Multi-scene centralized energy management system and automated intelligent replenishment device

Abstract

The present application relates to the technical field of intelligent energy management and automation control, and particularly relates to a multi-scene centralized energy management system and an automated intelligent replenishment device, which comprises a data acquisition module that acquires energy data such as fuel inventory liquid level, fuel transaction volume and environmental temperature of a gas station in real time through a sensor array; a communication transmission module that encapsulates and compresses the data and transmits the data to a central control platform through a wireless network; a central processing module that stores the data after cleaning the data and generates an inventory warning sign by using a machine learning model for time series prediction and anomaly detection; a decision generation module that generates an energy replenishment instruction in combination with a rule engine and an optimization algorithm; and a device control module that converts the instruction into a control signal to drive an execution mechanism to complete a replenishment task and feeds back a device operating state. The present application realizes centralized processing and intelligent scheduling of multi-scene gas station energy data through modular closed-loop control, effectively improving energy management efficiency and replenishment automation level.

Description

Technical Field

[0001] This invention relates to the field of intelligent energy management and automated control technology, and in particular to a multi-scenario centralized energy management system and automated intelligent replenishment equipment. Background Technology

[0002] The multi-scenario centralized energy management system is an architecture that integrates energy data from multiple gas stations based on a central control platform. It collects information such as fuel inventory, consumption rate and environmental parameters in real time, and optimizes energy allocation plans by combining data analysis models to achieve efficient resource allocation. The automated intelligent replenishment equipment relies on embedded sensing units and actuators to autonomously complete fuel refueling or energy storage medium replenishment tasks according to the instructions issued by the management system.

[0003] In existing gas station operations, due to the wide distribution and independent management of stations, data collection mainly relies on manual operations, such as manually recording refueling transactions and inventory changes, leading to data update delays and insufficient accuracy. For example, gas station attendants need to enter daily refueling volumes afterward, which may cause data inconsistencies or omissions; at the same time, tank level monitoring relies on periodic checks and cannot provide real-time feedback on inventory status, affecting the efficiency of energy replenishment scheduling. This decentralized management architecture lacks centralized data processing capabilities, making the monitoring system slow to respond and hindering automated replenishment decisions from being optimized based on real-time data, thereby reducing overall operational efficiency. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a multi-scenario centralized energy management system and automated intelligent replenishment equipment. This invention solves the technical problems of untimely data collection, low monitoring efficiency, and insufficient automation of energy replenishment caused by the decentralized management of gas stations and reliance on manual operation.

[0005] To solve the above-mentioned technical problems, the specific contents of the present invention are as follows:

[0006] In a first aspect, the multi-scenario centralized energy management system provided by the present invention includes: The data acquisition module is deployed at multiple gas stations. It collects energy-related data through a sensor array composed of multiple sensors. The energy-related data includes fuel inventory level, refueling transaction volume and ambient temperature. The energy-related data is then sent to the communication transmission module. The communication transmission module receives energy-related data from the data acquisition module, encapsulates the energy-related data to generate standardized data packets, and transmits the standardized data packets to the central control platform via a wireless network. The central processing module, deployed on the central control platform, receives standardized data packets from the communication transmission module, performs data cleaning and data storage on the energy-related data in the standardized data packets, obtains cleaned data, inputs the cleaned data into a machine learning model for time series prediction and anomaly detection, generates a processing result including inventory warning signs, and sends the processing result to the decision generation module. The decision generation module receives the processing results from the central processing module, inputs the processing results into the rule engine and optimization algorithm, generates an energy replenishment instruction including the replenishment amount and the target site, and sends the energy replenishment instruction to the equipment control module. The equipment control module receives energy replenishment instructions from the decision generation module, converts the energy replenishment instructions into control signals and sends them to the automated intelligent replenishment equipment, drives the actuator of the automated intelligent replenishment equipment to complete the replenishment task, collects the operating status data of the automated intelligent replenishment equipment, and sends the operating status data to the communication transmission module.

[0007] Furthermore, the multi-scenario centralized energy management system of the present invention also includes: The communication transmission module receives operating status data from the device control module and transmits the operating status data to the central processing module; The central processing module uses the running status data to adjust the parameters of the machine learning model, and uses the adjusted machine learning model to perform time series prediction and anomaly detection on the subsequently received and cleaned data, generating new processing results, and sending the new processing results to the decision generation module. The decision generation module receives the new processing result, inputs the new processing result into the rule engine and optimization algorithm, generates an updated energy replenishment instruction, and sends the updated energy replenishment instruction to the device control module. The data acquisition module receives updated energy replenishment instructions from the decision generation module, and adjusts the data acquisition strategy of the sensor array composed of multiple sensors according to the target station and replenishment amount in the updated energy replenishment instructions. When the data acquisition module collects energy-related data: The capacitance change signal is obtained through the level probe in the sensor array composed of multiple sensors, the pulse signal is obtained through the flow meter in the sensor array composed of multiple sensors, and the thermocouple signal is obtained through the temperature sensor in the sensor array composed of multiple sensors. The capacitance change signal, the pulse signal, and the thermocouple signal are amplified using an instrumentation amplifier, and high-frequency noise is filtered out from the amplified signals using an active low-pass filter. The analog signal after high-frequency noise is filtered out is input into the analog-to-digital converter to obtain the corresponding digital signal; A medium-range filter is applied to the digital signal to remove pulse interference. The digital signal after removing pulse interference is then subjected to a minimum-maximum normalization method to map the numerical ranges of the digital signals converted from the capacitance change signal, the pulse signal, and the thermocouple signal to a unified interval, thereby generating structured data. The structured data is appended with a timestamp and the corresponding sensor identifier to generate preprocessed structured data, and the preprocessed structured data is sent to the communication transmission module.

[0008] Furthermore, the multi-scenario centralized energy management system of the present invention also includes: After determining the abnormal state of the device based on the operating status data, the device control module sends the device abnormal state code through the User Datagram Protocol broadcast port. The data acquisition module receives the device's abnormal status code, triggers an interrupt service routine, and increases the data sampling frequency of the sensor array composed of multiple sensors from the reference frequency to a preset higher frequency.

[0009] Furthermore, in the multi-scenario centralized energy management system of the present invention, the communication transmission module is configured as follows: A frame header including a station identifier and a sequence number is added to the preprocessed structured data from the data acquisition module. A cyclic redundancy check code is calculated for the data after adding the frame header. The cyclic redundancy check code is appended to the end of the data to form a transmission message. Before sending the transmission message, monitor the signal strength and transmission delay of the primary 5G wireless network; When the signal strength of the primary 5G wireless network is detected to be lower than the first threshold, or the transmission delay is detected to be higher than the second threshold, the transmission channel is switched from the primary 5G wireless network to the backup 4G wireless network. Using the channel of the switched or primary 5G wireless network, a data compression algorithm is selected from a preset set of data compression algorithms based on the current network transmission delay value to compress the transmission message, and the compressed transmission message is sent to the central control platform.

