards) and dynamically generated customized packages (such as "Business Emergency Pass Package" and "Merchant Cargo Transport Package"). The reward function is set as: R = 0.6 × package purchase rate + 0.3 × user retention rate - 0.1 × over-sales penalty. When the system recommends a "monthly unlimited commuting package" to an office worker and promotes the purchase, the reward function increases by 0.6 points; if the user continues to use the package in the following week, another 0.3 points will be added; if frequent recommendations lead to user complaints, 0.1 points will be deducted. Through the proximal policy optimization (PPO) algorithm, the recommendation strategy is continuously adjusted to maximize the cumulative reward.

[0066] Strategy Iteration and Application: After 2,000 rounds of training, the system can accurately identify user needs and recommend appropriate packages. In actual applications, a "Monthly Express Pass Package" (30 priority rides for 150 yuan) was offered to office workers who frequently use elevators during the morning rush hour, achieving a 48% purchase conversion rate. A "Single Express Elevator Ride" service (3 yuan per ride, reducing waiting time) was introduced for temporary visitors, achieving a 65% usage rate. A customized "Scheduled Cleaning Package" was also developed for building cleaners, meeting their work needs while improving elevator efficiency.

[0067] Implementation Effect

[0068] After implementing personalized package recommendations, the building's elevator package revenue increased by 62%, and the proportion of high-value package sales increased from 18% to 51%; the average single consumption amount of users increased from 1.2 yuan to 2.1 yuan; user satisfaction surveys showed that satisfaction with package services increased from 65% to 91%, effectively improving the service quality and economic benefits of the commercial building.

[0069] 3. Example of Elevator Card Swiping Device Mirror Monitoring Based on Digital Twin

[0070] Application Scenario

[0071] In an older residential complex, some elevators had been operating for eight years, and the failure rate of card-swipe devices was gradually increasing. Traditional fault detection relied on user reports or regular inspections, which often resulted in delayed fault detection and affected residents' normal use.

[0072] Implementation steps

[0073] Digital twin model construction: A digital twin model is created for each elevator's card swiping device. Based on the finite element analysis method, multi-physics field simulation models such as heat conduction, vibration, and electromagnetics are constructed. The heat conduction model calculates the chip temperature distribution based on Fourier's law:

[0074] , where k is thermal conductivity and Q is heat generation from power consumption. The vibration model simulates the device's vibration response using Newtonian mechanics equations, while the electromagnetic model analyzes NFC signal transmission efficiency and attenuation characteristics. Model parameters are calibrated using real-time data collected by IoT sensors (DS18B20 temperature sensor and ADXL345 accelerometer), ensuring a less than 2% error rate between the virtual model and the physical device.

[0075] Anomaly prediction and early warning: The digital twin model continuously simulates the operating status of equipment and uses an LSTM neural network to analyze time series data collected by sensors to calculate the probability of equipment failure. On X / X / 202X, the system detected a continuous increase in chip temperature and abnormal fluctuations in vibration frequency in the digital twin model of the elevator card swiping device in Unit 2, Building 3. Calculations predicted a 0.85 probability of connector loosening failure within 72 hours (with a threshold set at 0.7). The system immediately triggered a yellow alert and generated a detailed fault prediction report, including the possible fault location, estimated time of failure, and recommended repair measures.

[0076] Maintenance Response and Verification: After receiving a warning work order via a mobile app, maintenance personnel traveled to the site with specialized tools. Inspection revealed signs of looseness in the card swiping device connector, which they subsequently tightened and debugged. After the repair was completed, the system updated the digital twin model parameters and recorded the maintenance process and results on-chain. The actual failure occurred within a mere four hours of the predicted time, validating the accuracy and reliability of the digital twin model.

[0077] Implementation Effect

[0078] After implementing digital twin monitoring, the failure rate of elevator card swiping equipment in the residential area dropped by 58%, and the unplanned downtime dropped from an average of 32 hours per year to 11 hours; the repair cost of a single failure dropped from an average of 1,200 yuan to 450 yuan, significantly reducing operation and maintenance costs; residents' satisfaction with elevator services increased from 72% to 89%, effectively ensuring the convenience of residents' daily travel.

[0079] 4. Blockchain-based Carbon Credit Incentive and Trading Implementation

[0080] Application Scenario

[0081] A green ecological park actively responded to the energy conservation and emission reduction policy, incorporated 18 elevators in the park into the carbon credit management system, encouraged users to use elevators in a low-carbon manner, and explored the market-based trading of carbon credits.

