A charging pile management method and system based on the Internet of Things

Abstract

The application relates to the technical field of Internet of Things, and discloses a charging pile management method and system based on Internet of Things. The method comprises the following steps: collecting charging pile parameters by an Internet of Things sensor to obtain original data; distributing an authorization level through user identity verification; calculating a warning signal through cable temperature anomaly detection; performing clustering analysis on user charging behaviors; performing load optimization scheduling through LSTM prediction; and calculating a bill through multi-period dynamic electricity price calculation. The application realizes real-time monitoring and anomaly detection on the internal cable temperature of a charging pile through Internet of Things technology, identifies potential safety hazards in a timely manner, and realizes intelligent charging load prediction and scheduling in combination with user behavior characteristics.

Description

Technical Field

[0001] This application relates to the field of Internet of Things (IoT) technology, and in particular to a charging pile management method and system based on IoT. Background Technology

[0002] With the rapid development of the electric vehicle industry, the construction of charging infrastructure is becoming increasingly important. Traditional charging pile management systems typically only provide basic charging functions, relying mainly on simple control systems to manage the charging process. These systems generally use wired communication for data transmission and a fixed electricity price billing model, lacking intelligent and personalized features. In existing technologies, the operational status monitoring of charging piles mainly relies on manual inspections, resulting in discontinuous data collection and difficulty in achieving real-time monitoring; user authentication often uses single identification methods, leading to low security; charging resource allocation is usually based on a first-come, first-served principle, without considering the differentiated needs of users; electricity price calculation methods are fixed and cannot be dynamically adjusted according to grid load conditions; and the maintenance of charging equipment is mainly reactive, making it difficult to predict fault occurrences.

[0003] However, traditional charging pile management methods have significant shortcomings: First, the lack of real-time temperature monitoring of key internal components such as cables makes it impossible to detect potential safety hazards in a timely manner; second, user data is stored in a scattered manner and lacks in-depth analysis, making it impossible to uncover user charging behavior characteristics and provide personalized services; third, the accuracy of charging load prediction is low and the scheduling of charging resources is unreasonable, resulting in uneven utilization of charging piles and long waiting times during peak periods; in addition, the billing system lacks flexibility and cannot implement differentiated pricing based on grid load and user level; finally, equipment maintenance is mostly handled after the fact, lacking predictive maintenance capabilities, which affects the reliability and service life of charging facilities. Summary of the Invention

[0004] This application provides a charging pile management method and system based on the Internet of Things (IoT), which is used to monitor and detect anomalies in the internal cable temperature of the charging pile in real time through IoT technology, identify potential safety hazards in a timely manner, and realize intelligent charging load prediction and scheduling by combining user behavior characteristics.

[0005] Firstly, this application provides an IoT-based charging pile management method, which includes: real-time data acquisition of the charging pile by collecting voltage, current, temperature, and power parameters through IoT sensors to obtain raw charging pile data; user identification and authorization management based on the raw charging pile data by verifying user information and assigning authorization levels through a multi-factor authentication mechanism to obtain user identity authorization data; and temperature anomaly detection of the internal cables of the charging pile based on the user identity authorization data and the raw charging pile data by comparing the measured temperature with a set temperature threshold and calculating the temperature rise rate to obtain charging safety data. Early warning signal; based on the charging safety early warning signal and the original data of the charging pile, feature extraction and cluster analysis are performed on user charging behavior. The user behavior feature vector is calculated by improving the K-Means clustering algorithm to obtain the user behavior clustering result; based on the user behavior clustering result and the original data of the charging pile, the charging load is predicted and scheduled. The load prediction curve is calculated by using the LSTM neural network model and the charging scheduling strategy is optimized to obtain the charging scheduling scheme; based on the charging scheduling scheme and the user identity authorization data, the electricity price is dynamically calculated in multiple time periods. The real-time charging rate is calculated by multiplying the benchmark electricity price by the time period coefficient and the fee accumulation processing is performed to obtain the user charging bill.

[0006] Secondly, this application provides an Internet of Things (IoT)-based charging pile management system, which includes:

[0007] The data acquisition module is used to collect real-time data from the charging pile. It collects voltage, current, temperature and power parameters through IoT sensors to obtain the raw data of the charging pile.

[0008] The authorization module is used to identify and authorize users based on the original data of the charging pile, verify user information and assign authorization levels through a multi-factor authentication mechanism, and obtain user identity authorization data.

[0009] The detection module is used to detect abnormal temperatures in the internal cables of the charging pile based on the user identity authorization data and the original data of the charging pile. It compares the measured temperature with the set temperature threshold and calculates the temperature rise rate to obtain a charging safety warning signal.

[0010] The analysis module is used to extract features and perform cluster analysis on user charging behavior based on the charging safety warning signal and the original data of the charging pile. It calculates user behavior feature vectors by improving the K-Means clustering algorithm to obtain user behavior clustering results.

[0011] The scheduling module is used to predict and schedule the charging load based on the user behavior clustering results and the original data of the charging piles. It calculates the load prediction curve through the LSTM neural network model and optimizes the charging scheduling strategy to obtain the charging scheduling scheme.

[0012] The calculation module is used to perform multi-period dynamic calculation of electricity price based on the charging scheduling scheme and the user identity authorization data, calculate the real-time charging rate by multiplying the benchmark electricity price by the period coefficient, and perform fee accumulation processing to obtain the user's charging bill.

[0013] Thirdly, an Internet of Things (IoT) based charging pile management device is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the IoT-based charging pile management device to execute the aforementioned IoT-based charging pile management method.

[0014] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored in the computer and, when executed on a computer, cause the computer to perform the above-described Internet of Things-based charging pile management method.

[0015] The technical solution provided in this application utilizes IoT sensors to collect real-time data from charging piles, enabling precise monitoring of key parameters such as voltage, current, temperature, and power, thus laying a data foundation for charging safety and efficient management. A multi-factor authentication mechanism combined with authorization level allocation not only improves system security but also implements differentiated service strategies, enhancing user experience while optimizing resource allocation efficiency. Personalized temperature threshold settings and real-time temperature rise rate calculations based on user identity authorization data construct a precise charging safety early warning mechanism, significantly reducing safety risks during charging and protecting the safety of charging equipment and user property. An improved K-Means clustering algorithm extracts and classifies user charging behavior features, revealing the charging patterns and preferences of different user groups. These refined user profiles provide a basis for subsequent optimized allocation of charging resources. The LSTM neural network model fully utilizes temporal characteristics and multi-dimensional features to achieve high-precision charging load prediction. Its recursive structure effectively captures long-term dependencies, improving prediction accuracy by more than 15% compared to traditional time series prediction methods, providing a reliable prediction foundation for intelligent scheduling strategies. The multi-period dynamic electricity pricing mechanism combines user levels with time-period coefficients to form a flexible price incentive system, effectively guiding users to charge during off-peak hours, reducing peak load on the power grid, and improving the overall operational efficiency of the charging network. Overall, this solution deeply integrates IoT technology with artificial intelligence algorithms, particularly the application of K-Means clustering in user behavior analysis and the advantages of LSTM neural networks in load forecasting, to jointly construct an adaptive, efficient, and safe charging pile management system. This integration not only enhances the intelligence level of electric vehicle charging infrastructure but also optimizes the utilization of power grid resources through peak shaving and valley filling. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of one embodiment of the charging pile management method based on the Internet of Things in this application.

[0018] Figure 2 This is a schematic diagram of one embodiment of the IoT-based charging pile management system in this application.

[0019] Figure 3 This is a schematic block diagram of the structure of the charging pile management device based on the Internet of Things in this embodiment of the invention. Detailed Implementation

[0020] This application provides a charging pile management method and system based on the Internet of Things. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0021] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the charging pile management method based on the Internet of Things in this application includes:

[0022] Step S101: Real-time data acquisition of the charging pile, collecting voltage parameters, current parameters, temperature parameters and power parameters through IoT sensors to obtain the raw data of the charging pile;

[0023] Step S102: Based on the original data of the charging pile, perform user identification and authorization management, verify user information and assign authorization level through a multi-factor authentication mechanism to obtain user identity authorization data;

[0024] Step S103: Based on user identity authorization data and original charging pile data, perform temperature anomaly detection on the internal cables of the charging pile, compare the set temperature threshold with the measured temperature and calculate the temperature rise rate to obtain a charging safety warning signal.

[0025] Step S104: Based on the charging safety warning signal and the original data of the charging pile, perform feature extraction and cluster analysis on the user's charging behavior, calculate the user behavior feature vector by improving the K-Means clustering algorithm, and obtain the user behavior clustering result;

[0026] Step S105: Based on the user behavior clustering results and the original data of the charging pile, predict and schedule the charging load, calculate the load prediction curve through the LSTM neural network model and optimize the charging scheduling strategy to obtain the charging scheduling scheme.

[0027] Step S106: Calculate the electricity price dynamically across multiple time periods based on the charging scheduling plan and user identity authorization data. Calculate the real-time charging rate by multiplying the benchmark electricity price by the time period coefficient and perform cumulative fee processing to obtain the user's charging bill.

[0028] It is understood that the executing entity of this application can be an IoT-based charging pile management system, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiments use a server as an example for illustration.

[0029] Specifically, the charging pile collects data in real time through internal IoT sensor modules. Voltage sensors collect input and output voltage data every 5 seconds to ensure real-time monitoring of voltage changes during charging. Current sensors also collect input and output current data at 5-second intervals to record current fluctuations during charging. Temperature sensors collect the internal temperature and charging interface temperature of the charging pile every 30 seconds to monitor temperature trends. Power sensors collect charging power data every 5 seconds to reflect the energy transfer status during charging. This raw data undergoes preliminary processing by the charging pile's built-in edge computing unit, including data verification, outlier filtering, and data compression, before being packaged into standard data packets. Each data packet contains a unique charging pile ID, a timestamp, various parameter values, and a checksum. The data packets are encrypted using the AES-256 encryption algorithm and transmitted to the central server via the IoT communication module. Upon receiving the data packets, the central server decrypts, verifies, and indexes and stores them according to the charging pile ID and timestamp, forming the charging pile's raw data. For example, when a user connects to a charging station and starts charging, the voltage sensor records an output voltage of 220V, the current sensor records an output current of 32A, the temperature sensor records a charging interface temperature of 35℃, and the power sensor records a charging power of 7.04kW. These data are encrypted and then transmitted to the central server, where data records are created in the database indexed by the charging station number "CP20230001" and the timestamp "2023-06-15 14:30:25".