[0010] Furthermore, in the multi-scenario centralized energy management system of the present invention, the central processing module is configured as follows: Integrity verification is performed on the transmission messages from the communication transmission module, and energy-related data is extracted from the transmission messages that pass the verification. The rule engine is used to identify and remove outliers in the extracted energy-related data. For time series data points with missing data, linear interpolation is used to generate imputation values ​​to obtain cleaned data. The cleaned data is written into a time-series database in chronological order, and an inverted index is created for the cleaned data written into the time-series database. The cleaned data is input into a long short-term memory network model, which uses a sliding window of a preset time length to extract the cleaned data for training and learning, and outputs the inventory prediction value for future periods. The cleaned data is input into the isolated forest algorithm model, which calculates the path length of each data point in the cleaned data, calculates anomaly scores based on the path lengths, and identifies and outputs anomaly scores that characterize equipment failures or data anomalies. The inventory forecast value is combined with the anomaly score to generate the processing result including the inventory warning flag.

[0011] Furthermore, in the multi-scenario centralized energy management system of the present invention, the central processing module is further configured as follows: Read the post-cleaning operation data of multiple gas stations from the time-series database, and input the operation data into the K-means clustering algorithm; The K-means clustering algorithm calculates the sum of squared errors for different numbers of clusters based on the elbow rule, selects the value corresponding to the inflection point of the sum of squared errors as the optimal number of clusters, and clusters the running data according to the optimal number of clusters, assigning a category label to each group. A mapping relationship is established between the category labels and the corresponding group of gas station operation data. Based on the mapping relationship, a personalized management strategy is generated. The personalized management strategy includes the data monitoring frequency and early warning threshold settings corresponding to different category labels. The personalized management strategy is sent to the decision generation module.

[0012] Furthermore, in the multi-scenario centralized energy management system of the present invention, the decision generation module is configured as follows: Receive the processing results from the central processing module; The processing result is input into the rule engine, which uses the Rete algorithm to match the processing result with predefined supply rules and outputs the initial supply scheme set corresponding to the successfully matched rules. The initial supply scheme set is encoded as a chromosome population and input into a genetic algorithm. The genetic algorithm performs selection, crossover, and mutation operations on the chromosome population, calculates the fitness value of each chromosome, and performs multiple generations of iteration with Pareto optimality as the goal, and outputs the optimal supply scheme. Meanwhile, the processing result is used as the environmental state input to the Q-learning reinforcement learning algorithm. The Q-learning reinforcement learning algorithm selects actions based on the Q-value table, simulates the long-term benefits of different supply decisions, updates the decision threshold, and outputs recommended decisions. The optimal supply scheme output by the genetic algorithm and the recommended decision input by the Q-learning reinforcement learning algorithm are weighted and voted on. A comprehensive score is calculated according to preset weights, and the scheme with the highest comprehensive score is selected as the final scheme to generate the energy supply instruction.

[0013] Furthermore, in the multi-scenario centralized energy management system of the present invention, the decision generation module is further configured as follows: Receive the supply task execution result from the device control module; When the status field in the supply task execution result indicates that the task has failed or the completion time field exceeds the preset time limit, perform the following operations: The rule engine is triggered to recalculate the priority weight of the currently active supply rule; The chromosome population of the genetic algorithm is reset, and an initial population is generated using a new random initialization method; Based on the reward signal in the result of the supply task, update the Q value of the corresponding state and action pair in the Q value table of the Q-learning reinforcement learning algorithm.

[0014] Furthermore, in the multi-scenario centralized energy management system of the present invention, the equipment control module is configured as follows: Parse the energy replenishment command from the decision generation module, and extract the replenishment amount parameter and target station parameter from the energy replenishment command; A proportional-integral-derivative control algorithm is used to calculate the control quantity value based on the feed quantity parameter, and the control quantity value is converted into a pulse width modulation signal with a corresponding duty cycle. The pulse width modulation signal is output to the motor pump and solenoid valve in the automated intelligent refueling equipment. The speed of the motor pump and the opening of the solenoid valve are controlled by adjusting the duty cycle of the pulse width modulation signal, thereby regulating the fuel flow rate. The operating current of the motor pump is collected in real time by a current sensor, and the coil temperature of the solenoid valve is collected in real time by a temperature sensor. The collected operating current and coil temperature are combined to form operating status data; The operating current value in the operating status data is compared with the preset current safety threshold, and the coil temperature value is compared with the preset temperature safety threshold. When the comparison results show that the operating current value continuously exceeds the current safety threshold for multiple consecutive sampling periods, or the coil temperature value exceeds the temperature safety threshold, the equipment is judged to be in an abnormal state. In response to the determination of abnormal equipment status, a cut-off command is sent to the power controller of the automated intelligent supply equipment, and an alarm signal including an abnormality type code and equipment identifier is generated; The alarm signal is sent to the communication transmission module.

[0015] Secondly, the present invention provides an automated intelligent replenishment device, applied to the multi-scenario centralized energy management system as described above, comprising: The sensor array, composed of multiple sensors, is deployed at the gas station site to collect energy-related data, including fuel inventory level, refueling transaction volume, and ambient temperature, and then sends the collected energy-related data to the communication interface. The communication interface receives energy replenishment instructions from the decision generation module in the multi-scenario centralized energy management system and forwards the energy replenishment instructions to the control unit; at the same time, it receives working parameters from the status monitoring unit, encapsulates the working parameters into operating status data, and sends them to the communication transmission module in the multi-scenario centralized energy management system. The control unit receives an energy replenishment command from the communication interface, parses the energy replenishment command to obtain replenishment quantity parameters and target station parameters, generates a drive signal based on the replenishment quantity parameters, and sends the drive signal to the actuator. The actuator, including a motor pump and a solenoid valve, receives a drive signal from the control unit and adjusts the motor pump speed and the solenoid valve opening according to the drive signal to perform a fuel replenishment operation. The status monitoring unit collects the operating current of the motor pump and the coil temperature of the solenoid valve in the actuator in real time as operating parameters, and sends the operating parameters to the communication interface. The control unit also receives control commands from the communication interface issued by the multi-scenario centralized energy management system, and adjusts the generation logic of the drive signal according to the control commands.

[0016] Beneficial effects of this invention: This invention provides a multi-scenario centralized energy management system and automated intelligent replenishment equipment. Through a modular architecture, it achieves closed-loop control of data acquisition, transmission, processing, decision-making, and execution, effectively improving energy management efficiency. The data acquisition module collects multi-dimensional energy data from the gas station in real time using a sensor array. Combined with adaptive network switching and data compression technology in the communication transmission module, it ensures the timeliness of data acquisition and the reliability of transmission. The central processing module uses a machine learning model to perform time-series prediction and anomaly detection on the cleaned data, accurately generating inventory warning indicators to provide precise basis for decision-making. The decision generation module integrates rule engines, genetic algorithms, and reinforcement learning algorithms to dynamically generate optimized replenishment instructions, realizing intelligent adjustment of replenishment strategies. The equipment control module drives the actuators to complete replenishment tasks through high-precision control signals and provides real-time feedback of equipment status data, forming the basis for continuous system optimization. The modules work collaboratively through bidirectional data flow, overcoming the problems of data update delays, low monitoring efficiency, and insufficient automation in existing technologies, significantly improving the automation level and system response speed of energy replenishment at multi-scenario gas stations. Attached Figure Description

[0017] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on the drawings without creative effort.

[0018] Figure 1 This is a system architecture diagram of the multi-scenario centralized energy management system of the present invention. Detailed Implementation

[0019] To make the technical solution of the present invention clearer, the present invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. The present invention provided by various embodiments will be described in detail below with reference to the accompanying drawings. To better understand the purpose of the present invention, the present invention will be described in further detail below.