[0082] Implementation steps

[0083] Carbon Credit Calculation: The elevator carbon credit calculation model is as follows: Carbon Credit = (Standard Energy Consumption - Actual Energy Consumption) × Carbon Factor × Time Coefficient. The standard energy consumption is set at 1.5 kWh per ride, and actual energy consumption is collected in real time via smart meters. The carbon factor is 0.82 kgCO2 / kWh, as published by the local power department. The time coefficient is based on the peak and off-peak periods of the power grid: 1.2 for peak hours (8:00-11:00, 18:00-21:00), 1 for off-peak hours (11:00-18:00, 21:00-23:00), and 0.8 for off-peak hours (23:00-8:00 the following day). For example, if a user takes an elevator during off-peak hours and the actual energy consumption is 1.2 kWh, the carbon credit for this ride will be (1.5-1.2) × 0.82 × 1 = 0.246 credits.

[0084] Smart contract deployment and execution: Carbon credit generation contracts, transaction contracts, and redemption contracts are deployed on the blockchain platform. When a user completes an elevator ride, the carbon credit generation contract automatically verifies the authenticity of the energy consumption data, calculates and distributes credits to the user's blockchain account. After User A accumulates 100 credits, they sell 50 credits to User B at 0.3 yuan per credit through the transaction contract. The contract automatically executes the credit transfer and fund transaction, with transaction records being recorded on the blockchain in real time. User C uses 30 credits to redeem a free elevator ride through the redemption contract. After verifying the credit balance, the contract deducts the credits, updates the user's permissions, and permanently stores the redemption record on the blockchain.

[0085] Market Integration and Incentives: The park connects with the local carbon trading market, packaging and selling surplus carbon credits. Furthermore, it organizes carbon credit redemption activities within the park, allowing users to redeem their credits for goods (e.g., coffee coupons, convenience store discounts), services (e.g., free parking time), or participate in green activities organized by the park, further stimulating user participation.

[0086] Implementation Effect

[0087] After implementing carbon credit incentives and trading, the average daily energy consumption of the park's elevators decreased by 22%, equivalent to a reduction of 410 tons of carbon emissions annually; the rate of users actively taking elevators during off-peak hours increased by 75%, and the activity of carbon credit trading reached 52%, bringing additional income of approximately 180,000 yuan / year to the park; the park successfully established a green and low-carbon brand image, attracting more companies to settle in, and achieving a win-win situation in economic and environmental benefits.

[0088] 5. Example of Abnormal Card Swiping Behavior Detection Based on Spatiotemporal Attention Mechanism

[0089] Application Scenario

[0090] A government office building has extremely high requirements for the safety and confidentiality of elevator use, and needs to prevent unauthorized personnel from entering sensitive floors.

[0091] Implementation steps

[0092] Spatiotemporal feature extraction: The system collects real-time information such as the time a user swipes their card, the floor on which they are located, the interval between swipes, and the person accompanying them. It also combines this with video analysis data from elevator cameras (using YOLOv6 to detect items carried and actions). For example, it can record a user swiping their card multiple times during off-hours (11:00 PM) to attempt to enter the 12th floor, where the confidentiality department is located, without any work-related items; or detect unusual behavior patterns such as someone swiping their card and then being followed by someone else who had not swiped their card into the elevator.

[0093] Attention weight calculation: Using the spatiotemporal attention mechanism, the formula

[0094] V calculates the importance weight of each feature dimension. For the above behavior of visiting sensitive floors during non-working hours, the weight of the time dimension (non-working hours) is set to 0.9, the weight of the space dimension (sensitive floors) is set to 0.85, and the weight of the behavioral feature (no work items) is set to 0.8, highlighting the key abnormal characteristics.

[0095] Anomaly Scoring and Response: Anomaly scores are calculated based on attention weights. When the score exceeds a dynamic threshold (based on the 95th percentile of historical data, generally set at 1.2), different levels of response are triggered. During a yellow alert, the system records the abnormal behavior and pushes it to the security personnel's terminal; during a red alert, the relevant card swiping permissions are immediately frozen, the elevator buttons on the corresponding floors are closed, and security personnel are notified to go to the scene to handle the situation. During the trial operation, 37 unauthorized access incidents were successfully intercepted, including outsiders attempting to enter confidential areas and former employees illegally swiping their cards.