[0030] The system performs user identification and authorization management based on the original data from charging piles. First, it receives user registration information via a mobile application and generates a unique user identifier code, including key information such as user name, contact information, vehicle information, and payment method. Upon arrival at a charging pile, the system uses NFC near-field communication technology to identify the electronic identity credential on the user's mobile phone, or scans a QR code on the charging pile and confirms identity within the application, or reads the user's dedicated charging card information via RFID, performing multiple identity verifications. User identity information is transmitted to the central server for verification using RSA asymmetric encryption. The system calculates a user credit score based on their historical charging records, credit rating, and account balance, mapping the score to different authorization levels: Diamond Member (90-100 points), Gold Member (80-89 points), Silver Member (70-79 points), and Regular Member (0-69 points). Users at different authorization levels enjoy different charging priorities, discount rates, and value-added services. For higher-level users, the system allows for scheduled charging. After user authentication is successful, the system automatically allocates the maximum charging power and allowed charging time based on the user's authorization level and the current charging pile load, forming user identity authorization data. For example, when user "Mr. Zhang" authenticates his identity at charging pile "CP20230001" via NFC, the system finds that the user has accumulated 50 charging sessions, an average charging volume of 25kWh, and a timely payment rate of 98%. Based on the scoring rules, his credit score is calculated to be 85 points, classifying him as a Gold Member, allocating a maximum charging power of 11kW, allowing a maximum continuous charging time of 5 hours, and enjoying a 9.3% discount.

[0031] The system detects abnormal temperatures in the internal cables of charging piles based on user authorization data and raw charging pile data. It extracts cable temperature data from the most recent 30 minutes from the raw charging pile data to construct a temperature time series. Temperature thresholds are personalized according to the user's authorization level by multiplying the base temperature threshold by a safety factor corresponding to the user's authorization level. The system divides the temperature data into time periods, calculating the average temperature value every 5 minutes and recording it as a temperature change data point. Then, it calculates the temperature difference between adjacent time points and divides it by the time interval to obtain the temperature rise rate for each time period. The temperature rise rate is compared with three preset intervals: normal rate, warning rate, and dangerous rate, to classify the temperature change level. Finally, based on the temperature change level and the difference between the current measured temperature and the personalized temperature threshold, a safety risk value is calculated, generating a charging safety warning signal. For example, the cable temperature data of charging pile "CP20230001" shows that within a continuous 30-minute period, the average temperature every 5 minutes was 42℃, 45℃, 47℃, 51℃, 54℃, and 58℃, respectively, with calculated temperature rise rates of 0.6℃ / min, 0.4℃ / min, 0.8℃ / min, 0.6℃ / min, and 0.8℃ / min. User "Mr. Zhang's" Gold Membership level corresponds to a safety coefficient of 0.95, a base temperature threshold of 60℃, and a personalized temperature threshold of 57℃. The current temperature of 58℃ exceeds the personalized temperature threshold, and the temperature rise rate of 0.8℃ / min in the last time period falls within the warning rate range. The system calculates a safety risk value of 0.85, triggering a high-level warning signal. Based on the charging safety warning signal and the original data of the charging pile, the system performs feature extraction and cluster analysis on the user's charging behavior. The system extracts the user's charging history records for the past 90 days from the original data of the charging pile, including data such as charging time, charging duration, charging amount, and charging frequency. These original features are processed over time, statistically analyzing the charging frequency distribution of users over seven days (Monday to Sunday) and within 24 hours each day, forming a seven-dimensional weekly vector and a 24-dimensional daily vector. Simultaneously, a user charging safety index is calculated based on charging safety warning signals, which is the ratio of the frequency of safety warnings triggered during a user's historical charging process to the total number of charging attempts. All features are normalized, mapping each dimension of data to the 0-1 interval to eliminate dimensional differences and obtain standardized feature vectors. The system employs an improved K-Means clustering algorithm, determining the optimal number of clusters by calculating the silhouette coefficient under different K values ​​and selecting the K value corresponding to the maximum silhouette coefficient. The silhouette coefficient is calculated as the average of the difference between the similarity of a sample to its own cluster and the similarity to other clusters, ranging from -1 to 1, with values ​​closer to 1 indicating better clustering performance. Initial cluster centers are selected using the K-means++ method to improve the stability of the clustering algorithm.The system performs iterative clustering calculations. It stops iterating when the change in cluster centers is less than a preset threshold of 0.001 or when the maximum number of iterations (50) is reached, thus obtaining the user behavior clustering results. For example, by performing feature extraction and clustering analysis on the charging data of 5,000 users, the system tried different cluster numbers from K=3 to K=8, and calculated the corresponding silhouette coefficients as 0.52, 0.68, 0.61, 0.55, 0.48, and 0.43, respectively. The system selected K=4, corresponding to the maximum silhouette coefficient of 0.68, as the optimal number of clusters, and finally divided users into four categories: "peak-hour charging users," "nighttime slow charging users," "weekend charging users," and "emergency fast charging users."

[0032] The system predicts and schedules charging load based on user behavior clustering results and raw charging pile data. Nearly 30 days of historical load data are organized into hourly data points and time-stamped to construct charging load time-series data. This data is then augmented with time features (hour, date, weekday / weekend identifier), weather features (temperature, humidity, weather conditions), and user behavior clustering ratio features as input variables, forming a multi-dimensional input feature matrix. The system employs a sliding window method, setting a 24-hour input window length and a 24-hour prediction target length to organize the data, resulting in input-output sample pairs. Based on these sample pairs, an LSTM (Long Short-Term Memory) neural network model is constructed. This model includes an input layer, two LSTM hidden layers, a Dropout layer, and a fully connected output layer. LSTM networks are specifically designed for processing time-series data, capturing long-term dependencies and making them suitable for load prediction tasks. After training, the system inputs the current 24-hour load data into the model to predict the regional charging load curve for the next 24 hours. Based on the prediction results, the system divides the prediction period into peak periods (load rate > 80%), off-peak periods (load rate 50%-80%), and low-peak periods (load rate < 50%), and calculates the scheduling priority weight of each charging pile to form a charging scheduling scheme. For example, based on historical data and an LSTM model, the system predicts that 8:00-10:00 AM and 5:00-7:00 PM the next day will be peak periods, 1:00-5:00 AM will be low-peak periods, and the remaining time periods will be off-peak periods. In a region with 10 charging piles, the system calculates the weight value of each charging pile according to the scheduling priority weight formula, prioritizing the allocation of new charging demand to charging piles with higher weight values, while guiding non-urgent charging demand to avoid peak periods and fill low-peak periods, thus achieving grid load balance. The system performs multi-period dynamic calculation of electricity prices based on the charging scheduling scheme and user authorization data. The system divides 24 hours into three periods: peak, off-peak, and low-peak, assigning period coefficients of 1.5, 1.0, and 0.7 respectively. Simultaneously, the system calculates tiered discounts based on user authorization data, setting discount rates of 0.9, 0.93, 0.95, and 1.0 for Diamond, Gold, Silver, and Regular members, respectively. The system calculates the real-time rate based on the base electricity price, time-period coefficient, and user discount coefficient: Real-time rate = Base electricity price × Time-period coefficient × User discount coefficient. During charging, the system calculates and records the cumulative cost for each kilowatt-hour charged, generating a real-time charging fee. After charging, the system generates an electronic bill, recording the start and end times of charging, electricity consumption for each time period, time-period rate, total electricity consumption, total cost, and discount amount. Finally, the system uses blockchain technology to store the transaction data, calculates the hash value of the bill data, and adds a timestamp to create an immutable transaction record, ensuring the authenticity and reliability of the charging bill.For example, user "Mr. Zhang" uses charging station "CP20230001" on weekdays, charging across both off-peak and peak hours, for a total of 30 kWh charged, with 20 kWh charged during off-peak hours and 10 kWh during peak hours. The local benchmark electricity price is 1.0 yuan / kWh, the off-peak coefficient is 0.7, and the peak coefficient is 1.5. As a gold member, he enjoys a discount rate of 0.93. The final calculated off-peak cost is 13.02 yuan (1.0 × 0.7 × 0.93 × 20), and the peak cost is 13.95 yuan (1.0 × 1.5 × 0.93 × 10), for a total cost of 26.97 yuan. The system generates an electronic bill and records the transaction data using blockchain technology to ensure the bill is tamper-proof.

[0033] In this embodiment, real-time data acquisition from charging piles via IoT sensors enables precise monitoring of key parameters such as voltage, current, temperature, and power, laying a data foundation for charging safety and efficient management. A multi-factor authentication mechanism combined with authorization level allocation not only improves system security but also implements differentiated service strategies, enhancing user experience while optimizing resource allocation efficiency. Personalized temperature threshold settings and real-time temperature rise rate calculations based on user identity authorization data construct a precise charging safety early warning mechanism, significantly reducing safety risks during charging and protecting the safety of charging equipment and user property. An improved K-Means clustering algorithm extracts and classifies user charging behavior features, revealing the charging patterns and preferences of different user groups. These refined user profiles provide a basis for subsequent optimized allocation of charging resources. The LSTM neural network model fully utilizes temporal characteristics and multi-dimensional features to achieve high-precision charging load prediction. Its recursive structure effectively captures long-term dependencies, improving prediction accuracy by more than 15% compared to traditional time series prediction methods, providing a reliable prediction foundation for intelligent scheduling strategies. The multi-period dynamic electricity pricing mechanism combines user levels with time-period coefficients to form a flexible price incentive system, effectively guiding users to charge during off-peak hours, reducing peak load on the power grid, and improving the overall operational efficiency of the charging network. Overall, this solution deeply integrates IoT technology with artificial intelligence algorithms, particularly the application of K-Means clustering in user behavior analysis and the advantages of LSTM neural networks in load forecasting, to jointly construct an adaptive, efficient, and safe charging pile management system. This integration not only enhances the intelligence level of electric vehicle charging infrastructure but also optimizes the utilization of power grid resources through peak shaving and valley filling.