[0020] Firstly, please refer to Figure 1 The multi-scenario centralized energy management system provided by this invention includes: The data acquisition module is deployed at multiple gas stations. It collects energy-related data through a sensor array composed of multiple sensors. The energy-related data includes fuel inventory level, refueling transaction volume and ambient temperature. The energy-related data is then sent to the communication transmission module. The communication transmission module receives energy-related data from the data acquisition module, encapsulates the energy-related data to generate standardized data packets, and transmits the standardized data packets to the central control platform via a wireless network. The central processing module, deployed on the central control platform, receives standardized data packets from the communication transmission module, performs data cleaning and data storage on the energy-related data in the standardized data packets, obtains cleaned data, inputs the cleaned data into a machine learning model for time series prediction and anomaly detection, generates a processing result including inventory warning signs, and sends the processing result to the decision generation module. The decision generation module receives the processing results from the central processing module, inputs the processing results into the rule engine and optimization algorithm, generates an energy replenishment instruction including the replenishment amount and the target site, and sends the energy replenishment instruction to the equipment control module. The equipment control module receives energy replenishment instructions from the decision generation module, converts the energy replenishment instructions into control signals and sends them to the automated intelligent replenishment equipment, drives the actuator of the automated intelligent replenishment equipment to complete the replenishment task, collects the operating status data of the automated intelligent replenishment equipment, and sends the operating status data to the communication transmission module.

[0021] The multi-scenario centralized energy management system integrates real-time data from multiple gas stations based on a central control platform, achieving automated energy management. The data acquisition module is deployed at the gas station, collecting energy-related data such as fuel inventory level, refueling transaction volume, and ambient temperature through a sensor array. The sensor array includes a level probe, a flow meter, and a temperature sensor. The level probe detects the liquid level based on the principle of capacitance change, the flow meter measures the flow rate through pulse signals, and the temperature sensor collects the ambient temperature. The acquired analog signals are amplified by an instrumentation amplifier, then filtered by an active low-pass filter to remove high-frequency noise, and converted into digital signals by an analog-to-digital converter. A medium-range filter is used to remove pulse interference from the digital signals, and a minimum-maximum normalization method is used to map the data to a unified range, generating structured data. The structured data is then sent to the communication transmission module after being appended with timestamps and sensor identifiers. The data acquisition module dynamically adjusts the sampling frequency based on feedback from the equipment control module, for example, increasing the sampling rate when receiving equipment abnormality codes.

[0022] The communication transmission module receives structured data from the data acquisition module, encapsulates it, and generates standardized data packets. The encapsulation process includes adding a frame header containing a station identifier and sequence number, calculating a cyclic redundancy check (CRC) code, and appending it to the data tail to form the transmission message. The communication transmission module monitors the signal strength and transmission delay of the primary 5G wireless network. When the signal strength is below a threshold or the delay is above a threshold, it switches to the backup 4G wireless network. Based on the current network delay, it selects a compression algorithm from a preset set of data compression algorithms to compress the transmission message and sends it to the central control platform via the wireless network. An acknowledgment mechanism is used during transmission to ensure data integrity.

[0023] The central processing module, deployed on the central control platform, receives standardized data packets from the communication transmission module. The module performs data integrity verification, extracts energy-related data, and uses a rule engine to identify and remove outliers. For missing data points, linear interpolation is used to generate imputation values, resulting in cleaned data. The cleaned data is written to a time-series database in chronological order, and an inverted index is created to optimize query efficiency. The cleaned data is then input into a Long Short-Term Memory (LSTM) network model for time-series forecasting, outputting inventory forecasts for future periods. Simultaneously, an Isolation Forest algorithm is input to calculate anomaly scores, identifying equipment malfunctions or data anomalies. The inventory forecasts and anomaly scores are combined to generate a processing result including inventory warning indicators, which is sent to the decision generation module. The central processing module also reads operational data from multiple gas stations from the time-series database, inputs it into a K-means clustering algorithm, and uses the elbow rule to determine the optimal number of clusters to group the gas stations and generate personalized management strategies.

[0024] The decision generation module receives the processing results from the central processing module and inputs them into the rule engine. The rule engine uses the Rete algorithm to match the processing results with predefined supply rules, outputting an initial supply scheme set. This initial supply scheme set is encoded as a chromosome population and input into a genetic algorithm. The genetic algorithm iteratively optimizes the scheme through selection, crossover, and mutation operations, aiming for Pareto optimality. Simultaneously, the processing results are used as environmental state input to a Q-learning reinforcement learning algorithm. The algorithm selects actions based on the Q-value table and updates the decision threshold, outputting a recommended decision. The optimal supply scheme from the genetic algorithm and the recommended decision from the reinforcement learning algorithm are combined using a weighted voting mechanism to calculate a comprehensive score. The scheme with the highest score is selected to generate an energy supply instruction including the supply amount and target site. The decision generation module receives the supply task execution results from the equipment control module. When a task fails or times out, it recalculates the rule priorities, resets the genetic algorithm population, and updates the Q-value table.

[0025] The equipment control module receives energy replenishment commands from the decision generation module, parses the commands to extract replenishment quantity parameters and target site parameters. It employs a preset control algorithm (proportional-integral-derivative control) to calculate control values ​​based on the replenishment quantity parameters and converts these values ​​into pulse-width modulation (PWM) signals. These PWM signals are output to the motor pump and solenoid valve of the automated intelligent replenishment equipment. By adjusting the signal duty cycle, the module controls the motor pump speed and solenoid valve opening, thereby regulating fuel flow. The equipment control module collects the motor pump operating current via a current sensor and the solenoid valve coil temperature via a temperature sensor, combining these data to form operating status data. This operating status data is compared with preset safety thresholds. If the operating current or coil temperature continuously exceeds the limit, an abnormal equipment state is determined, a cut-off command is sent to the power controller, and an alarm signal is generated. The alarm signal is sent to the communication transmission module, while the operating status data is simultaneously fed back to the central processing module for model parameter adjustment.

[0026] The system forms a closed-loop control through bidirectional data flow between modules. Real-time data from the data acquisition module is transmitted with low latency via the communication transmission module. The central processing module intelligently analyzes and optimizes the input for decision-making. The adaptive strategies of the decision generation module drive the equipment control module to execute the replenishment task. Equipment status feedback adjusts system behavior in real time. This architecture improves energy management efficiency, automates replenishment, and is suitable for energy management in various scenarios such as gas stations.

[0027] During system operation, the communication transmission module continuously monitors the operational status data messages sent by the device control module via an embedded network interface. These messages are encapsulated using a lightweight transmission protocol and include the device identifier, operational parameters, and a timestamp. The communication transmission module parses the message content, verifies the data format, and then re-encapsulates it into standardized data packets, which are then transmitted to the central processing module via the wireless network link. An acknowledgment mechanism is employed during transmission, awaiting a response from the receiver to ensure data integrity.

[0028] After receiving operational status data, the central processing module merges it with historical datasets for parameter optimization of the machine learning model. The weight matrix of the Long Short-Term Memory (LSTM) network model is adjusted using a gradient descent algorithm to adapt to changes in equipment status. The adjusted model then performs time-series forecasting on subsequently cleaned data, outputting more accurate inventory predictions. Simultaneously, the tree structure parameters of the Isolation Forest algorithm model are updated based on operational status data, improving anomaly detection sensitivity.