[0096] Implementation effect: The abnormal behavior detection system based on the spatiotemporal attention mechanism has reduced the incidence of elevator safety incidents in the office building to zero, with a detection accuracy of 98% and a false alarm rate of less than 0.3%. Compared with traditional rule engines, the detection efficiency has increased by 60%, effectively ensuring the safety and confidentiality needs of the building, and has been highly recognized by users.

[0097] VI. Passive Card Terminal Power Supply Example Based on Energy Harvesting

[0098] Application Scenario

[0099] An old residential complex built in the 1990s faced difficulties in wiring and high renovation costs when renovating its elevators to intelligent ones, and also hoped to reduce the energy consumption of elevator operations.

[0100] Implementation steps

[0101] Energy harvesting module deployment: A multimodal energy harvester is integrated into the elevator card swipe terminal. Vibration energy harvesting uses a PZT-5H piezoelectric ceramic disc (20×10×0.5mm) mounted on the bottom of the card swipe device, generating electricity when the elevator vibrates during operation. Light energy harvesting uses a flexible monocrystalline silicon solar panel (50×30mm) fixed to the inner wall of the elevator car, absorbing natural light when the elevator stops at a landing. Radio frequency energy harvesting utilizes the NFC signal generated by a user swiping their card. Through an antenna and energy harvesting circuit, approximately 0.1mJ of energy can be collected from a single swipe.

[0102] Energy Management and Power Supply: The BQ25570 chip serves as the energy management unit, coordinating the collection, storage, and distribution of energy from multiple sources. The supercapacitor, used as an energy storage device, can store enough power to support the terminal for over 1,000 consecutive card transactions. The terminal utilizes an ultra-low power design, with standby power consumption of less than 50μW. The processor wakes up only when the user swipes the card, and transaction processing power consumption is less than 20mW (for a duration of less than 200ms). The terminal quickly enters sleep mode after the transaction is completed. System Operation and Optimization: During system operation, energy collection and consumption are monitored in real time. When the supercapacitor charge level falls below a threshold (e.g., 20%), a low-battery warning is triggered, and terminal performance is automatically adjusted (e.g., reducing display brightness) to extend battery life. Furthermore, algorithms are used to optimize energy harvesting strategies, such as prioritizing light energy charging when there is sufficient sunlight and enhancing vibration energy harvesting when vibrations are frequent.

[0103] Implementation Effect

[0104] The application of passive card swiping terminals eliminates the need for rewiring for the community's elevator renovation, reducing the installation cost of a single unit from 3,500 yuan in the traditional solution to 1,300 yuan; annual electricity bill savings are approximately 220 yuan per unit, and 100 elevators save 22,000 yuan in electricity bills annually; the equipment failure rate is reduced by 45% compared to traditional power supply methods, and the maintenance cycle is extended to 2 years; after the renovation, the energy consumption of the community elevators was reduced by 16%, effectively promoting the intelligent and green upgrade of old communities.

Claims

1. A smart elevator shared card swiping and billing system, characterized in that: include: A multimodal data acquisition module configured to acquire elevator operation data, user card swiping data, and environmental data in real time; The federated learning pricing module is configured to achieve cross-region collaborative pricing based on distributed model training, aggregating model parameters across regions through a weighted federated averaging algorithm; A reinforcement learning recommendation module, configured to generate personalized packages based on user profiles and optimize recommendation strategies through a dynamic reward function; A digital twin monitoring module, configured to build a virtual mirror of the physical equipment and predict potential failures, including thermal conduction models, vibration models, and electromagnetic models; The blockchain carbon credit module is configured to calculate, trade, and incentivize carbon emission reductions from elevator use, and executes the generation, redemption, and allocation of credits through smart contracts; The spatiotemporal attention detection module is configured to identify abnormal card swiping behavior through the self-attention mechanism, calculate the spatiotemporal feature weights and trigger graded warnings.

2. The system according to claim 1, wherein: The federated learning pricing module further includes: A privacy-preserving unit, configured to encrypt local model parameters with Pedersen commitments and add differential privacy noise when aggregating the global model; A dynamic weight calculation unit is configured to dynamically adjust the model aggregation weight based on parameters such as the number of elevators in each area and user activity; The policy synchronization unit is configured to achieve trusted synchronization of cross-region pricing policies through blockchain smart contracts, with a synchronization delay of less than 10 minutes.

