Energy allocation optimization control device and system adaptive to regional endowment

Abstract

The invention relates to the technical field of gas station network energy intelligent allocation optimization control, in particular to an energy allocation optimization control device and system adaptive to regional endowment, and the device comprises a data acquisition module which obtains real-time data from a distributed gas station and forms an original data flow; the data processing module performs data cleaning and standardization to generate a structured data set; the natural endowment analysis module extracts energy natural endowment characteristics and generates a regional natural endowment portrait; the allocation algorithm module dynamically calculates an energy allocation strategy and generates an allocation instruction; and the control execution module converts the instruction into a control signal and feeds back an execution result. According to the invention, through a modular design and a closed-loop control mechanism, an adaptive communication protocol, multiple simulation optimization and a machine learning algorithm are adopted, the problem of difficulty in real-time optimization control of energy allocation caused by regional communication infrastructure difference and unbalanced energy endowment is effectively solved, and the energy utilization efficiency and the management effect are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent energy allocation and optimization control technology for gas station networks, and particularly to an energy allocation and optimization control device and system adapted to the regional characteristics. Background Technology

[0002] In the application scenario of intelligent energy allocation and optimization control technology for gas station networks, adapting to regional endowment refers to dynamically adjusting based on the inherent characteristics of different geographical locations, such as energy resource distribution, climate conditions, and energy consumption habits. Energy allocation involves the rational scheduling and distribution of energy such as electricity or fuel required by gas stations, while optimization control adjusts the allocation process through integrated computing models and strategy algorithms, thereby improving energy utilization efficiency and maintaining the stability of network operation.

[0003] Existing intelligent energy allocation and optimization control technologies for gas station networks suffer from the following technical challenges: Firstly, differences in regional communication infrastructure lead to unstable and delayed data transmission. Secondly, uneven energy resources make it difficult for resource allocation algorithms to dynamically adapt to the needs of multiple stations, resulting in difficulties in real-time energy allocation optimization and control. For example, in remote areas with poor communication conditions, intelligent communication boxes rely on 4G or wireless networks to upload tank level and transaction data; however, signal fluctuations can cause data synchronization interruptions, preventing the central platform from obtaining accurate inventory information in real time and affecting fuel allocation decisions. Simultaneously, significant differences in energy types and inventory levels among different gas stations necessitate that the optimization control system coordinate heterogeneous data to achieve balanced allocation; however, uneven resource allocation increases the complexity of real-time scheduling, leading to response delays and reduced efficiency. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an energy distribution optimization control device and system adapted to regional endowments. This invention solves the technical problem that real-time optimization control of energy distribution is difficult due to differences in regional communication infrastructure and uneven energy endowments.

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

[0006] In a first aspect, the energy allocation optimization control device adapted to regional endowments provided by the present invention includes: The data acquisition module is used to acquire real-time data from the smart communication box and sensor devices of the distributed gas station. The real-time data includes oil tank level, refueling transaction records, energy inventory level, site geographical location information, environmental parameters and equipment operating status, and sends the real-time data as a raw data stream to the data processing module. The data processing module is used to receive the raw data stream, perform data cleaning, integrity verification and format standardization on the raw data stream, generate a structured dataset, and send the structured dataset to the endowment analysis module. The endowment analysis module is used to receive the structured dataset, extract regional energy endowment features from the structured dataset, use machine learning clustering algorithms to divide gas station sites into regions, generate regional endowment profiles, and send the regional endowment profiles to the allocation algorithm module. The allocation algorithm module is used to receive the regional endowment profile and the real-time data output by the data processing module, dynamically calculate the energy allocation strategy using an optimization model, generate energy allocation instructions, and send the energy allocation instructions to the control execution module. The control execution module is used to receive the energy allocation instruction, convert the energy allocation instruction into a control signal and send it to the gas station field equipment, and collect the instruction execution results and feed them back to the data acquisition module.

[0007] Furthermore, in the energy allocation optimization control device adapted to regional endowments described in this invention, the data acquisition module is also used for: Initialize the communication connection with the distributed gas stations, and select RS485 wired communication or LORA wireless communication protocol according to the communication infrastructure conditions; Data request instructions are sent to each gas station at preset time intervals, and a polling scheduling algorithm is used to allocate the sequence of data collection tasks. Receive station response data and initiate a redundant transmission mechanism to retransmit data in response to communication delays or interruptions; The successfully received data is added with a timestamp and a site identifier, temporarily stored in a local cache, and the resulting raw data stream is output to the data processing module.