[0029] The decision generation module receives new processing results, triggering the rule engine to re-evaluate the replenishment rules. The Rete algorithm quickly matches the current state with the rule base conditions to generate an initial set of replenishment schemes. The genetic algorithm reinitializes the population, using chromosome encoding to represent replenishment parameters, and iteratively optimizes through selection, crossover, and mutation operations. The reinforcement learning algorithm adjusts the decision threshold based on the Q-value table to simulate long-term gains. A weighted voting mechanism integrates the outputs of multiple algorithms to generate updated energy replenishment instructions.

[0030] After receiving the update command, the data acquisition module dynamically adjusts the acquisition strategy of the sensor array. For the target site, the microprocessor modifies the sampling timer settings, prioritizing the acquisition of high-frequency liquid level data. During acquisition, the liquid level probe detects capacitance changes and outputs an analog voltage signal, the flow meter generates a pulse sequence, and the temperature sensor provides thermocouple voltage. The instrumentation amplifier amplifies the weak signal, and an active low-pass filter suppresses high-frequency noise. The amplified analog signal is converted to a digital signal by an analog-to-digital converter. A medium-range filter is used to eliminate pulse interference, and a minimum-maximum normalization method maps the digital signal to the zero-to-one interval. The processed data is appended with a precise timestamp and sensor identifier, forming a structured message that is sent to the communication transmission module.

[0031] The equipment control module monitors the operating parameters of the actuators in real time. Current sensors collect the operating current of the motor pump, and temperature sensors collect the temperature of the solenoid valve coil. When the current value exceeds the safety threshold or the temperature value exceeds the limit within a continuous sampling period, the state machine model determines an abnormal equipment state. The control unit immediately sends an abnormal state code via the user datagram protocol broadcast port. The code includes an abnormality type code and a unique equipment identifier.

[0032] The network interface of the data acquisition module listens for User Datagram Protocol (UDP) broadcast messages. Upon receiving an abnormal status code, the microprocessor triggers a high-priority interrupt service routine. This interrupt routine modifies the clock divider settings, increasing the sensor sampling frequency from the baseline value to a higher frequency, enabling intensive data acquisition. For example, the sampling rate of a liquid level sensor is increased to quickly capture abnormal fluctuations.

[0033] The communication transmission module adds a frame header to the preprocessed structured data. The frame header fields include the gas station identifier and a global sequence number. The module calculates a cyclic redundancy check (CRC) code, generates a checksum using a standard polynomial algorithm, and appends it to the end of the message to form a complete transmission message. Before transmission, the communication module monitors the received signal strength indicator and round-trip time (RTT) of the primary 5G wireless network. When the signal strength falls below a threshold or the delay exceeds a threshold, the channel management unit automatically switches to the backup 4G wireless network. The switching process uses a link-keeping protocol to maintain session continuity. Based on the real-time network delay value, the compression algorithm selector dynamically selects a compression method from the algorithm library to compress the transmission message, reducing the data size before transmission.

[0034] The central processing module verifies the received transmission messages using cyclic redundancy check (CRC) codes. After confirming data integrity, it extracts energy-related data fields. The rule engine uses statistical methods to identify outliers, such as removing abnormal values ​​based on the three-standard-deviation principle. For missing points in the time series, linear interpolation between adjacent data points is used to calculate imputation values. The cleaned data is stored in the time-series database in timestamp order, and the database engine automatically creates an inverted index to optimize time-range queries. The Long Short-Term Memory (LSTM) network model is trained using a sliding window approach to extract historical data. A gating mechanism learns long-term dependencies and outputs inventory forecasts. The Isolation Forest algorithm constructs a random tree structure and calculates anomaly scores based on the path length of data points. The predicted values ​​and scores are combined to generate inventory warning indicators, constituting the processing result.

[0035] The central processing module reads historical operational data from multiple gas stations from a time-series database. This data includes dimensions such as fuel inventory levels, refueling transaction volume, and ambient temperature. The data preprocessing unit standardizes the raw data, eliminating dimensional differences, and constructs a feature matrix. The K-means clustering algorithm uses Euclidean distance as a similarity measure to calculate the sum of squared errors within clusters under different numbers of clusters. The algorithm plots a curve of the sum of squared errors as a function of the number of clusters based on the elbow rule, identifying the value corresponding to the inflection point of the curve as the optimal number of clusters. The clustering process employs an iterative optimization approach, randomly initializing cluster centers and continuously reallocating data points and updating cluster centers until the cluster allocation stabilizes.

[0036] After clustering, the system assigns business-meaning category labels to each group. For example, high-traffic gas stations are labeled as category A, and low-traffic stations as category B. A mapping relationship between category labels and gas station operational characteristics is established based on the clustering results, forming group profiles. The personalized management strategy engine dynamically generates monitoring parameter configurations according to the operational characteristics of different groups. For category A high-traffic stations, a higher data collection frequency and a stricter inventory warning threshold are set; for category B low-traffic stations, a normal monitoring frequency and a relatively lenient threshold setting are used. The strategy is sent to the decision generation module via configuration messages, providing differentiated judgment criteria for subsequent decisions.

[0037] After receiving the processing results, the decision generation module uses the Rete algorithm for pattern matching in the rule engine. The algorithm matches facts such as inventory warning indicators and anomaly scores from the processing results with conditions in the rule base to generate an initial set of replenishment schemes that meet the triggering conditions. The rule matching process optimizes matching efficiency by constructing network nodes; when new facts are added to the working memory, pattern matching is performed along the network nodes.

[0038] The genetic algorithm encodes the initial resupply plan as a population of chromosomes, each chromosome including decision variables such as resupply amount and resupply time. The fitness function comprehensively considers multiple objectives such as transportation cost, inventory turnover rate, and resupply efficiency. The algorithm retains high-quality individuals through a roulette wheel selection mechanism and exchanges chromosome segments using a single-point crossover operation, introducing diversity through basic positional mutations. During the iteration process, the Pareto front solution set is recorded, and the final output non-dominated optimal solution is the optimal resupply plan.

[0039] Meanwhile, the Q-learning reinforcement learning algorithm treats the processing results as the environment state, constructing a state-action value function table. The algorithm employs an ε-greedy strategy to balance exploration and exploitation, iteratively updating the Q-value through the Bellman equation. The reward function design includes immediate rewards and long-term benefits; the algorithm simulates the cumulative rewards of different replenishment decisions, outputting a recommendation decision based on value iteration.

[0040] The weighted voting mechanism assigns fixed weights to both the genetic algorithm and reinforcement learning decisions, calculating a comprehensive score for each scheme. The score calculation considers factors such as scheme feasibility, cost-effectiveness, and risk coefficients, ultimately selecting the scheme with the highest score to generate the energy replenishment command. A weight calibration mechanism is incorporated into the voting process, dynamically adjusting the weight allocation based on historical performance.

[0041] Upon receiving the supply task execution results from the device control module, the decision generation module initiates the strategy optimization process. The rule engine parses the status fields in the execution results, and when task failure or timeout is detected, it recalculates the priority weights of each supply rule. The weight adjustment is based on the historical execution success rate and urgency of the rules, increasing the trigger probability of high-priority rules.