[0034] In one specific embodiment, the process of performing step S101 may specifically include the following steps:

[0035] The input and output voltages of the charging pile are collected by a voltage sensor. The collected voltage data is sampled and stored every 5 seconds to obtain voltage parameters.

[0036] The charging pile's input and output currents are collected by a current sensor, and the collected current data is sampled and stored every 5 seconds to obtain current parameters.

[0037] The internal temperature of the charging pile and the temperature of the charging interface are collected by temperature sensors. The collected temperature data is sampled and stored every 30 seconds to obtain temperature parameters.

[0038] The power sensor collects the real-time charging power of the charging pile, and the collected power data is sampled and stored every 5 seconds to obtain the power parameters.

[0039] The voltage, current, temperature, and power parameters are encrypted using AES-256, and a unique identifier and timestamp of the charging pile are added to obtain an encrypted data packet.

[0040] The encrypted data packets are transmitted to the central server via the IoT communication module. The encrypted data packets are then decrypted, verified, and indexed for storage, thus obtaining the original data of the charging pile.

[0041] Specifically, sensors monitor the charging pile status in real time to achieve intelligent management and optimized control. Voltage sensors installed inside the charging pile need to cover both the input and output ends, employing a high-precision resistor voltage divider principle to convert high voltage into a measurable low voltage signal. Each charging pile is equipped with at least two sets of voltage sensors, monitoring the voltage status on the input and output sides respectively. The voltage sensors collect data every 5 seconds. After collection, the sensors convert the analog signal into a digital signal to form voltage parameters. These voltage parameters include four key data points: measurement time, input voltage value, output voltage value, and voltage stability score. The voltage stability score is calculated by determining the standard deviation of the most recent five sampled values ​​and is used to assess voltage fluctuations. Simultaneously, current sensors utilize the Hall effect principle, measuring the current during charging without disconnecting the circuit. Current sensors are also divided into input and output sets, monitoring the input current obtained by the charging pile from the grid and the current output to the electric vehicle, respectively. The current data is also sampled every 5 seconds, synchronously with the voltage sampling. The current parameters record three key data points: measurement time, input current value, output current value, and current change rate. The current change rate is calculated by the difference between two adjacent sampling values ​​and is used to monitor sudden changes in current.

[0042] Temperature sensors employ thermistor technology and are distributed in key areas inside the charging pile and at the charging interface. The internal temperature sensor primarily monitors the temperature status of core components such as the transformer and rectifier, while the interface temperature sensor focuses on monitoring temperature changes at the connection point between the charging plug and the vehicle. Temperature data is sampled every 30 seconds, lower than the sampling frequency for voltage and current, primarily because temperature changes are relatively slow and a high sampling frequency is unnecessary. Temperature parameters include four key data points: measurement time, internal temperature value, interface temperature value, and temperature anomaly flag. The temperature anomaly flag determines whether the current temperature is within a safe range based on a preset threshold. The power sensor multiplies the voltage and current signals using a multiplication circuit to directly measure the real-time power during charging. Power data is sampled every 5 seconds, synchronized with the voltage and current sampling. Power parameters record four key data points: measurement time, real-time power value, cumulative energy consumption, and power factor. The power factor is an important indicator of charging efficiency, calculated as the ratio of active power to apparent power.

[0043] All collected parameter data undergoes secure encryption before transmission. The encryption process employs the AES-256 algorithm, a symmetric encryption algorithm known for its high security and computational efficiency. Before encryption, the charging station's local processing unit adds a unique charging station identifier (e.g., "CP20240001") and a timestamp accurate to milliseconds (e.g., "2024-04-14 15:30:25.789") to each data packet to ensure data traceability. AES-256 encryption requires a 256-bit key, which is assigned by the central server during charging station initialization and updated periodically to enhance system security. The encryption process converts the raw data into ciphertext, ensuring that even if data is intercepted during transmission, the content cannot be deciphered without the correct key.

[0044] The encrypted data packets are transmitted to the central server via the IoT communication module. The communication module supports multiple network connection methods, primarily including 4G / 5G mobile networks and wired Ethernet, automatically selecting the optimal connection method based on the charging pile deployment environment. The transmission process uses the MQTT (Message Queuing Telemetry Transport) protocol, a lightweight publish / subscribe messaging protocol particularly suitable for data transmission from IoT devices. The MQTT protocol features low bandwidth consumption, a reliable message delivery mechanism, and support for unstable network environments. Upon receiving the encrypted data packets, the central server first performs a data integrity check, verifying whether the packets were corrupted during transmission. If the check passes, the server decrypts the data packets using the corresponding AES-256 key, restoring the original parameter data. The decrypted data undergoes secondary verification, including timestamp verification (ensuring data timeliness) and charging pile ID verification (ensuring reliable data source). Verified data is indexed and stored according to the charging pile ID and timestamp, using a distributed database structure to ensure high-concurrency read / write performance and secure data backup.

[0045] In one specific embodiment, the process of performing step S102 may specifically include the following steps:

[0046] User information is obtained based on the original data of the charging pile. User real-name registration information is received through a mobile application and a unique user identification code is generated to obtain basic user information.

[0047] Multiple authentication methods are used to verify the user's basic information, including NFC near-field communication technology, QR code scanning or RFID radio frequency identification, to obtain the authentication result.

[0048] The authentication result is processed by RSA asymmetric encryption, and the user's identity information is protected and transmitted through the encryption algorithm to obtain encrypted identity data;

[0049] Credit ratings are calculated for users based on encrypted identity data and historical charging records. The credit score is obtained by weighting and summing three indicators: cumulative number of charging times, average charging amount, and payment timeliness.

[0050] Users are assigned authorization levels based on their credit scores. By mapping credit scores to four levels—Diamond Member, Gold Member, Silver Member, and Regular Member—user authorization levels are obtained.

[0051] Based on the user's authorization level, the upper limit of charging power and the charging time are allocated to the user. Personalized charging parameters are calculated through a differentiated resource allocation algorithm for different user levels to obtain user identity authorization data.

[0052] Specifically, user information is obtained through raw data from charging stations, starting with a mobile application. Users need to download a dedicated charging station management application from their mobile app store and register with their real names. Real-name registration requires users to fill in basic personal information, including name, contact number, ID card number, vehicle brand, vehicle model, license plate number, and default payment method. After collecting this information, the application cross-verifies it with the public security department's identity information database through a backend data verification interface to ensure the authenticity and validity of the user's identity information. After successful verification, the application's backend server generates a unique user identifier (UUID) for the user, in the format of a 32-bit hexadecimal string, such as "a1b2c3d4-e5f6-g7h8-i9j0-k1l2m3n4o5p6". This identifier is bound to the user's real-name information, forming a basic user information data packet, which is stored in the user data center.

[0053] When a user completes registration and arrives at a charging station to charge, multiple identity verifications of their basic information are required. The charging station supports three authentication methods: NFC (Near Field Communication), QR code scanning, and RFID (Radio Frequency Identification). NFC utilizes the NFC chip built into the mobile device; the user simply brings their phone close to the NFC reading area on the charging station, and the electronic identity credential stored on the phone can be read by the charging station. QR code scanning generates a dynamic QR code on the charging station's display screen; the user scans the code using a mobile application and confirms their identity within the application. RFID uses a dedicated charging card; the user brings the card close to the reader area on the charging station, and the reader captures the RFID signal stored in the card and extracts the user's identity information. Users can choose one of these three verification methods based on their personal habits and the site conditions. During the verification process, the charging station combines the acquired user identification code with the charging station ID, verification timestamp, and verification method code to form the identity verification result.

[0054] The authentication result needs to be securely transmitted to the central server for further processing, therefore, the RSA asymmetric encryption algorithm is used for protection. RSA encryption uses a pair of keys: a public key and a private key. The central server pre-distributes the public key to each charging station, while the private key is securely stored on the server side. The authentication result is encrypted using the public key to generate ciphertext, which can only be decrypted by the central server possessing the corresponding private key. The encryption process first converts the authentication result into a binary data stream, then encrypts it in blocks, with each block having a maximum length of the key length minus 11 bytes (for padding requirements). The encrypted data blocks are then merged to form complete ciphertext, which is then Base64 encoded into a transmittable string format to obtain the encrypted identity data.

[0055] After receiving the encrypted identity data, the central server uses its private key to decrypt and verify the user's identity. Upon successful verification, the server retrieves the user's historical charging records from the database and calculates a credit rating. The credit rating is based on three core indicators: cumulative charging frequency, average charging amount, and payment timeliness. Cumulative charging frequency reflects the user's system usage frequency and loyalty, obtained directly from the total number of historical charging records in the database. Average charging amount is calculated by dividing the total charging amount of all historical charging records by the number of charging sessions, reflecting the user's electricity consumption. Payment timeliness is obtained by analyzing the user's historical payment records and calculating the proportion of timely payments (timely payments divided by total payment times), reflecting the user's creditworthiness. These three indicators are weighted and summed to arrive at the user's credit score.

[0056] Credit Score = Weight of Cumulative Charging Times × Normalized Value of Cumulative Charging Times + Weight of Average Charging Amount × Normalized Value of Average Charging Amount + Weight of Payment Timeliness × Payment Timeliness Ratio

[0057] The weights for each indicator are as follows: cumulative charging frequency (0.3), average charging volume (0.2), and payment timeliness (0.5). The cumulative charging frequency and average charging volume need to be normalized, mapping the original values ​​to the 0-1 range. The calculation method is to divide the current value by the system's preset maximum reference value. The final calculated credit score ranges from 0 to 100. Based on the user's credit score, the system divides users into four authorization levels: Diamond Member, Gold Member, Silver Member, and Regular Member. The classification criteria are: 90-100 points corresponds to Diamond Member, 80-89 points to Gold Member, 70-79 points to Silver Member, and 0-69 points to Regular Member. Different membership levels enjoy different service benefits, such as charging priority, billing discounts, and reservation privileges.