[0008] The data processing module is also used to generate data quality indicators and send them to the data acquisition module; the data acquisition module is also used to dynamically adjust the sending frequency of the data request command according to the data quality indicators.

[0009] The allocation algorithm module is also used to generate strategy execution effect data and send it to the endowment analysis module; the endowment analysis module is also used to update the clustering model parameters based on the strategy execution effect data.

[0010] Furthermore, in the energy allocation optimization control device adapted to regional endowments described in this invention, the data processing module is also used for: Receive and parse the raw data stream to verify data integrity and compliance with numerical range; For the missing parts in the validated data, multiple simulation optimizations were used to fill them in, including: using the Kriging interpolation algorithm to fill the gaps in the geographically distributed data, using the ARIMA model to predict the missing values ​​in the time series, and using Monte Carlo simulation to assess the uncertainty of the data; The isolated forest algorithm is used to detect and remove outlier data points after the data is filled in. The data after removing outliers is converted into a unified JSON format, and then the fields are standardized and the data types are normalized to generate the structured dataset and output to the endowment analysis module.

[0011] Furthermore, in the energy allocation optimization control device adapted to regional endowments described in this invention, the data processing module is also used for: Calculate the quality assessment metrics for the structured dataset, including data completeness, data accuracy, and data timeliness; The quality assessment indicators are sent to the data acquisition module; The data acquisition module is also used to dynamically adjust the data acquisition frequency based on the received quality assessment indicators: when the data integrity rate or the data accuracy rate is lower than a preset threshold, the frequency of sending data request instructions is increased; when both the data integrity rate and the data accuracy rate are higher than the preset threshold, the frequency of sending data request instructions is decreased.

[0012] Furthermore, in the energy allocation optimization control device adapted to regional endowments described in this invention, the endowment analysis module is also used for: Extract multidimensional feature vectors from the structured dataset. These multidimensional feature vectors include the distribution ratio of energy types, historical consumption pattern trends, climate condition influence coefficients, and energy consumption habit parameters. Principal component analysis algorithm is used to reduce the dimensionality of the multidimensional feature vector, calculate the variance contribution rate of each principal component, and retain the principal components with a cumulative variance contribution rate of more than 85% to form the dimensionality-reduced feature vector. The K-means++ clustering algorithm is used to perform cluster analysis on the dimensionality-reduced feature vectors. Initial cluster centers are selected based on the maximum-minimum distance principle. The Euclidean distance from the stations to the cluster centers is calculated iteratively and the station labels are reassigned until the clustering results are stable. A regional endowment profile is generated based on the stabilized clustering results. The regional endowment profile includes static endowment indicators and dynamic trend predictions, and the energy demand priority weight and elasticity coefficient of each cluster region are calculated.

[0013] Furthermore, in the energy allocation optimization control device adapted to regional endowments described in this invention, the endowment analysis module is also used for: Receive strategy execution effect data sent by the allocation algorithm module, wherein the strategy execution effect data includes the deviation between the actual consumption data and the predicted value in the region; When the deviation continues to exceed the preset threshold, the cluster center parameters are updated using an incremental learning method. Based on the updated cluster center parameters, the distance from the site to the cluster center is recalculated and new site labels are assigned; Based on the new site labels, update the energy demand priority weights and resilience coefficients in the regional endowment profile.

[0014] Furthermore, in the energy allocation optimization control device adapted to regional endowments described in this invention, the allocation algorithm module is further used for: Align the regional endowment profile with the timestamps of the real-time data to establish a data association mapping table; Based on the data association mapping table, a mixed integer linear programming model is constructed. The decision variables are defined as the allocation quantity and the binary authorization flag. The objective function is set as minimizing the weighted sum of energy waste cost and communication delay impact. The constraints include real-time inventory capacity, demand forecast value and communication status weight. After constructing the mixed-integer linear programming model, the Q-learning reinforcement learning algorithm is used to construct the state-action value function, and the data flow weights of priority stations in the model are dynamically adjusted according to historical reward signals. Using the weighted mixed-integer linear programming model, different allocation strategies are pre-run in a digital twin simulation environment to evaluate key performance indicators and select the Pareto optimal solution set. The selected Pareto optimal solution is encoded into a control command sequence, which includes the quantity of oil to be allocated, refueling authorization mode parameters, and price adjustment coefficient.