[0042] The genetic algorithm employs a novel initialization method to reconstruct the chromosome population, increasing population diversity. The initialization process incorporates the distribution characteristics of historical high-quality solutions, intelligently allocating points within the solution space. The Q-learning algorithm updates the Q-value table based on reward signals from the execution results, decreasing the Q-value for state-action pairs that lead to task failure and increasing the Q-value for successful decisions. The update process utilizes a temporal difference learning method to progressively optimize long-term decision quality.

[0043] When the equipment control module parses the energy replenishment command, the command parsing unit extracts the replenishment quantity parameter and target station identifier from the command. The proportional-integral-derivative (PID) control algorithm calculates the control quantity based on the replenishment quantity deviation. The proportional term processes the current error, the integral term eliminates the accumulated error, and the derivative term predicts the trend of change. After the algorithm outputs the control quantity value, the signal conversion unit converts it into a pulse width modulation signal with the corresponding duty cycle.

[0044] A pulse width modulation (PWM) signal drives the actuator, controlling the motor pump speed and solenoid valve opening by adjusting the signal duty cycle. The motor pump speed is linearly related to the PWM signal duty cycle, while the solenoid valve opening increases with increasing duty cycle. The status monitoring unit collects real-time data on the motor pump operating current and the solenoid valve coil temperature. The current sensor uses the Hall effect principle to detect the current value, and the temperature sensor measures the temperature based on the resistance-temperature characteristic.

[0045] The equipment control module combines the collected operating current and coil temperature into operating status data, which is then compared and analyzed with preset safety thresholds. The comparator uses a window comparison method; when the operating current continuously exceeds the current safety threshold within a continuous sampling period, or the coil temperature momentarily exceeds the temperature safety threshold, the state machine transitions to an abnormal state. The abnormal judgment logic incorporates a debouncing mechanism to avoid misjudgments caused by instantaneous fluctuations.

[0046] Upon determining an abnormal equipment condition, the control unit immediately sends a cut-off command to the power controller. This cut-off command is secured with a hard-wired connection for reliability. Simultaneously, an alarm signal is generated, including an anomaly type code and equipment identification. This alarm signal is uploaded to the central processing platform via a communication transmission module. A priority transmission mechanism is used for the alarm signal to minimize anomaly response time.

[0047] Secondly, the present invention provides an automated intelligent replenishment device, applied to the multi-scenario centralized energy management system as described above, comprising: The sensor array, composed of multiple sensors, is deployed at the gas station site to collect energy-related data, including fuel inventory level, refueling transaction volume, and ambient temperature, and then sends the collected energy-related data to the communication interface. The communication interface receives energy replenishment instructions from the decision generation module in the multi-scenario centralized energy management system and forwards the energy replenishment instructions to the control unit; at the same time, it receives working parameters from the status monitoring unit, encapsulates the working parameters into operating status data, and sends them to the communication transmission module in the multi-scenario centralized energy management system. The control unit receives an energy replenishment command from the communication interface, parses the energy replenishment command to obtain replenishment quantity parameters and target station parameters, generates a drive signal based on the replenishment quantity parameters, and sends the drive signal to the actuator. The actuator, including a motor pump and a solenoid valve, receives a drive signal from the control unit and adjusts the motor pump speed and the solenoid valve opening according to the drive signal to perform a fuel replenishment operation. The status monitoring unit collects the operating current of the motor pump and the coil temperature of the solenoid valve in the actuator in real time as operating parameters, and sends the operating parameters to the communication interface. The control unit also receives control commands from the communication interface issued by the multi-scenario centralized energy management system, and adjusts the generation logic of the drive signal according to the control commands.

[0048] During system operation, the communication transmission module continuously monitors the operational status data messages sent by the device control module via an embedded network interface. These messages are encapsulated using a lightweight transmission protocol and include the device identifier, operational parameters, and a timestamp. The communication transmission module parses the message content, verifies the data format, and then re-encapsulates it into standardized data packets, which are then transmitted to the central processing module via the wireless network link. An acknowledgment mechanism is employed during transmission, awaiting a response from the receiver to ensure data integrity.

[0049] After receiving operational status data, the central processing module merges it with historical datasets for parameter optimization of the machine learning model. The weight matrix of the Long Short-Term Memory (LSTM) network model is adjusted using a gradient descent algorithm to adapt to changes in equipment status. The adjusted model then performs time-series forecasting on subsequently cleaned data, outputting more accurate inventory predictions. Simultaneously, the tree structure parameters of the Isolation Forest algorithm model are updated based on operational status data, improving anomaly detection sensitivity.

[0050] The decision generation module receives new processing results, triggering the rule engine to re-evaluate the replenishment rules. The Rete algorithm quickly matches the current state with the rule base conditions to generate an initial set of replenishment schemes. The genetic algorithm reinitializes the population, using chromosome encoding to represent replenishment parameters, and iteratively optimizes through selection, crossover, and mutation operations. The reinforcement learning algorithm adjusts the decision threshold based on the Q-value table to simulate long-term gains. A weighted voting mechanism integrates the outputs of multiple algorithms to generate updated energy replenishment instructions.

[0051] After receiving the update command, the data acquisition module dynamically adjusts the acquisition strategy of the sensor array. For the target site, the microprocessor modifies the sampling timer settings, prioritizing the acquisition of high-frequency liquid level data. During acquisition, the liquid level probe detects capacitance changes and outputs an analog voltage signal, the flow meter generates a pulse sequence, and the temperature sensor provides thermocouple voltage. The instrumentation amplifier amplifies the weak signal, and an active low-pass filter suppresses high-frequency noise. The amplified analog signal is converted to a digital signal by an analog-to-digital converter. A medium-range filter is used to eliminate pulse interference, and a minimum-maximum normalization method maps the digital signal to the zero-to-one interval. The processed data is appended with a precise timestamp and sensor identifier, forming a structured message that is sent to the communication transmission module.

[0052] The equipment control module monitors the operating parameters of the actuators in real time. Current sensors collect the operating current of the motor pump, and temperature sensors collect the temperature of the solenoid valve coil. When the current value exceeds the safety threshold or the temperature value exceeds the limit within a continuous sampling period, the state machine model determines an abnormal equipment state. The control unit immediately sends an abnormal state code via the user datagram protocol broadcast port. The code includes an abnormality type code and a unique equipment identifier.

[0053] The network interface of the data acquisition module listens for User Datagram Protocol (UDP) broadcast messages. Upon receiving an abnormal status code, the microprocessor triggers a high-priority interrupt service routine. This interrupt routine modifies the clock divider settings, increasing the sensor sampling frequency from the baseline value to a higher frequency, enabling intensive data acquisition. For example, the sampling rate of a liquid level sensor is increased to quickly capture abnormal fluctuations.

[0054] The communication transmission module adds a frame header to the preprocessed structured data. The frame header fields include the gas station identifier and a global sequence number. The module calculates a cyclic redundancy check (CRC) code, generates a checksum using a standard polynomial algorithm, and appends it to the end of the message to form a complete transmission message. Before transmission, the communication module monitors the received signal strength indicator and round-trip time (RTT) of the primary 5G wireless network. When the signal strength falls below a threshold or the delay exceeds a threshold, the channel management unit automatically switches to the backup 4G wireless network. The switching process uses a link-keeping protocol to maintain session continuity. Based on the real-time network delay value, the compression algorithm selector dynamically selects a compression method from the algorithm library to compress the transmission message, reducing the data size before transmission.