[0058] Based on user authorization levels, the system assigns personalized charging power limits and charging durations to users. The charging power limit determines the maximum charging rate a user can use: Diamond members can use the highest charging power (e.g., 350kW super-fast charging), Gold members the next highest (e.g., 150kW fast charging), Silver members the next lowest (e.g., 60kW medium-speed charging), and Regular members the lowest (e.g., 40kW standard charging). Charging duration is the maximum allowed time for a single charge: Diamond members have no time limit, Gold members have 8 hours, Silver members have 6 hours, and Regular members have 4 hours. When the grid load is high, these limits are dynamically adjusted by multiplying them by the current grid load rate. All this information is combined to form user authorization data, which guides subsequent charging process control.

[0059] For example, a user, Mr. Li, completes real-name registration through a mobile application, and the system generates a unique user identifier code for him: "u7d8f9g0-h1j2-k3l4-m5n6-p7q8r9s0t1". Upon arriving at the charging station, Mr. Li chooses to authenticate his identity using his mobile phone's NFC method. He brings his phone close to the NFC reading area of ​​charging pile CP20240415, and the charging pile successfully reads Mr. Li's electronic identity credential, generating an authentication result. This result includes the user identifier code, the charging pile ID "CP20240415", the verification time "2024-04-15 09:30:45", and the verification method code "NFC-01". The charging pile uses an RSA public key to encrypt this information, generating encrypted identity data and transmitting it to the central server. After decryption, the server retrieves Mr. Li's historical charging records: a total of 45 charging sessions, a total historical charging volume of 1125 kWh, an average charging volume of 25 kWh per session, and 48 out of 50 payments were made on time. The system divides the cumulative number of charging attempts (45) by the maximum reference value (100) to obtain a normalized value of 0.45, and divides the average charging amount (25kWh) by the reference value (50kWh) to obtain a normalized value of 0.5. The payment timeliness is calculated as 48 / 50 = 0.96. Substituting these values ​​into the formula, the credit score is calculated as follows: Credit Score = 0.3 × 0.45 + 0.2 × 0.5 + 0.5 × 0.96 = 0.135 + 0.1 + 0.48 = 0.715 × 100 = 71.5 points. Based on this score, Mr. Li is awarded Silver Membership. The system allocates a maximum charging power of 60kW and a maximum charging time of 6 hours to him, forming complete user authorization data to guide subsequent charging process control.

[0060] In one specific embodiment, the process of executing step S103 may specifically include the following steps:

[0061] Temperature parameters are extracted from the original data of the charging pile, and temperature time series is constructed by selecting the internal cable temperature data of the charging pile in the most recent 30 minutes to obtain historical cable temperature data.

[0062] The temperature threshold is set in a personalized manner based on the user's identity authorization data. The personalized temperature threshold is obtained by multiplying the basic temperature threshold by the security coefficient corresponding to the user's authorization level.

[0063] The historical cable temperature data is divided into time periods. The average temperature value is calculated every 5 minutes and the data points are recorded to obtain the temperature change data points.

[0064] The rate of temperature rise is calculated based on temperature change data points. The rate of temperature rise for each time period is obtained by dividing the temperature difference between adjacent time points by the time interval.

[0065] The temperature rise rate value is compared with the preset rate standard, and the safety level is divided by setting three ranges: normal rate, warning rate and dangerous rate, so as to obtain the temperature change level.

[0066] The safety risk value is calculated based on the temperature change level and the difference between the current measured temperature and the personalized temperature threshold. The cable safety status is comprehensively evaluated by weighted summation to obtain a charging safety warning signal.

[0067] Specifically, the internal cables of charging piles are prone to high temperatures during high-current charging. If not detected and addressed promptly, this can lead to safety accidents such as insulation aging, short circuits, or even fires. The first step is to extract temperature parameters from the charging pile's raw data. This raw data refers to various real-time monitoring data collected by IoT sensors and transmitted to a central server, including parameters such as voltage, current, temperature, humidity, and power. Temperature parameters are collected by temperature sensors distributed throughout key parts of the charging pile, recording data every 30 seconds. The temperature data is timestamped, allowing filtering of internal cable temperature data from the most recent 30 minutes. The data filtering process first calculates the current time minus 30 minutes as the starting time point, then extracts all temperature records from the database according to the charging pile ID and time range. These cable temperature data are arranged chronologically, forming a complete temperature time series, i.e., historical cable temperature data. Personalized temperature thresholds are then set based on user authorization data. This user authorization data includes user authorization level information, with different levels corresponding to different safety coefficients. The safety coefficient is an adjustment parameter set based on a safety assessment of the user's historical charging behavior, used to adjust the base temperature threshold. The base temperature threshold refers to the maximum permissible temperature of the charging cable under normal operating conditions, typically set at 60℃. Diamond members, due to their historically safer and more reliable charging behavior, have a safety factor of 1.05, allowing for operating temperatures slightly higher than the base temperature. Gold members have a safety factor of 1.0, strictly adhering to the base temperature threshold. Silver members have a safety factor of 0.95, requiring operating temperatures lower than the base temperature. Regular members have a safety factor of 0.9, requiring the most stringent temperature control. By multiplying the base temperature threshold by the corresponding safety factor, a personalized temperature threshold applicable to the current user is calculated. For example, a Diamond member's personalized temperature threshold is 60℃ × 1.05 = 63℃, while a regular member's personalized temperature threshold is 60℃ × 0.9 = 54℃.

[0068] Dividing the historical cable temperature data into time periods reduces data noise and helps extract temperature trends. Since the raw data is collected every 30 seconds, there will be 60 data points within 30 minutes. Direct analysis of these data points is susceptible to short-term fluctuations. Therefore, the 30-minute data is divided into 5-minute intervals, each containing 10 raw data points. The arithmetic mean of the 10 temperature data points within each interval is taken as the representative temperature for that interval, and the midpoint of that interval is recorded. This compresses the 30-minute data into 6 temperature change data points, each representing the average temperature within a 5-minute interval, with a 5-minute time interval between data points. Calculating the temperature rise rate based on these temperature change data points is a crucial method for detecting temperature anomalies. The temperature rise rate refers to the rate of temperature increase per unit time, reflecting how quickly the charging pile cable heats up. The calculation method is to take the temperature difference between two adjacent time points and divide it by the time interval between the two points (fixed at 5 minutes). For example, if the average temperature in the first time period is 40℃ and the average temperature in the second time period is 42.5℃, then the temperature rise rate between these two time periods is (42.5-40) / 5 = 0.5℃ / minute. For six temperature change data points, the temperature rise rate values ​​for five consecutive time periods can be calculated. A positive temperature rise rate indicates that the temperature is rising, a negative rate indicates that the temperature is falling, and zero indicates that the temperature remains stable.

[0069] The calculated temperature rise rate is compared with a preset rate standard to classify the safety level. The rate standard includes three intervals: normal rate interval (0-0.3℃ / minute), warning rate interval (0.3-0.8℃ / minute), and dangerous rate interval (greater than 0.8℃ / minute). The normal rate interval indicates that the cable temperature rises slowly or remains stable, which is a safe state; the warning rate interval indicates that the cable temperature rises relatively quickly, requiring attention but not yet reaching a dangerous level; the dangerous rate interval indicates that the cable temperature rises extremely quickly, posing a safety hazard. By comparing the calculated temperature rise rate for each time period with these three intervals, the temperature change level for each time period is determined. For example, if the temperature rise rate for a certain time period is 0.5℃ / minute, then the temperature change level for that time period is "warning level". For 5 consecutive time periods, 5 temperature change levels will be obtained.

[0070] A safety risk value is calculated based on the temperature change level and the difference between the current measured temperature and the personalized temperature threshold. The current measured temperature refers to the latest temperature sensor reading, while the personalized temperature threshold is the previously calculated upper temperature limit for the current user. The difference between the two reflects how close the current temperature is to the safe upper limit; a smaller difference indicates a closer proximity to a dangerous state. Simultaneously, the temperature change level reflects the trend of temperature changes; the appearance of a warning or danger level indicates that the temperature may continue to rise rapidly. Taking both factors into account, the safety risk value is calculated using a weighted summation method.

[0071] Safety risk value = 0.6 × temperature proximity + 0.4 × temperature change trend

[0072] Among them, temperature proximity = (current measured temperature / personalized temperature threshold), with a value range of 0-1. The closer to 1, the closer to the temperature threshold. Temperature change trend is a comprehensive assessment of the temperature change level in the most recent time period. The calculation method is to assign a value to the temperature change level in the most recent three time periods (0.1 for normal, 0.5 for warning, and 1.0 for danger), and then multiply it by the weights of 0.1, 0.3, and 0.6 respectively according to the time from farthest to most recent, and add them together to obtain the final temperature change trend, with a value range of 0.1-1.0.

[0073] The calculated safety risk value ranges from 0 to 1, with higher values ​​indicating higher safety risks. Warning levels are set based on these values: 0-0.5 indicates a safe state with no warning triggered; 0.5-0.7 is a low-level warning, notifying charging station management personnel; 0.7-0.9 is a medium-level warning, notifying the user and suggesting a reduction in charging power; and 0.9 and above is a high-level warning, automatically reducing charging power or interrupting the charging process and notifying emergency personnel. The final warning level, together with the corresponding automatic control strategy, constitutes a complete charging safety warning signal.

[0074] In one specific embodiment, the process of executing step S104 may specifically include the following steps:

[0075] The user's charging history is extracted from the original data of the charging pile. By filtering the charging time, charging duration, charging amount and charging frequency in the past 90 days, the user's original charging characteristics are obtained.

[0076] The original charging characteristics of users are processed in terms of time dimension. By statistically analyzing the charging frequency distribution of each user over seven days from Monday to Sunday and within 24 hours each day, user time preference characteristics are obtained.

[0077] The user's charging behavior is rated based on the charging safety warning signals. The user's charging safety index is obtained by calculating the ratio of the frequency of safety warnings triggered during the user's historical charging process to the total number of charging times.

[0078] Normalize the user time preference characteristics and user charging safety index by mapping the data of each dimension to the 0-1 interval to eliminate the difference in units and obtain a standardized feature vector.