[0015] Furthermore, in the energy allocation optimization control device adapted to regional endowments described in this invention, the allocation algorithm module is further used for: A physical model of the gas station network is established for the digital twin simulation environment, and the real-time data and the regional endowment profile are imported to drive the simulation. In the simulation environment after the driver is activated, multiple candidate allocation strategies are executed, and system behavior data under each strategy is monitored and recorded. Based on the monitored system behavior data, calculate the energy utilization efficiency, response time and inventory balance index corresponding to each candidate allocation strategy; Based on the calculated indicators, a multi-objective optimization algorithm is used to select the Pareto optimal solution set from all candidate strategies.

[0016] Furthermore, in the energy allocation optimization control device adapted to regional endowments described in this invention, the control execution module is further used for: Verify the digital signature and parameter range of the energy allocation instruction; The verified energy distribution command is parsed, the command opcode is mapped to the corresponding device driver function, and a control signal is generated. The control signal is used to adjust the fuel dispenser authorization mode, set a preset fuel amount threshold, or trigger the fuel tank alarm rule. After generating the control signal, a fuzzy logic controller is used to handle the uncertainty in instruction execution, and smooth control is achieved through a fuzzy rule base. After the control signal is issued, the command execution status of the equipment at the gas station is monitored, and the actual refueling volume, equipment operating status and alarm information are collected as the execution results. The execution result, along with the corresponding instruction identifier, execution timestamp, and result status code, is encoded into a feedback data packet and transmitted to the data acquisition module via the uplink channel.

[0017] Secondly, the energy allocation optimization control system adapted to regional endowments provided by the present invention is applied to the energy allocation optimization control device adapted to regional endowments as described above, comprising: The data acquisition module is used to acquire real-time data from the smart communication box and sensor devices of the distributed gas station. The real-time data includes oil tank level, refueling transaction records, energy inventory level, site geographical location information, environmental parameters and equipment operating status, and sends the real-time data as a raw data stream to the data processing module. The data processing module is used to receive the raw data stream, perform data cleaning, integrity verification and format standardization on the raw data stream, generate a structured dataset, and send the structured dataset to the endowment analysis module. The endowment analysis module is used to receive the structured dataset, extract regional energy endowment features from the structured dataset, use machine learning clustering algorithms to divide gas station sites into regions, generate regional endowment profiles, and send the regional endowment profiles to the allocation algorithm module. The allocation algorithm module is used to receive the regional endowment profile and the real-time data output by the data processing module, dynamically calculate the energy allocation strategy using an optimization model, generate energy allocation instructions, and send the energy allocation instructions to the control execution module. The control execution module is used to receive the energy allocation command, convert the energy allocation command into a control signal and send it to the gas station field equipment, and collect the command execution results and feed them back to the data acquisition module.

[0018] Beneficial effects of this invention: This invention achieves dynamic optimization of energy allocation through modular design. The data acquisition module adaptively selects communication protocols to effectively address regional differences in communication infrastructure. The data processing module employs multiple simulation optimization techniques to improve data quality. The endowment analysis module generates regional endowment profiles based on machine learning clustering algorithms to accurately reflect energy distribution characteristics. The allocation algorithm module integrates optimization models and reinforcement learning to dynamically calculate allocation strategies. The control execution module continuously adjusts operation commands through a closed-loop feedback mechanism. The entire system enhances its adaptive capabilities through bidirectional interaction and collaboration between modules, thereby solving the problem of real-time optimization and control of energy allocation caused by regional communication differences and uneven energy endowment. Attached Figure Description

[0019] 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.

[0020] Figure 1 This is a system architecture diagram of the energy allocation optimization control system adapted to the regional endowment of the present invention. Detailed Implementation

[0021] 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.