[0055] The central processing module verifies the received transmission messages using cyclic redundancy check (CRC) codes. After confirming data integrity, it extracts energy-related data fields. The rule engine uses statistical methods to identify outliers, such as removing abnormal values ​​based on the three-standard-deviation principle. For missing points in the time series, linear interpolation between adjacent data points is used to calculate imputation values. The cleaned data is stored in the time-series database in timestamp order, and the database engine automatically creates an inverted index to optimize time-range queries. The Long Short-Term Memory (LSTM) network model is trained using a sliding window approach to extract historical data. A gating mechanism learns long-term dependencies and outputs inventory forecasts. The Isolation Forest algorithm constructs a random tree structure and calculates anomaly scores based on the path length of data points. The predicted values ​​and scores are combined to generate inventory warning indicators, constituting the processing result.

[0056] In the central processing module, when the Long Short-Term Memory (LSTM) network model performs time series prediction, it uses the following formulas to calculate the forget gate, input gate, cell state, and output gate: ; in, Indicates the forget gate at time step The output value, This represents the sigmoid activation function. The weight matrix represents the forget gate. This represents the hidden state of the previous time step. This represents the input data at the current time step. The bias term representing the forget gate. This represents matrix multiplication. Indicates the vector and Perform the splicing operation.

[0057] ; in, Indicates the input gate at time step The output value, This represents the weight matrix of the input gate. This represents the bias term of the input gate.

[0058] ; in, Indicates the state of candidate cells. This represents the hyperbolic tangent activation function. The weight matrix representing the cell state. Bias terms representing cell state.

[0059] ; in, This indicates the cell state at the current time step. This indicates the cell state at the previous time step. This indicates element-wise multiplication.

[0060] ; in, Indicates the output gate at time step The output value, This represents the weight matrix of the output gate. This represents the bias term of the output gate.

[0061] ; in, This represents the hidden state at the current time step. The model trains the weight matrix and bias term by minimizing the prediction error, and outputs the inventory prediction value for future periods.

[0062] The Isolation Forest algorithm uses the following formula to calculate anomaly scores: ; in, Representing data points Abnormal scores, Representing data points The expected value of the path length in an isolated tree. Representing data points The path length from the root node to the leaf node. Represents a given number of samples The average path length at that time is used for normalization. This indicates the number of training samples.

[0063] When the K-means clustering algorithm determines the optimal number of clusters based on the elbow rule, it calculates the sum of squared errors: ; in, Represents the sum of squared errors. Indicates the number of clusters. Indicates the number of data points. Indicates the first Data points, Indicates the first The center point of each cluster, Represents Euclidean distance. This indicates a summation operation.

[0064] In the decision generation module, when the genetic algorithm optimizes the supply scheme, the fitness function is defined as: ; in, This represents the fitness value of a chromosome. This represents the total cost of the supply plan. The selection operation uses a roulette wheel strategy, and the probability of a chromosome being selected is: ; in, Indicates the first The probability of selection for each chromosome Indicates the first Fitness values ​​of each chromosome Indicates population size.

[0065] When updating the Q-value table, the Q-learning reinforcement learning algorithm uses the following formula: ; in, Indicates the state Next action Q value, Indicates the learning rate. This indicates an immediate reward. Indicates the discount factor. This represents the maximum Q value of the next state. Indicates the current state. Indicates the current action. Indicates the next state.

[0066] The weighted voting mechanism uses the following formula to calculate the overall score: ; in, This represents the overall score. This represents the optimal supply scheme output by the genetic algorithm. The weight, This represents the recommendation decision output by the Q-learning algorithm. The weight, It represents the multiplication operation.

[0067] In data preprocessing, the formula for the min-max normalization method is: ; in, This represents the normalized data. Represents the original data. Represents the dataset The minimum value, Represents the dataset The maximum value. The formula for linear interpolation to fill in missing values ​​is: ; in, Indicates a point in time The fill value, This represents data from the previous point in time. This indicates data at the next point in time.

[0068] In the implementation of the multi-scenario centralized energy management system, the system achieves centralized processing and intelligent control of gas station energy data through a modular architecture. Data acquisition modules are deployed at each gas station. Level probes in the sensor array detect fuel inventory levels based on capacitance changes, flow meters measure refueling transactions via pulse signals, and temperature sensors collect ambient temperature data. The acquired analog signals are amplified by an instrumentation amplifier and then filtered out for high-frequency noise by an active low-pass filter. An analog-to-digital converter converts the processed analog signals into digital signals, a median filter eliminates pulse interference, and a minimum-maximum normalization method maps the data to a uniform numerical range. The pre-processed data is appended with timestamps and sensor identifiers, forming structured data packets that are sent to the communication transmission module.

[0069] After receiving structured data, the communication transmission module adds a frame header including a site identifier and calculates a cyclic redundancy check (CRC) code to construct the transmission message. The module monitors the 5G wireless network signal strength in real time and automatically switches to a backup 4G network channel when the signal quality falls below a threshold. Based on network latency, it dynamically selects a data compression algorithm to compress the message before transmitting it to the central control platform via the wireless network. An acknowledgment and retransmission mechanism ensures data integrity during transmission.

[0070] The central processing module verifies the integrity of received messages and extracts energy data. The rule engine identifies and removes outliers, and fills in missing data using linear interpolation. The cleaned data is stored in a database in time series format. A Long Short-Term Memory (LSTM) network model learns historical consumption patterns using a sliding window approach and outputs inventory forecasts. An Isolation Forest algorithm calculates anomaly scores based on path length to identify equipment failure risks. The forecasts and anomaly scores are combined to generate a processing result including warning indicators, which is then pushed to the decision generation module.

[0071] The decision generation module's rule engine uses the Rete algorithm to match predefined supply rules and generate an initial supply scheme set. A genetic algorithm encodes the schemes into a chromosome population and iteratively optimizes them through selection, crossover, and mutation operations. A Q-learning reinforcement learning algorithm simulates the long-term payoffs of different decisions based on a Q-value table and outputs recommended schemes. A weighted voting mechanism integrates the results of both algorithms to generate energy supply instructions that include the supply amount and the target site.

[0072] After parsing the instructions, the equipment control module uses a proportional-integral-derivative (PI-DE) control algorithm to calculate the control quantity, which is then converted into a pulse-width modulation (PWM) signal to drive the actuator. The motor pump adjusts its speed according to the pulse signal, and the solenoid valve controls its opening to achieve precise fuel replenishment. A current sensor collects the operating current in real time, and a temperature sensor monitors the coil temperature; the data is combined to provide the operating status feedback to the system. When a continuous overcurrent or abnormal temperature is detected, the power supply is immediately cut off and an alarm signal is sent.

[0073] During system operation, the data acquisition module dynamically adjusts the sampling frequency based on equipment status. When the communication transmission module receives an exception code from the equipment control module, it triggers an interrupt service routine to increase the sampling rate. The central processing module uses operational status data to adjust the parameters of the machine learning model, and the decision generation module updates the replenishment strategy based on real-time feedback, forming a closed-loop optimization. This architecture enables the system to adaptively respond to complex scenarios such as gas station flow fluctuations and equipment failures, achieving fully automated management of the energy replenishment process.