[0079] The optimal number of clusters is obtained by adaptively determining the K value based on the standardized feature vector, by calculating the silhouette coefficient under different K values ​​and selecting the K value corresponding to the maximum silhouette coefficient.

[0080] The standardized feature vectors are grouped and calculated based on the optimal number of clusters. The cluster centers are initially selected from samples far from the center point and iteratively updated until the cluster centers are stable, thus obtaining the user behavior clustering results.

[0081] Specifically, in IoT-based charging pile management methods, user charging behavior analysis and clustering are key steps in improving the efficiency of charging resource scheduling. First, extracting user charging history records from raw charging pile data requires clarifying the data source and selection criteria. Raw charging pile data is stored in a distributed database, containing complete records of each user's charging session, such as user ID, charging pile ID, charging start time, charging end time, charging amount, and charging cost. To obtain effective user behavior characteristics, it is necessary to extract user charging data from the past 90 days. This 90-day time span covers seasonal changes in user charging habits without including too much historical data that could cause the analysis to deviate from current behavior patterns. The data extraction process uses SQL queries to filter data from the database by user ID and time range, obtaining the raw dataset. From this raw data, four key features are extracted: charging time (specific to the hour and day of the week), charging duration (duration from start to finish), charging amount (kilowatt-hours per charge), and charging frequency (number of charges within 90 days). These data collectively constitute the raw user charging characteristics, serving as the foundation for subsequent analysis.

[0082] Next, the original user charging characteristics are processed over time to uncover patterns in user charging preferences. This time-dimensional processing consists of two parts: weekly distribution and daily distribution. The weekly distribution tracks the number of times a user charges each day of the week, forming a 7-dimensional frequency vector that reflects whether the user prefers to charge on weekdays or weekends. Specifically, each charging record within 90 days is categorized and counted according to the week of the day (Monday to Sunday), then divided by the user's total charging count to obtain the charging frequency for each week. Similarly, the daily distribution tracks the number of times a user charges in each hour within a 24-hour period, forming a 24-dimensional frequency vector that reflects whether the user is more accustomed to charging during the day, at night, or in the early morning. This is calculated by categorizing and counting each charging record within 90 days according to the hour in which charging began (0:00-23:00), then dividing by the total charging count to obtain the charging frequency for each hour. These two distribution vectors are combined to form the user's time preference characteristics, a crucial dimension for characterizing user charging habits.

[0083] Assessing user charging safety by rating their charging behavior based on charging safety warning signals is a crucial step in evaluating the safety of user charging behavior. Charging safety warning signals are safety risk indicators generated by a temperature anomaly detection module during the charging process, and are categorized into three levels: low, medium, and high. The safety rating process first extracts all events that triggered safety warnings from the user's historical charging records from the database, recording the warning level and the number of occurrences. Then, the ratio of the frequency of safety warnings to the total number of charging attempts is calculated to obtain the user's charging safety index. The calculation formula is: Safety Index = 1 - (Weighted Warning Count / Total Charging Attempts), where the weighted warning count is obtained by multiplying low-level warnings by 0.3, medium-level warnings by 0.6, and high-level warnings by 1.0, and then summing the results. The safety index ranges from 0 to 1; a value closer to 1 indicates safer user charging behavior, while a value closer to 0 indicates a greater risk of safety hazards.

[0084] Normalization of user time preference features and charging safety index aims to eliminate dimensional differences between features, ensuring each feature has equal weight in cluster analysis. The normalization process employs a min-max normalization method, mapping the original value of each feature to the 0-1 range. The specific calculation formula is: Normalized value = (Original value - Minimum value) / (Maximum value - Minimum value). For the frequency vectors of the weekday and dayday distributions, each dimension is normalized individually; for the charging safety index, since it is already within the 0-1 range, no further normalization is needed. Finally, all normalized feature vectors are merged into a single 32-dimensional standardized feature vector (7-dimensional weekday distribution + 24-dimensional daytime distribution + 1-dimensional safety index). This vector comprehensively describes the user's charging behavior characteristics, providing standardized input data for subsequent cluster analysis.

[0085] Adaptive determination of the K-value based on standardized feature vectors is a core step in solving the most critical problem in cluster analysis—determining the optimal number of clusters. The K-means algorithm requires pre-specifying the number of clusters K, and different K values ​​lead to different clustering results. To objectively determine the most suitable K value, the silhouette coefficient method is used for evaluation. The silhouette coefficient is an indicator of cluster quality, comprehensively considering intra-cluster similarity and inter-cluster dissimilarity. The calculation process first selects a series of candidate K values, typically from 2 to 8 or 10. A complete K-means clustering is performed for each K value, and then the corresponding silhouette coefficient is calculated. The silhouette coefficient is calculated as follows: for each sample point, the average distance *a* to other points in the same cluster and the average distance *b* to all points in the nearest other cluster are calculated. Then, the silhouette value of that point is calculated using (ba) / max(a,b). Finally, the average of all silhouette values ​​is taken as the overall silhouette coefficient. The silhouette coefficient ranges from -1 to 1; a larger value indicates better clustering. By comparing the silhouette coefficients corresponding to different K values, the K value with the largest silhouette coefficient is selected as the optimal number of clusters.

[0086] The standardized feature vectors are grouped and calculated based on the optimal number of clusters to complete the final user behavior clustering. This step uses an improved K-means++ algorithm, which optimizes the selection of initial cluster centers compared to the traditional K-means algorithm. Traditional K-means randomly selects initial centers, which can easily lead to unstable clustering results; while the K-means++ algorithm selects initial centers using a probability distribution method, ensuring that these points are as far apart as possible, improving clustering stability. The specific steps are as follows: First, a sample point is randomly selected as the first cluster center. Then, the shortest distance from each remaining sample point to the selected center is calculated, and probability sampling is performed using the square of the distance as a weight to select the next center point. This process is repeated until K centers are selected. After the initial centers are determined, the standard K-means iterative process is executed: each sample is assigned to the nearest center point to form K clusters, and then the center point of each cluster (the average of all sample points) is recalculated. These two steps are repeated until the center point positions no longer change significantly or the maximum number of iterations is reached (usually set to 50). Finally, each user is assigned to a specific cluster, forming a complete user behavior clustering result.

[0087] For example, an electric vehicle charging station operator used this method to analyze the charging behavior of 5,000 users. First, charging records from the past 90 days were extracted from the database. User A had 25 charging records during these 90 days, with an average charging volume of 35 kWh per charge and an average charging time of 2.5 hours. Statistical analysis revealed that User A primarily charged on Mondays, Wednesdays, and Fridays, accounting for 20%, 25%, and 30% of the total charging frequency, respectively, with very few charging on weekends. In terms of intraday distribution, the user mainly charged between 6:00 PM and 8:00 PM, accounting for 60% of the total charging frequency. Regarding safety rating, User A triggered a low-level warning twice, a medium-level warning once, and no high-level warnings in the 25 charging sessions, resulting in a safety index of 0.88 (1 - (2 × 0.3 + 1 × 0.6 + 0 × 1.0) / 25). After normalizing these data, a 32-dimensional standardized feature vector for User A was formed. The operator performed the same processing on all 5,000 users and then tested the clustering effect from K=2 to K=8 using the silhouette coefficient method. The highest silhouette coefficient (0.68) was found at K=4, thus determining the optimal number of clusters to be 4. Using the K-means++ algorithm for clustering, user A was ultimately classified as a "weekday evening peak charging group." This group of users is characterized by charging at fixed times after work on weekdays, exhibiting strong regularity in their charging behavior and a high level of safety awareness. Based on this clustering result, the charging station operator developed differentiated service strategies for different user types, such as reserving specific charging piles for "weekday evening peak charging groups" and offering off-peak charging discounts to encourage some users to adjust their charging times and alleviate charging pressure during the evening peak hours.

[0088] In one specific embodiment, the process of executing step S105 may specifically include the following steps:

[0089] A time-series dataset was constructed based on user behavior clustering results and original charging pile data. By organizing nearly 30 days of historical load data into hourly data points and adding time labels, charging load time-series data was obtained.

[0090] The charging load time series data is augmented with features by adding time features, weather features, and user behavior clustering ratio features as input variables to obtain a multidimensional input feature matrix.

[0091] The multidimensional input feature matrix is ​​segmented by a sliding window. The data is organized by setting 24 hours as the input window length and the next 24 hours as the prediction target length, resulting in input-output sample pairs.

[0092] An LSTM network structure is constructed for the input-output sample pairs. By setting up a network structure with an input layer, two LSTM hidden layers, a Dropout layer, and a fully connected output layer, the time series data is processed to obtain the load prediction model.

[0093] The charging load for the next 24 hours is calculated based on the load prediction model. By inputting the load data of the current 24 hours into the model and gradually rolling the prediction of the load value for subsequent periods, the load prediction curve is obtained.

[0094] Based on the load forecast curve, the charging piles are scheduled according to resources. By dividing the forecast period into peak period, off-peak period and low period and calculating the scheduling priority weight of each charging pile, a charging scheduling scheme is obtained.

[0095] Specifically, a time-series dataset was constructed based on user behavior clustering results and raw charging pile data. This process involved data extraction, organization, and labeling. User behavior clustering results provided classifications of charging characteristics for different user groups, such as "peak-hour charging users," "nighttime slow-charging users," "weekend charging users," and "emergency fast-charging users." The raw charging pile data contained real-time operating parameters for each charging pile, such as voltage, current, and power. Through database queries, load data for all charging piles over the past 30 days was extracted, and the total regional load value was calculated hourly, forming a 720-hour historical load time series. The load value was calculated by adding the power data of all active charging piles within the same time period, representing the total power demand of the entire charging network at a given moment. To facilitate subsequent processing, each load data point was labeled with a corresponding time tag, including year, month, day, hour, day of the week, and whether it was a holiday, thus forming the basic charging load time-series data.