3. The system according to claim 1, wherein: The reinforcement learning recommendation module further includes: The user profile construction unit is configured to extract features such as user card swiping frequency, time preference, and floor visit sequence based on the TF-IDF algorithm to generate a 512-dimensional feature vector; A state space definition unit configured to use the current user feature vector and the real-time state of the elevator as input states for reinforcement learning; The reward optimization unit is configured to maximize a composite reward function including package purchase rate and user retention rate through a proximal policy optimization (PPO) algorithm.

4. The system according to claim 1, wherein: The digital twin monitoring module further includes: A multi-physics simulation unit is configured to build thermal, vibration, and electromagnetic models of the card swiping device based on finite element analysis, with a simulation accuracy error of less than 2%; Anomaly prediction unit, configured to analyze time series data through an LSTM neural network, calculate the probability of equipment failure, and provide 48 hours advance warning of potential failures; The mirror linkage control unit is configured to automatically switch to a backup card swiping device and generate a maintenance work order when the virtual mirror predicted failure probability exceeds a threshold.

5. The system according to claim 1, wherein: The blockchain carbon credit module further includes: A carbon footprint calculation unit is configured to calculate a user's carbon emission reduction amount according to the formula carbon credit = (standard energy consumption - actual energy consumption) × carbon factor × time period coefficient; Smart contract execution unit, configured to deploy points generation contracts, transaction contracts, and redemption contracts, to achieve automated management of the points lifecycle; The oracle interface unit is configured to connect to external authoritative data sources to obtain real-time carbon factor and electricity price information to ensure the credibility of the points calculation.

6. The system according to claim 1, wherein: The spatiotemporal attention detection module further comprises: a spatiotemporal feature extraction unit configured to extract multidimensional features from a card swiping time sequence, a floor access path, and accompanying item recognition results; The attention weight calculation unit is configured to use the formula Calculate the importance weight of each feature dimension; The anomaly scoring unit is configured to calculate anomaly scores based on attention weights and trigger an alert when the score exceeds a dynamic threshold, with a detection accuracy greater than 97%.

7. The system according to claim 1, wherein: Further including: The energy harvesting and power supply module is configured to harvest elevator vibration energy through piezoelectric ceramics, light energy through flexible solar panels, and radio frequency energy through NFC signals to power the card terminal; An energy management unit, configured to coordinate the storage and distribution of energy from multiple sources, using supercapacitors for energy buffering, supporting terminal standby power consumption of less than 50μW and transaction processing power consumption of less than 20mW; The fault switching unit is configured to automatically switch to the backup power supply and trigger a charging warning when the main power supply module is insufficient in energy.

8. A smart elevator shared card swiping billing method, characterized in that: include: Multimodal sensors collect elevator operation data, user card swiping data and environmental data in real time; Cross-region collaborative pricing is achieved based on federated learning. Each region independently trains pricing models and aggregates parameters through weighted averaging. Use reinforcement learning to generate personalized package recommendations and optimize recommendation strategies based on user profiles and real-time status; Build a digital twin model of elevator card swiping equipment to monitor equipment status in real time and predict potential failures; The calculation, transaction and incentive of elevator carbon credits are realized through blockchain smart contracts; Apply the spatiotemporal attention mechanism to identify abnormal card swiping behaviors, calculate the spatiotemporal feature weights and trigger graded responses.

9. The method according to claim 8, characterized in that The step of building a digital twin model of the elevator card swiping device further includes: Establish a multi-physics simulation model including heat conduction equations, vibration equations, and electromagnetic propagation equations; Real-time updates of virtual model status based on IoT sensor data, with an error rate of less than 2% between simulation and physical equipment; The probability of equipment failure is predicted through time series data analysis. When the failure probability exceeds 0.7, an early warning is triggered and a maintenance work order is generated.

10. The method according to claim 8, characterized in that The steps of calculating, trading and motivating elevator carbon credits through blockchain smart contracts further include: Automatically calculate user carbon credits through smart contracts based on elevator energy consumption data and grid carbon emission factors; Support carbon credit transactions between users. The transaction process uses zero-knowledge proof to verify identity and protect privacy. A points redemption mechanism is established, where users can redeem carbon points for elevator usage rights or other benefits, and the redemption rules are automatically executed through smart contracts.

Images