[0022] In a first aspect, the energy allocation optimization control device adapted to regional endowments provided by the present invention includes: The data acquisition module is used to acquire real-time data from the smart communication box and sensor devices of the distributed gas station. The real-time data includes oil tank level, refueling transaction records, energy inventory level, site geographical location information, environmental parameters and equipment operating status, and sends the real-time data as a raw data stream to the data processing module. The data processing module is used to receive the raw data stream, perform data cleaning, integrity verification and format standardization on the raw data stream, generate a structured dataset, and send the structured dataset to the endowment analysis module. The endowment analysis module is used to receive the structured dataset, extract regional energy endowment features from the structured dataset, use machine learning clustering algorithms to divide gas station sites into regions, generate regional endowment profiles, and send the regional endowment profiles to the allocation algorithm module. The allocation algorithm module is used to receive the regional endowment profile and the real-time data output by the data processing module, dynamically calculate the energy allocation strategy using an optimization model, generate energy allocation instructions, and send the energy allocation instructions to the control execution module. The control execution module is used to receive the energy allocation instruction, convert the energy allocation instruction into a control signal and send it to the gas station field equipment, and collect the instruction execution results and feed them back to the data acquisition module.

[0023] The energy allocation optimization control device, adapted to the region's specific conditions, achieves dynamic optimization of energy allocation through modular design. The data acquisition module obtains real-time data from the smart communication boxes and sensor devices of the distributed gas stations. This real-time data includes tank level, refueling transaction records, energy inventory levels, station geographical location information, environmental parameters, and equipment operating status. The data acquisition module initializes its communication connection with the gas stations, selecting either RS485 wired communication or LoRa wireless communication protocol based on the communication infrastructure conditions to adapt to different regional network environments. The data acquisition module sends data request commands to each station at preset time intervals, employing a polling scheduling algorithm to allocate the acquisition task sequence and effectively manage the data acquisition priority across multiple stations. When communication delays or interruptions occur, the data acquisition module activates a redundant transmission mechanism for data retransmission, improving data acquisition reliability. Successfully received data is timestamped and includes a station identifier, temporarily stored in a local buffer, and output as a raw data stream to the data processing module. This process ensures that the data acquisition stage can cope with communication fluctuations, providing continuous data input for downstream processing.

[0024] After receiving the raw data stream, the data processing module performs data cleaning, integrity verification, and format standardization. The module parses the data stream, verifies data integrity and numerical range compliance, and identifies missing or outlier parts. To address missing data, multiple simulation optimization techniques are employed for data completion, including using Kriging interpolation to handle geographically distributed data gaps, using the ARIMA model to predict missing time-series values, and evaluating data uncertainty through Monte Carlo simulation. The module also applies the Isolation Forest algorithm to detect and remove outlier data points and eliminate noise interference. The processed data is converted to a unified JSON format, and field standardization and data type normalization are performed to generate a structured dataset, which is then sent to the endowment analysis module. Simultaneously, the data processing module calculates data quality assessment metrics, such as data completeness and accuracy, and feeds these metrics back to the data acquisition module to dynamically adjust the acquisition frequency, forming an initial feedback loop to optimize data stream quality.

[0025] The endowment analysis module extracts regional energy endowment features from a structured dataset. The feature vector includes the distribution ratio of energy types, historical consumption pattern trends, climate condition influence coefficients, and energy consumption habit parameters. The module uses principal component analysis (PCA) to reduce the dimensionality of the feature vector, retaining principal components whose cumulative variance contribution exceeds a threshold to form the reduced-dimensional feature vector. Subsequently, a K-means enhanced clustering algorithm is applied to divide the sites into regions. Initial cluster centers are selected based on the maximum-minimum distance principle, and site labels are assigned iteratively by calculating Euclidean distance until the clustering results stabilize. The endowment analysis module generates a regional endowment profile, including static endowment indicators and dynamic trend predictions, and calculates the energy demand priority weight and elasticity coefficient for each cluster region. The endowment analysis module receives strategy execution performance data from the allocation algorithm module. When the deviation between actual consumption and prediction continues to exceed limits, incremental learning is used to update the clustering model parameters, achieving adaptive optimization of regional division.

[0026] The allocation algorithm module integrates regional endowment profiles and real-time data to dynamically calculate energy allocation strategies. It aligns data timestamps and establishes a data association mapping table to ensure consistency across multiple data sources. The module constructs a mixed-integer linear programming model, defining decision variables such as allocation quantity and binary authorization flags. It sets an objective function that minimizes the weighted sum of energy waste costs and communication delay impacts, and sets constraints including real-time inventory capacity, demand forecasts, and communication status weights. The module employs Q-learning reinforcement learning to construct a state-action value function, dynamically adjusting the data flow weights of priority stations based on historical reward signals to enhance model robustness. The module pre-simulates different allocation strategies in a digital twin simulation environment, evaluating energy utilization efficiency, response time, and inventory balance indicators. It selects the Pareto optimal solution set and encodes it into a control command sequence, including oil allocation quantities, refueling authorization mode parameters, and price adjustment coefficients.