[0096] Feature augmentation processing is performed on the charging load time-series data. The original load data only contains historical load values ​​and cannot reflect the various factors influencing load changes. Therefore, three important features need to be introduced: time features, weather features, and user behavior clustering ratio features. Time features include hourly indicators (one-hot encoding of 0-23), date indicators (1-31), month indicators (1-12), day of the week indicators (0-6), and whether it is a holiday indicator (0 or 1). These features reflect the periodic variation pattern of the load. Weather features are obtained from meteorological data APIs, including temperature (degrees Celsius), humidity (percentage), weather conditions (classified encoding such as sunny, cloudy, rainy, snowy, etc.), and wind speed (meters per second). These factors directly affect the travel and charging behavior of electric vehicle users. User behavior clustering ratio features reflect the proportional distribution of various user groups in the current area, such as "morning peak charging users" accounting for 30%, "nighttime slow charging users" accounting for 25%, "weekend charging users" accounting for 20%, and "emergency fast charging users" accounting for 25%. This proportional distribution is calculated based on user clustering results and effectively reflects the impact of user group structure on charging load. All these features, aligned with the original load data by time, together form a multidimensional input feature matrix, where each row represents a time point and each column represents a feature dimension. Sliding window segmentation of the multidimensional input feature matrix is ​​used to transform continuous time-series data into a format suitable for training machine learning models. The sliding window method extracts multiple input-output sample pairs from continuous time series by setting a fixed-length input window and a prediction target window. Specifically, a 24-hour window is set as the input window length, using 24 consecutive hours of multidimensional feature data as model input; the load value for the next 24 hours is used as the prediction target, meaning the model needs to predict the hourly charging load within the next day. The segmentation process starts from the beginning of the historical data, using data from hours 1 to 24 as the input of the first sample and the load from hours 25 to 48 as its corresponding output; then the window slides forward one time step, using data from hours 2 to 25 as the input of the second sample and the load from hours 26 to 49 as its output; and so on, until the entire historical dataset is covered. This sliding window segmentation method can maximize the use of limited historical data to generate a large number of training sample pairs, providing sufficient training data for deep learning models.

[0097] Constructing an LSTM network structure for the input-output sample pairs is a core step in time series prediction using deep learning techniques. LSTM (Long Short-Term Memory) networks are a special type of recurrent neural network structure specifically designed for processing sequential data. They effectively capture long-term dependencies in time series data, making them ideal for data with significant temporal correlations, such as charging loads. The network structure consists of multiple layers: first, an input layer that receives an input matrix of 24 hours × feature dimensions; then, two LSTM hidden layers, the first containing 128 neurons and the second containing 64 neurons, enabling the network to learn complex temporal patterns and multi-level features in the data; next, a Dropout layer that randomly discards 50% of the neuron connections to prevent overfitting; and finally, a fully connected output layer containing 24 neurons, corresponding to the predicted load value for the next 24 hours. The network training employs the Adam optimizer with mean squared error (MSE) as the loss function. The training process includes data normalization (scaling all features to the 0-1 range), batch training (64 samples per batch), and an early stopping mechanism (training stops when the validation set loss no longer decreases for 5 consecutive cycles). After training, the model performance is evaluated on the validation set to obtain the final load prediction model. Calculating the charging load for the next 24 hours based on the load prediction model is the step of applying the trained model to actual prediction. Specifically, historical data from the 24 hours prior to the current time is taken, including load values ​​and all extended features, to form the model input matrix. Simultaneously, the time features (known) and weather features (obtained from the weather forecast API) for the next 24 hours, as well as the current user clustering ratio features (assuming they remain unchanged in the short term), are prepared. This data is input into the trained LSTM model, which outputs the predicted load values ​​for the next 24 hours all at once. To improve prediction accuracy, a rolling prediction strategy can be used: first predict the load for the next hour, then add this predicted value to the historical data, update the input window, and then predict the next hour, repeating this cycle 24 times. Although this method involves a large amount of computation, it effectively reduces the accumulation of prediction errors. The final 24 predicted values ​​form a complete load forecast curve, reflecting the changing trend of charging demand over the next day.

[0098] Resource scheduling of charging piles based on load forecast curves is the final step in optimizing charging resource allocation and improving system efficiency. First, different time periods need to be categorized based on the predicted load curve. Specifically, two load rate thresholds are set: a peak threshold (typically 80% of rated capacity) and a valley threshold (typically 50% of rated capacity). Periods with predicted load rates higher than the peak threshold are classified as peak periods, and periods with load rates lower than the valley threshold are classified as valley periods; the remaining periods are off-peak periods. For each charging pile, its scheduling priority weight is calculated as the basis for resource allocation. The formula for calculating the scheduling priority weight is: W = α × (1 - current load rate) + β × (1 - failure rate) + γ × (1 - predicted congestion) + δ × battery health status, where α, β, γ, and δ are weight parameters, and α + β + γ + δ = 1. The current load rate refers to the ratio of the charging pile's current power to its rated power; the failure rate is the probability of failure calculated based on historical failure records; the predicted congestion level is the utilization forecast calculated based on the predicted load and charging pile capacity; and the battery health status is the health index of the charging pile's battery pack, reflecting the charging pile's reliability. Based on the calculated weight values, all charging piles are ranked, with higher-weighted piles receiving new charging tasks first. Simultaneously, for reservable charging demands, they are prioritized for scheduling during off-peak hours to achieve peak shaving and valley filling; for peak-hour charging demands, users are guided to charging stations with lower loads to avoid regional congestion.

[0099] In one specific embodiment, the process of executing step S106 may specifically include the following steps:

[0100] The charging time period is divided according to the charging scheduling scheme. The 24 hours are divided into three periods: peak period, off-peak period and low-peak period, and time period coefficients of 1.5, 1.0 and 0.7 are assigned respectively to obtain the time period electricity price coefficient.

[0101] The user identity authorization data is used to calculate the level discount. By setting discount rates of 0.9, 0.93, 0.95 and 1.0 for Diamond members, Gold members, Silver members and Ordinary members respectively, the user discount coefficient is obtained.

[0102] The real-time charging rate is calculated based on the time-period electricity price coefficient and the current market benchmark electricity price. The real-time rate for the user is obtained by multiplying the benchmark electricity price by the corresponding time-period coefficient and then by the user discount coefficient.

[0103] The cost of a single charge is accumulated based on the user's real-time rate and the amount of electricity charged each time. The accumulated cost is calculated and recorded for every 1 kWh of electricity charged, thus obtaining the real-time charging cost.

[0104] Electronic bills are generated for real-time charging costs. By recording the start and end times of charging, the electricity consumption in each time period, the time period rate, the total electricity consumption, the total cost, and the discount amount, a detailed charging bill is obtained.

[0105] Detailed charging bills are stored using blockchain technology. By calculating the hash value of the bill data and adding a timestamp, an immutable transaction record is formed, resulting in the user's charging bill.

[0106] Specifically, the charging time period is divided according to the charging scheduling scheme. The charging scheduling scheme is based on the 24-hour charging load prediction curve generated by the LSTM load prediction model. This curve reflects the grid load status at different time periods. By analyzing the load curve, the 24 hours are divided into three periods with different characteristics: peak period, off-peak period, and valley period. The peak period refers to the time period when the charging load exceeds the preset threshold (usually 80% of the grid capacity), mainly concentrated between 7:00-9:00 in the morning and 17:00-19:00 in the evening. At this time, the grid pressure is high and the power supply cost is high. The off-peak period refers to the time period when the charging load is at a moderate level (usually 50%-80% of the grid capacity), such as 9:00-17:00 and 19:00-22:00. At this time, the grid operates stably. The valley period refers to the time period when the charging load is below the preset threshold (usually below 50% of the grid capacity), mainly concentrated between 22:00 at night and 6:00 the next day. At this time, the grid has a large amount of spare capacity and the power supply cost is low. To encourage users to charge during off-peak hours, different time-based pricing coefficients are assigned to these three periods: 1.5 for peak hours (1.5 times the base price), 1.0 for off-peak hours (equal to the base price), and 0.7 for off-peak hours (0.7 times the base price). These time-based pricing coefficients directly affect the final charging rate and are a crucial means of peak shaving and valley filling. Tiered discounts are calculated based on user authorization data, which includes user ID, authorization level, and credit score, generated during the user identification and authorization management phase. Based on users' credit scores and historical charging behavior, users are divided into four levels: Diamond Member, Gold Member, Silver Member, and Regular Member. To encourage users to upgrade their membership level and maintain good charging habits, different discount rates are set for different membership levels: Diamond members enjoy the highest discount, with a discount rate of 0.9, receiving a 10% discount; Gold members are next, with a discount rate of 0.93, receiving a 9.3% discount; Silver members are next, with a discount rate of 0.95, receiving a 9.5% discount; Regular members do not enjoy any discount, with a discount rate of 1.0, and are charged at the original price. These discount rates are stored in the user database. When a user charges, the system automatically reads the corresponding discount rate as an important parameter for calculating the real-time rate.

[0107] Real-time charging rates are calculated based on time-of-use electricity price coefficients and the current market benchmark electricity price. The market benchmark electricity price refers to the current standard electricity price in the electricity market, usually expressed in yuan / kWh, and is determined by the grid operator based on factors such as generation costs, grid operation and maintenance costs, and government subsidies, and is updated regularly. The formula for calculating the real-time charging rate is: Real-time rate = Benchmark electricity price × Time-of-use coefficient × User discount coefficient. For example, when the benchmark electricity price is 1.0 yuan / kWh, if a user charges during peak hours and is a silver member, the real-time rate is 1.0 × 1.5 × 0.95 = 1.425 yuan / kWh. This calculation method comprehensively considers time factors and user level factors, which can both guide users to charge during off-peak hours and incentivize users to upgrade their membership levels, thereby achieving the dual goals of grid load balancing and improved user loyalty.

[0108] The cost of a single charge is accumulated based on the user's real-time tariff and the amount of electricity charged each time. During the charging process, the charging pile's built-in electricity metering module monitors the charging amount in real time, triggering a cost calculation and recording every 1 kWh of electricity accumulated. Specifically, the latest measured 1 kWh of electricity is multiplied by the real-time tariff for the current period to obtain the cost corresponding to that 1 kWh, which is then added to the total cost. For example, if a user charges 3 kWh during off-peak hours (when the real-time tariff is 0.7 yuan / kWh), this cost is 3 × 0.7 = 2.1 yuan. If, during off-peak hours, another 2 kWh is charged (when the real-time tariff is 1.0 yuan / kWh), this cost is 2 × 1.0 = 2.0 yuan. The final total cost is 2.1 + 2.0 = 4.1 yuan. This time-based and electricity-based cumulative billing method ensures the accuracy and fairness of the billing, while also making it easier for users to understand the composition of charging costs.