[0074] In a typical application scenario, an abnormal fluctuation in the liquid level data at a gas station prompts the data acquisition module to increase the sampling frequency and capture detailed changes. The communication transmission module prioritizes transmitting the abnormal data to the central platform. The central processing module uses an isolated forest algorithm to identify equipment failure risks, and the decision generation module immediately generates maintenance instructions. Simultaneously, the system automatically dispatches backup gas stations to increase replenishment, ensuring a stable regional energy supply. The response process is completed within seconds, significantly improving operational efficiency and safety performance.

[0075] In a typical application embodiment of a multi-scenario centralized energy management system, the system achieves real-time acquisition and intelligent replenishment of energy data at gas stations through a modular architecture. The data acquisition module is deployed at multiple gas stations. A level probe in the sensor array detects the fuel inventory level based on capacitance changes, a flow meter measures refueling transactions via pulse signals, and a temperature sensor collects ambient temperature data. The acquired analog signals are amplified by an instrumentation amplifier, filtered by an active low-pass filter to remove high-frequency noise, and converted to digital format by an analog-to-digital converter. A medium-range filter is used to eliminate pulse interference in the digital signals, and a minimum-maximum normalization method maps the data to the zero-to-one interval, generating structured data with timestamps. The communication transmission module adds station identifiers and sequence number frame headers to the data and constructs transmission messages after calculating cyclic redundancy check codes. The module dynamically monitors the 5G wireless network signal quality, automatically switching to a 4G backup network when the signal strength falls below a threshold, and optimizing transmission efficiency by selecting a compression algorithm based on real-time latency. The central processing module performs integrity verification on the received data, a rule engine identifies outliers and fills in missing values ​​using linear interpolation, and the cleaned data is stored in a time-series database. The Long Short-Term Memory (LSTM) network model is trained using 30 days of historical data as a sliding window, outputting inventory forecasts for the next seven days. The Isolation Forest algorithm simultaneously calculates anomaly scores, combining them to generate inventory warning indicators. The decision generation module matches replenishment strategies using the Rete rule engine. A genetic algorithm optimizes replenishment schemes based on multiple objectives, including transportation costs and inventory turnover rate. A Q-learning reinforcement learning algorithm simulates long-term returns to generate recommended decisions, and a weighted voting mechanism comprehensively outputs the optimal replenishment command. After parsing the command, the equipment control module uses a proportional-integral-derivative (PID) controller to calculate the control quantity and generate a pulse width modulation (PWM) signal, driving the motor pump and solenoid valve to complete fuel replenishment. Operating status data is fed back to the system in real time, forming a closed-loop control flow of acquisition, transmission, processing, decision-making, and execution, achieving fully automated management of gas station group resource scheduling.

[0076] In the equipment failure emergency handling embodiment, the system exhibits dynamic adaptive capabilities. When the operating current of a gas station's motor pump continuously exceeds the safety threshold, the equipment control module immediately triggers the abnormal diagnosis process. The current data collected by the status monitoring unit broadcasts the abnormal code via the User Datagram Protocol (UDP). After receiving the code, the data acquisition module increases the sensor sampling frequency to three times the baseline value, and the level probe captures inventory fluctuation details in a high-frequency mode. The communication transmission module prioritizes abnormal data as an emergency transmission level and uses a minimum compression delay algorithm to ensure data real-time performance. The central processing module isolates abnormal site data, uses the Isolation Forest algorithm to mark fault point characteristics, and dynamically adjusts the prediction model parameters using a Long Short-Term Memory (LSTM) network. The decision generation module activates the emergency rule set, the genetic algorithm reinitializes the population to generate detour resupply plans, and the Q-learning algorithm updates the state value table to reduce the scheduling weight of the fault point. The system automatically assigns resupply tasks to surrounding gas stations, and personalized management strategies temporarily adjust the monitoring frequency to once per minute. The equipment control module executes the safety protocol, cuts off the power to the faulty equipment, generates a maintenance work order, and reroutes the resupply command to a backup station.

Claims

1. A multi-scenario centralized energy management system, characterized in that, include: The data acquisition module is deployed at multiple gas stations. It collects energy-related data through a sensor array composed of multiple sensors. The energy-related data includes fuel inventory level, refueling transaction volume and ambient temperature. The energy-related data is then sent to the communication transmission module. The communication transmission module receives energy-related data from the data acquisition module, encapsulates the energy-related data to generate standardized data packets, and transmits the standardized data packets to the central control platform via a wireless network. The central processing module, deployed on the central control platform, receives standardized data packets from the communication transmission module, performs data cleaning and data storage on the energy-related data in the standardized data packets, obtains cleaned data, inputs the cleaned data into a machine learning model for time series prediction and anomaly detection, generates a processing result including inventory warning signs, and sends the processing result to the decision generation module. The decision generation module receives the processing results from the central processing module, inputs the processing results into the rule engine and optimization algorithm, generates an energy replenishment instruction including the replenishment amount and the target site, and sends the energy replenishment instruction to the equipment control module. The equipment control module receives energy replenishment instructions from the decision generation module, converts the energy replenishment instructions into control signals and sends them to the automated intelligent replenishment equipment, drives the actuator of the automated intelligent replenishment equipment to complete the replenishment task, collects the operating status data of the automated intelligent replenishment equipment, and sends the operating status data to the communication transmission module.

2. The multi-scenario centralized energy management system according to claim 1, characterized in that, Also includes: The communication transmission module receives operating status data from the device control module and transmits the operating status data to the central processing module; The central processing module uses the running status data to adjust the parameters of the machine learning model, and uses the adjusted machine learning model to perform time series prediction and anomaly detection on the subsequently received and cleaned data, generating new processing results, and sending the new processing results to the decision generation module. The decision generation module receives the new processing result, inputs the new processing result into the rule engine and optimization algorithm, generates an updated energy replenishment instruction, and sends the updated energy replenishment instruction to the device control module. The data acquisition module receives updated energy replenishment instructions from the decision generation module, and adjusts the data acquisition strategy of the sensor array composed of multiple sensors according to the target station and replenishment amount in the updated energy replenishment instructions. When the data acquisition module collects energy-related data: The capacitance change signal is obtained through the level probe in the sensor array composed of multiple sensors, the pulse signal is obtained through the flow meter in the sensor array composed of multiple sensors, and the thermocouple signal is obtained through the temperature sensor in the sensor array composed of multiple sensors. The capacitance change signal, the pulse signal, and the thermocouple signal are amplified using an instrumentation amplifier, and high-frequency noise is filtered out from the amplified signals using an active low-pass filter. The analog signal after high-frequency noise is filtered out is input into the analog-to-digital converter to obtain the corresponding digital signal; A medium-range filter is applied to the digital signal to remove pulse interference. The digital signal after removing pulse interference is then subjected to a minimum-maximum normalization method to map the numerical ranges of the digital signals converted from the capacitance change signal, the pulse signal, and the thermocouple signal to a unified interval, thereby generating structured data. The structured data is appended with a timestamp and the corresponding sensor identifier to generate preprocessed structured data, and the preprocessed structured data is sent to the communication transmission module.