[0027] The control execution module receives energy allocation instructions, verifies the digital signature and parameter range, and then parses the instruction content. It maps the instruction opcode to the device driver function, generating control signals to adjust the fuel dispenser authorization mode, set preset refueling thresholds, or trigger tank alarm rules. The control execution module uses a fuzzy logic controller to handle uncertainties in instruction execution, achieving smooth control and improving operational stability through a fuzzy rule base. It monitors the instruction execution status of the gas station's on-site equipment, collecting actual refueling volume, equipment operating status, and alarm information as execution results. The execution results, along with the instruction identifier, execution timestamp, and result status code, are encoded into a feedback data packet and transmitted to the data acquisition module via the uplink channel, forming a closed-loop feedback mechanism. The entire data flow, from acquisition to execution and feedback, with modules interacting bidirectionally, enhances the system's adaptive capabilities, effectively addressing control challenges arising from regional communication differences and uneven energy distribution.

[0028] Secondly, please refer to Figure 1 The energy allocation optimization control system adapted to regional endowments provided by the present invention is applied to the energy allocation optimization control device adapted to regional endowments as described above, comprising: The data acquisition module is used to acquire real-time data from the smart communication box and sensor devices of the distributed gas station. The real-time data includes oil tank level, refueling transaction records, energy inventory level, site geographical location information, environmental parameters and equipment operating status, and sends the real-time data as a raw data stream to the data processing module. The data processing module is used to receive the raw data stream, perform data cleaning, integrity verification and format standardization on the raw data stream, generate a structured dataset, and send the structured dataset to the endowment analysis module. The endowment analysis module is used to receive the structured dataset, extract regional energy endowment features from the structured dataset, use machine learning clustering algorithms to divide gas station sites into regions, generate regional endowment profiles, and send the regional endowment profiles to the allocation algorithm module. The allocation algorithm module is used to receive the regional endowment profile and the real-time data output by the data processing module, dynamically calculate the energy allocation strategy using an optimization model, generate energy allocation instructions, and send the energy allocation instructions to the control execution module. The control execution module is used to receive the energy allocation command, convert the energy allocation command into a control signal and send it to the gas station field equipment, and collect the command execution results and feed them back to the data acquisition module.

[0029] When initializing communication connections with distributed gas stations, the data acquisition module automatically identifies available communication methods by scanning the network interface status of each station. When a stable wired network is detected, the RS485 protocol is prioritized to establish a physical layer connection. For wireless coverage areas, the module switches to LoRa for low-power wide-area transmission. This adaptive selection mechanism effectively matches differences in regional communication infrastructure. A polling scheduling algorithm assigns geographically partitioned numbers to stations and calculates the optimal acquisition sequence to minimize overall latency. The system sends data request commands in a fixed-time-slice loop, including station codes and data type identifiers, facilitating accurate response from target devices. In case of communication delays or interruptions, a redundant transmission mechanism triggers an exponential backoff retransmission strategy. The initial retry interval is set relatively short, gradually increasing after consecutive failures to avoid network congestion. Successfully received data packets are appended with a high-precision timestamp and a unique station identifier and temporarily stored in a local buffer with a circular buffer structure. The buffer uses a first-in, first-out (FIFO) strategy to manage data overflow risks, forming a continuous stream of raw data that is pushed to downstream modules.

[0030] The data quality indicators generated by the data processing module include data integrity rate, accuracy rate, and timeliness quantification values. These indicators are transmitted to the data acquisition module via a dedicated feedback channel. After parsing the quality indicators, the data acquisition module dynamically adjusts the acquisition frequency. For example, when the data integrity rate is below a threshold, the polling interval is shortened to increase the acquisition density; conversely, when the indicators are above the threshold, the interval is extended to reduce system load, forming a primary feedback loop. The allocation algorithm module calculates the strategy execution effect data, including statistical data on the deviation between actual regional energy consumption and predicted values. After receiving this data, the endowment analysis module initiates an incremental model learning process. When the deviation continues to exceed the tolerance, the cluster center coordinates are recalculated, and the site classification labels are iteratively updated, thereby optimizing the accuracy of the regional endowment profile.