[0109] The system generates electronic bills for real-time charging costs. Once charging is complete, the system automatically generates an electronic bill containing all necessary information. The bill includes: charging start and end times (accurate to the minute), such as "2024-04-15 18:30" to "2024-04-16 07:15"; electricity consumption for each period, such as "peak period: 2.5 kWh, off-peak period: 8 kWh, off-peak period: 15 kWh"; electricity price rate for each period, such as "peak period: 1.5 yuan / kWh, off-peak period: 1.0 yuan / kWh, off-peak period: 0.7 yuan / kWh"; user membership level and corresponding discount rate, such as "silver member: 0.95"; subtotal of costs for each period, such as "peak period: 3.56 yuan, off-peak period: 7.6 yuan, off-peak period: 9.98 yuan"; total electricity consumption "25.5 kWh"; total cost due "21.14 yuan"; discount amount "1.11 yuan"; actual payment amount "20.03 yuan". This detailed electronic bill is pushed to the user via their mobile application and is also stored in the system database for easy retrieval and verification. Storing the detailed charging bill using blockchain technology is a crucial step in ensuring the security and immutability of transaction data. Blockchain, a distributed database technology, is decentralized, immutable, fully traceable, and auditable, making it ideal for secure transaction data storage. The bill data storage process first calculates a hash value for the bill content using the SHA-256 algorithm, which maps data of any length to a fixed-length (256-bit) string, ensuring that different data have virtually no identical hash values, thus guaranteeing data uniqueness and integrity. Then, the bill data, hash value, and current timestamp are packaged together to form a transaction record. The timestamp is accurate to the millisecond, ensuring the transaction's temporal order. Next, this transaction record is broadcast to all nodes in the charging station network, where nodes verify the transaction's validity through a consensus mechanism. Once verified, the transaction record is added to a block. When a block reaches a predetermined size or time interval, the node links the current block with the hash value of the previous block to form a chain structure, and broadcasts the new block to the entire network, ensuring data immutability. Finally, users can query bill details in the blockchain explorer using the transaction ID or bill ID, ensuring transparent and traceable transactions.

[0110] For example, Mr. Zhang used charging station CP20240417 to charge his electric vehicle. Mr. Zhang completed identity verification with the charging station using his mobile phone's NFC function, and the system recognized him as a Gold Member (discount rate of 0.93). The base electricity price for the day was 1.0 yuan / kWh, and charging started at 20:00 in the evening (off-peak period, coefficient 1.0) and continued until 2:00 a.m. the next day (off-peak period, coefficient 0.7). During off-peak hours (8:00 PM - 10:00 PM), Mr. Zhang charged 8 kWh. The system calculated the real-time rate at 1.0 × 1.0 × 0.93 = 0.93 yuan / kWh, for a total cost of 8 × 0.93 = 7.44 yuan. During low-peak hours (10:00 PM - 2:00 AM), he charged 20 kWh. The system calculated the real-time rate at 1.0 × 0.7 × 0.93 = 0.651 yuan / kWh, for a total cost of 20 × 0.651 = 13.02 yuan. After charging, the system generated an electronic bill detailing the start and end times, electricity consumption during each period, the rate for each period, the total electricity consumption of 28 kWh, the total cost of 20.46 yuan, and a discount of 1.54 yuan (compared to a regular member). The system then calculates the SHA-256 hash value of the bill content, generating "3f7d8e2a1c9b6f5d4e2a1c9b8f7d6e5a4c3b2a1", adds the timestamp "2024-04-18 02:00:15.382", packages this information into a transaction record, and writes it into the blockchain. Mr. Zhang can view the bill details through a mobile application and verify the authenticity and immutability of the bill through a blockchain explorer.

[0111] The above describes the IoT-based charging pile management method in the embodiments of this application. The following describes the IoT-based charging pile management system in the embodiments of this application. Please refer to [link / reference]. Figure 2 One embodiment of the IoT-based charging pile management system in this application includes:

[0112] The data acquisition module 201 is used to acquire real-time data of the charging pile. It collects voltage parameters, current parameters, temperature parameters and power parameters through IoT sensors to obtain the raw data of the charging pile.

[0113] The authorization module 202 is used to identify and authorize users based on the original data of the charging pile, verify user information and assign authorization levels through a multi-factor authentication mechanism, and obtain user identity authorization data.

[0114] The detection module 203 is used to detect abnormal temperatures in the internal cables of the charging pile based on the user identity authorization data and the original data of the charging pile. It compares the measured temperature with the set temperature threshold and calculates the temperature rise rate to obtain a charging safety warning signal.

[0115] Analysis module 204 is used to perform feature extraction and cluster analysis on user charging behavior based on the charging safety warning signal and the original data of the charging pile. It calculates user behavior feature vectors by improving the K-Means clustering algorithm to obtain user behavior clustering results.

[0116] The scheduling module 205 is used to predict and schedule the charging load based on the user behavior clustering results and the original data of the charging pile, calculate the load prediction curve through the LSTM neural network model and optimize the charging scheduling strategy to obtain the charging scheduling scheme.

[0117] The calculation module 206 is used to perform multi-period dynamic calculation of the electricity price based on the charging scheduling scheme and the user identity authorization data, calculate the real-time charging rate by multiplying the benchmark electricity price by the time period coefficient, and perform fee accumulation processing to obtain the user's charging bill.

[0118] Through the collaborative efforts of the aforementioned components, real-time data collection from charging piles via IoT sensors enables precise monitoring of key parameters such as voltage, current, temperature, and power, laying a data foundation for charging safety and efficient management. A multi-factor authentication mechanism combined with authorization level allocation not only enhances system security but also implements differentiated service strategies, improving user experience while optimizing resource allocation efficiency. Personalized temperature threshold settings and real-time temperature rise rate calculations based on user authorization data construct a precise charging safety early warning mechanism, significantly reducing safety risks during charging and protecting charging equipment and user property. An improved K-Means clustering algorithm extracts and classifies user charging behavior features, revealing charging patterns and preferences of different user groups. These refined user profiles provide a basis for subsequent optimized allocation of charging resources. The LSTM neural network model fully utilizes temporal characteristics and multi-dimensional features to achieve high-precision charging load prediction. Its recursive structure effectively captures long-term dependencies, improving prediction accuracy by more than 15% compared to traditional time series prediction methods, providing a reliable prediction foundation for intelligent scheduling strategies. The multi-period dynamic electricity pricing mechanism combines user levels with time-period coefficients to form a flexible price incentive system, effectively guiding users to charge during off-peak hours, reducing peak load on the power grid, and improving the overall operational efficiency of the charging network. Overall, this solution deeply integrates IoT technology with artificial intelligence algorithms, particularly the application of K-Means clustering in user behavior analysis and the advantages of LSTM neural networks in load forecasting, to jointly construct an adaptive, efficient, and safe charging pile management system. This integration not only enhances the intelligence level of electric vehicle charging infrastructure but also optimizes the utilization of power grid resources through peak shaving and valley filling.

[0119] above Figure 2The IoT-based charging pile management system in this embodiment of the invention will be described in detail from the perspective of modular functional entities. The IoT-based charging pile management device in this embodiment of the invention will be described in detail from the perspective of hardware processing.

[0120] Figure 3 This is a schematic diagram of the structure of an IoT-based charging pile management device 300 provided in an embodiment of the present invention. The IoT-based charging pile management device 300 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 310 (e.g., one or more processors) and a memory 320, and one or more storage media 330 (e.g., one or more mass storage devices) for storing application programs 333 or data 332. The memory 320 and storage media 330 can be temporary or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the IoT-based charging pile management device 300. Furthermore, the processor 310 may be configured to communicate with the storage media 330 and execute the series of instruction operations in the storage media 330 on the IoT-based charging pile management device 300 to implement the steps of the aforementioned IoT-based charging pile management method.

[0121] The IoT-based charging pile management device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 331, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 3 The illustrated structure of the IoT-based charging pile management device does not constitute a limitation on the IoT-based charging pile management device provided by the present invention. It may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0122] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the Internet of Things-based charging pile management method.