3. The multi-scenario centralized energy management system according to claim 2, characterized in that, Also includes: After determining the abnormal state of the device based on the operating status data, the device control module sends the device abnormal state code through the User Datagram Protocol broadcast port. The data acquisition module receives the device's abnormal status code, triggers an interrupt service routine, and increases the data sampling frequency of the sensor array composed of multiple sensors from the reference frequency to a preset higher frequency.

4. The multi-scenario centralized energy management system according to claim 3, characterized in that, The communication transmission module is configured as follows: A frame header including a station identifier and a sequence number is added to the preprocessed structured data from the data acquisition module. A cyclic redundancy check code is calculated for the data after adding the frame header. The cyclic redundancy check code is appended to the end of the data to form a transmission message. Before sending the transmission message, monitor the signal strength and transmission delay of the primary 5G wireless network; When the signal strength of the primary 5G wireless network is detected to be lower than the first threshold, or the transmission delay is detected to be higher than the second threshold, the transmission channel is switched from the primary 5G wireless network to the backup 4G wireless network. Using the channel of the switched or primary 5G wireless network, a data compression algorithm is selected from a preset set of data compression algorithms based on the current network transmission delay value to compress the transmission message, and the compressed transmission message is sent to the central control platform.

5. The multi-scenario centralized energy management system according to claim 4, characterized in that, The central processing module is configured as follows: Integrity verification is performed on the transmission messages from the communication transmission module, and energy-related data is extracted from the transmission messages that pass the verification. The rule engine is used to identify and remove outliers in the extracted energy-related data. For time series data points with missing data, linear interpolation is used to generate imputation values ​​to obtain cleaned data. The cleaned data is written into a time-series database in chronological order, and an inverted index is created for the cleaned data written into the time-series database. The cleaned data is input into a long short-term memory network model, which uses a sliding window of a preset time length to extract the cleaned data for training and learning, and outputs the inventory prediction value for future periods. The cleaned data is input into the isolated forest algorithm model, which calculates the path length of each data point in the cleaned data, calculates anomaly scores based on the path lengths, and identifies and outputs anomaly scores that characterize equipment failures or data anomalies. The inventory forecast value is combined with the anomaly score to generate the processing result including the inventory warning flag.

6. The multi-scenario centralized energy management system according to claim 5, characterized in that, The central processing module is also configured to: Read the post-cleaning operation data of multiple gas stations from the time-series database, and input the operation data into the K-means clustering algorithm; The K-means clustering algorithm calculates the sum of squared errors for different numbers of clusters based on the elbow rule, selects the value corresponding to the inflection point of the sum of squared errors as the optimal number of clusters, and clusters the running data according to the optimal number of clusters, assigning a category label to each group. A mapping relationship is established between the category labels and the corresponding group of gas station operation data. Based on the mapping relationship, a personalized management strategy is generated. The personalized management strategy includes the data monitoring frequency and early warning threshold settings corresponding to different category labels. The personalized management strategy is sent to the decision generation module.

7. The multi-scenario centralized energy management system according to claim 6, characterized in that, The decision generation module is configured as follows: Receive the processing results from the central processing module; The processing result is input into the rule engine, which uses the Rete algorithm to match the processing result with predefined supply rules and outputs the initial supply scheme set corresponding to the successfully matched rules. The initial supply scheme set is encoded as a chromosome population and input into a genetic algorithm. The genetic algorithm performs selection, crossover, and mutation operations on the chromosome population, calculates the fitness value of each chromosome, and performs multiple generations of iteration with Pareto optimality as the goal, and outputs the optimal supply scheme. Meanwhile, the processing result is used as the environmental state input to the Q-learning reinforcement learning algorithm. The Q-learning reinforcement learning algorithm selects actions based on the Q-value table, simulates the long-term benefits of different supply decisions, updates the decision threshold, and outputs recommended decisions. The optimal replenishment scheme output by the genetic algorithm and the recommended decision input by the Q-learning reinforcement learning algorithm are weighted and voted on. A comprehensive score is calculated according to preset weights, and the scheme with the highest comprehensive score is selected as the final scheme to generate the energy replenishment command.

8. The multi-scenario centralized energy management system according to claim 7, characterized in that, The decision generation module is also configured to: Receive the supply task execution result from the device control module; When the status field in the supply task execution result indicates that the task has failed or the completion time field exceeds the preset time limit, perform the following operations: The rule engine is triggered to recalculate the priority weight of the currently active supply rule; The chromosome population of the genetic algorithm is reset, and an initial population is generated using a new random initialization method; Based on the reward signal in the result of the supply task, update the Q value of the corresponding state and action pair in the Q value table of the Q-learning reinforcement learning algorithm.

9. The multi-scenario centralized energy management system according to claim 8, characterized in that, The device control module is configured as follows: Parse the energy replenishment command from the decision generation module, and extract the replenishment amount parameter and target station parameter from the energy replenishment command; A preset control algorithm is used to calculate the control quantity value based on the supply quantity parameter, and the control quantity value is converted into a pulse width modulation signal with a corresponding duty cycle. The pulse width modulation signal is output to the motor pump and solenoid valve in the automated intelligent refueling equipment. The speed of the motor pump and the opening of the solenoid valve are controlled by adjusting the duty cycle of the pulse width modulation signal, thereby regulating the fuel flow rate. The operating current of the motor pump is collected in real time by a current sensor, and the coil temperature of the solenoid valve is collected in real time by a temperature sensor. The collected operating current and coil temperature are combined to form operating status data; The operating current value in the operating status data is compared with the preset current safety threshold, and the coil temperature value is compared with the preset temperature safety threshold. When the comparison results show that the operating current value continuously exceeds the current safety threshold for multiple consecutive sampling periods, or the coil temperature value exceeds the temperature safety threshold, the equipment is judged to be in an abnormal state. In response to the determination of abnormal equipment status, a cut-off command is sent to the power controller of the automated intelligent supply equipment, and an alarm signal including an abnormality type code and equipment identifier is generated; The alarm signal is sent to the communication transmission module.

10. An automated intelligent replenishment device, applied to a multi-scenario centralized energy management system as described in any one of claims 1 to 9, characterized in that, include: The sensor array, composed of multiple sensors, is deployed at the gas station site to collect energy-related data, including fuel inventory level, refueling transaction volume, and ambient temperature, and then sends the collected energy-related data to the communication interface. The communication interface receives energy replenishment instructions from the decision generation module in the multi-scenario centralized energy management system and forwards the energy replenishment instructions to the control unit; at the same time, it receives working parameters from the status monitoring unit, encapsulates the working parameters into operating status data, and sends them to the communication transmission module in the multi-scenario centralized energy management system. The control unit receives an energy replenishment command from the communication interface, parses the energy replenishment command to obtain replenishment quantity parameters and target station parameters, generates a drive signal based on the replenishment quantity parameters, and sends the drive signal to the actuator. The actuator, including a motor pump and a solenoid valve, receives a drive signal from the control unit and adjusts the motor pump speed and the solenoid valve opening according to the drive signal to perform a fuel replenishment operation. The status monitoring unit collects the operating current of the motor pump and the coil temperature of the solenoid valve in the actuator in real time as operating parameters, and sends the operating parameters to the communication interface. The control unit also receives control commands from the communication interface issued by the multi-scenario centralized energy management system, and adjusts the generation logic of the drive signal according to the control commands.

Images