[0123] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0124] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an IoT-based charging pile management device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0125] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for managing charging piles based on an Internet of Things, characterized in that, The method comprises: Real-time data acquisition of charging piles, collecting voltage parameters, current parameters, temperature parameters and power parameters through Internet of Things sensors to obtain charging pile raw data; Identity recognition and authorization management of users according to the charging pile raw data, verifying user information and assigning authorization levels through a multi-identity verification mechanism to obtain user identity authorization data; Temperature anomaly detection of internal cables of charging piles based on the user identity authorization data and the charging pile raw data, comparing the set temperature threshold with the measured temperature and calculating the temperature rise rate to obtain a charging safety warning signal; Feature extraction and cluster analysis of user charging behavior according to the charging safety warning signal and the charging pile raw data, calculating user behavior feature vectors through an improved K-Means clustering algorithm to obtain user behavior clustering results; Prediction and scheduling of charging load based on the user behavior clustering results and the charging pile raw data, calculating load prediction curves and optimizing charging scheduling strategies through an LSTM neural network model to obtain a charging scheduling scheme, including: constructing a time series data set according to the user behavior clustering results and the charging pile raw data, organizing and adding time labels to the historical load data of the past 30 days by one data point per hour to obtain charging load time series data; performing feature expansion processing on the charging load time series data by adding time features, weather features and user behavior clustering proportion features as input variables to obtain a multi-dimensional input feature matrix; performing sliding window segmentation on the multi-dimensional input feature matrix by setting 24 hours as the input window length and the next 24 hours as the prediction target length to organize data, obtaining input-output sample pairs; constructing an LSTM network structure for the input-output sample pairs by setting the network structure of the input layer, two LSTM hidden layers, the Dropout layer and the fully connected output layer to process time series data, obtaining a load prediction model; calculating the charging load of the next 24 hours based on the load prediction model by inputting the load data of the current 24 hours into the model and gradually rolling forward to predict the load values of the subsequent period, obtaining a load prediction curve; scheduling resources of charging piles according to the load prediction curve by dividing the prediction period into peak period, flat peak period and valley period and calculating the scheduling priority weight of each charging pile, obtaining the charging scheduling scheme; Multi-period dynamic calculation of electricity prices according to the charging scheduling scheme and the user identity authorization data, calculating real-time charging rates by multiplying the benchmark electricity price by the period coefficient and performing fee accumulation processing to obtain user charging bills. 2.The Internet of Things based charging pile management method according to claim 1, characterized in that, The real-time data acquisition of charging piles, collecting voltage parameters, current parameters, temperature parameters and power parameters through Internet of Things sensors to obtain charging pile raw data, comprises: Collecting charging pile input voltage and output voltage through voltage sensors, sampling and storing the collected voltage data every 5 seconds to obtain voltage parameters; The input current and output current of the charging pile are collected by the current sensor, the collected current data is sampled and stored every 5 seconds, and the current parameters are obtained; The internal temperature and charging interface temperature of the charging pile are collected by the temperature sensor, the collected temperature data is sampled and stored every 30 seconds, and the temperature parameters are obtained; The real-time charging power of the charging pile is collected by the power sensor, the collected power data is sampled and stored every 5 seconds, and the power parameters are obtained; The voltage parameters, current parameters, temperature parameters and power parameters are subjected to AES-256 encryption processing, and a charging pile unique identification code and a time stamp are added to obtain an encrypted data packet; The encrypted data packet is transmitted to the central server through the Internet of Things communication module, the encrypted data packet is decrypted, verified and indexed stored to obtain the charging pile original data. 3.The Internet of Things based charging pile management method according to claim 1, characterized in that, According to the charging pile original data, the user is identified and authorized, the user information is verified through a multi-identity verification mechanism, and the authorization level is assigned, to obtain user identity authorization data, including: According to the charging pile original data, the user information is obtained, the user real-name registration information is received through the mobile application program, and the unique user identification code is generated, to obtain the user basic information; The user basic information is subjected to multi-identity verification, and the user identity is verified through at least one of NFC near field communication technology, two-dimensional code scanning or RFID radio frequency identification, to obtain an identity verification result; The identity verification result is subjected to RSA asymmetric encryption processing, the user identity information is protected and transmitted through an encryption algorithm, and encrypted identity data is obtained; Based on the encrypted identity data and historical charging records, the user is credit rated, and the charging frequency, average charging capacity and payment timeliness are weighted and summed to obtain a user credit score; According to the user credit score, the user is authorized and classified, and the credit score is mapped to diamond members, gold members, silver members and ordinary members in four level intervals to obtain user authorization levels; Based on the user authorization level, the user is allocated with a charging power upper limit and a charging time length, and individualized charging parameters are calculated through a differentiated resource allocation algorithm for users of different levels to obtain the user identity authorization data. 4.The Internet of Things based charging pile management method according to claim 1, characterized in that, Based on the user identity authorization data and the charging pile original data, the temperature of the internal cable of the charging pile is detected, the set temperature threshold is compared with the measured temperature, and the temperature rise rate is calculated to obtain a charging safety warning signal, including: The temperature parameters are extracted from the charging pile original data, the internal cable temperature data of the charging pile in the last 30 minutes is selected to construct a temperature time sequence, and cable temperature historical data is obtained; According to the user identity authorization data, the temperature threshold is individually set, the basic temperature threshold is multiplied by the safety factor corresponding to the user authorization level to obtain an individualized temperature threshold; The cable temperature historical data is subjected to time period division processing, the average temperature value is calculated every 5 minutes, and the data points are recorded to obtain temperature change data points; According to the temperature change data points, a temperature rise rate is calculated by dividing the temperature difference between adjacent time points by the time interval to obtain a temperature rise rate value for each time period; The temperature rise rate value is compared with a preset rate standard, a safety level is divided by setting three intervals of normal rate, pre-warning rate and dangerous rate, and a temperature change level is obtained; Based on the temperature change level and the difference between the current measured temperature and the personalized temperature threshold, a safety risk value is calculated, the cable safety state is comprehensively evaluated by weighted summation, and the charging safety warning signal is obtained. 5.The Internet of Things based charging pile management method according to claim 1, characterized in that, According to the charging safety warning signal and the charging pile original data, user charging behavior is extracted and cluster analysis is performed, user behavior feature vectors are calculated by improving the K-Means clustering algorithm, and user behavior clustering results are obtained, including: User charging history records are extracted from the charging pile original data, charging time, charging duration, charging power and charging frequency in the last 90 days are filtered, and user charging original features are obtained; The user charging original features are processed in time dimension, the charging distribution frequency of each user in seven days from Monday to Sunday and 24 hours per day is counted, and user time preference features are obtained; According to the charging safety warning signal, the safety rating of user charging behavior is calculated, the ratio of the frequency of triggering safety warning to the total charging times in the user's historical charging process is calculated, and the user charging safety index is obtained; The user time preference features and the user charging safety index are normalized, the dimensional differences are eliminated by mapping the data in each dimension to the 0-1 interval, and the standardized feature vector is obtained; Based on the standardized feature vector, K value self-adaptation is determined, the silhouette coefficient under different K values is calculated, and the K value corresponding to the maximum silhouette coefficient is selected, and the optimal cluster number is obtained; According to the optimal cluster number, the standardized feature vector is grouped and calculated, the samples far from the center point are initially selected as the clustering center, and the clustering center is iteratively updated until it is stable, and the user behavior clustering result is obtained. 6.The Internet of Things based charging pile management method according to claim 1, characterized in that, According to the charging scheduling scheme and the user identity authorization data, the electricity price is dynamically calculated in multiple time periods, the real-time charging rate is calculated by multiplying the benchmark electricity price by the time period coefficient, and the fee accumulation processing is performed, and the user charging bill is obtained, including: According to the charging scheduling scheme, the charging period is divided, 24 hours is divided into peak period, flat peak period and valley period, and the time period coefficient is 1.5, 1.0 and 0.7 respectively, and the time period electricity price coefficient is obtained; The user identity authorization data is calculated by grade discount, the discount rates of diamond members, gold members, silver members and ordinary members are set to 0.9, 0.93, 0.95 and 1.0 respectively, and the user discount coefficient is obtained; Based on the time period electricity price coefficient and the current market benchmark electricity price, the real-time charging rate is calculated, the product of the benchmark electricity price and the corresponding time period coefficient is multiplied by the user discount coefficient, and the user real-time rate is obtained; According to the user real-time rate and the single charging electricity quantity, the single charging fee is accumulated, the accumulated fee is calculated once per 1 degree of charging electricity quantity and data is recorded, and real-time charging fee is obtained; The real-time charging fee is generated into an electronic bill, the charging start and end time, the electricity quantity of each period, the period rate, the total electricity quantity, the total fee and the preferential amount are recorded, and a charging detailed bill is obtained; The charging detailed bill is stored through transaction data storage based on the blockchain technology, the bill data hash value is calculated and a time stamp is added to form an unalterable transaction record, and the user charging bill is obtained.

7. An Internet of Things based charging pile management system, characterized in that, The application discloses an Internet of Things-based charging pile management system and a management method thereof. A collection module is configured to collect real-time data of the charging pile, collect voltage parameters, current parameters, temperature parameters and power parameters through Internet of Things sensors, and obtain charging pile original data. An authorization module is configured to identify and authorize users according to the charging pile original data, verify user information through a multi-identity authentication mechanism, and assign an authorization level, so as to obtain user identity authorization data. A detection module is configured to detect temperature abnormalities of internal cables of the charging pile based on the user identity authorization data and the charging pile original data, compare a set temperature threshold with a measured temperature, and calculate a temperature rise rate, so as to obtain a charging safety early warning signal. An analysis module is configured to extract and cluster analyze user charging behaviors according to the charging safety early warning signal and the charging pile original data, calculate user behavior feature vectors through an improved K-Means clustering algorithm, and obtain user behavior clustering results. The scheduling module is configured to predict and schedule the charging load based on the user behavior clustering result and the charging pile original data, calculate a load prediction curve by using an LSTM neural network model, and optimize a charging scheduling strategy to obtain a charging scheduling scheme, including: constructing a time series dataset according to the user behavior clustering result and the charging pile original data, organizing historical load data of nearly 30 days according to one data point per hour, and adding a time label to obtain charging load time series data; performing feature expansion processing on the charging load time series data, adding time features, weather features, and the user behavior clustering proportion features as input variables to obtain a multi-dimensional input feature matrix; performing sliding window segmentation on the multi-dimensional input feature matrix, setting 24 hours as an input window length and 24 hours in the future as a prediction target length to organize data, and obtaining an input-output sample pair; constructing an LSTM network structure on the input-output sample pair, processing time series data by setting an input layer, two layers of LSTM hidden layers, a Dropout layer, and a fully connected output layer, and obtaining a load prediction model; calculating the charging load in the next 24 hours based on the load prediction model, inputting the load data in the current 24 hours into the model, and gradually rolling forward to predict the load value in the subsequent period to obtain a load prediction curve; and scheduling resources of the charging pile according to the load prediction curve, dividing the prediction period into a peak period, a flat peak period, and a valley period, calculating the scheduling priority weight of each charging pile, and obtaining the charging scheduling scheme. The calculation module is configured to calculate the electricity price in multiple time periods according to the charging scheduling scheme and the user identity authorization data, calculate the real-time charging rate by multiplying the benchmark electricity price by the time period coefficient, and perform fee accumulation processing to obtain a user charging bill.

8. An Internet of Things based charging pile management device, characterized in that, The computer program is stored in the memory and can be run on the processor, and the processor implements the method for managing the charging pile based on the Internet of Things according to any one of claims 1 to 6 when the computer program is run.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is stored in the memory and can be run on the processor, and the processor implements the method for managing the charging pile based on the Internet of Things according to any one of claims 1 to 6 when the computer program is run.

Images