[0031] When the data processing module parses the raw data stream, it first verifies the compliance of the data packet checksum and field length. Failed data is logged and retransmitted. Verified data enters a multi-stage simulation optimization process. For spatial dimension missing values, a variogram model is constructed based on the geographical coordinates of neighboring sites using the Kriging interpolation algorithm. Time series gaps are filled by analyzing historical trends using the ARIMA model. Monte Carlo simulation generates multiple probabilistic scenarios to assess the range of data uncertainty. In the anomaly detection phase, an isolated forest algorithm is used to construct a decision tree cluster, identifying and removing data points that deviate from the mainstream distribution. The processed data is converted to a standardized JSON format, with field naming following camelCase rules. Numerical data is normalized to a unified dimension to ensure the structured dataset is compatible with downstream analysis requirements.

[0032] When calculating quality assessment metrics for structured datasets, the data processing module calculates the data integrity rate by measuring the proportion of missing fields, the data accuracy rate by verifying the compliance of numerical boundaries, and the data timeliness by measuring the latency from data collection to processing. These metrics are sent asynchronously to the data acquisition module via a message queue. The data acquisition module has built-in frequency adjustment logic: when any metric falls below a preset threshold, the frequency of data request commands is immediately increased; conversely, when all metrics are above the threshold, the frequency is decreased, thus achieving dynamic allocation of acquisition resources.

[0033] The endowment analysis module extracts multidimensional feature vectors from the structured dataset, including the distribution ratio of energy types, historical consumption pattern trends, climate condition influence coefficients, and energy consumption habit parameters. Principal component analysis (PCA) calculates the feature covariance matrix, retaining principal components whose cumulative variance contribution exceeds a threshold to form dimensionality-reduced feature vectors, eliminating redundant information between features. The K-means++ clustering algorithm initializes cluster centers using the maximum-minimum distance principle, avoiding local optima caused by random initial values. It iteratively calculates the Euclidean distance between stations and redistributes labels until the intra-cluster variance stabilizes. Finally, it generates a regional endowment profile including static endowment indicators and dynamic trend predictions, and calculates the energy demand priority weights and elasticity coefficients for each cluster region.

[0034] After receiving the strategy execution effect data from the allocation algorithm module, the endowment analysis module compares the deviation between the actual regional consumption and the predicted values. When the deviation continues to exceed a threshold, an incremental learning mechanism is triggered, using gradient descent to fine-tune the cluster center parameters, recalculate the distance from the stations to the new centers, and assign updated labels. Based on the new labels, the energy demand priority weights and elasticity coefficients in the regional endowment profile are dynamically adjusted, enabling the profile to continuously adapt to changes in energy consumption patterns.

[0035] The allocation algorithm module aligns regional endowment profiles with real-time data timestamps, establishing a data association mapping table with station codes as keys. A mixed-integer linear programming model defines allocation quantity as a continuous decision variable, binary authorization flags represent station activation status, and the objective function weights and combines energy waste costs and communication delay impact factors. Constraints embed real-time inventory capacity upper and lower limits and demand forecasts. A Q-learning reinforcement learning algorithm constructs a state space encompassing network load levels and inventory urgency, an action space corresponding to data flow weight adjustment strategies, and a reward function dynamically updates weight parameters based on the improvement of the objective function after historical strategy execution. A digital twin simulation environment constructs a gas station network physical model, imports real-time data to drive multi-round strategy pre-playing, evaluates energy utilization efficiency, response time, and inventory balance indicators, and uses a multi-objective optimization algorithm to select the Pareto optimal solution set, ultimately encoding a control command sequence including oil allocation quantity, refueling authorization mode parameters, and price adjustment coefficients.

[0036] After verifying the digital signature and parameter range of the energy distribution command, the control execution module parses the command opcode and maps it to the device driver function library. In the control signal generation stage, the fuel dispenser authorization mode setting function is called to adjust the preset refueling threshold or trigger the tank alarm rule. The fuzzy logic controller defines input variables such as the fuzzy set of device response status and output variables such as the fuzzy set of signal strength, achieving smooth control transitions through the fuzzy rule library. After the command is issued, changes in the device status register are monitored, and the actual refueling volume, operating status, and alarm information are collected as execution results. The feedback data packet encapsulates the original command identifier, execution timestamp, and result status code, and is transmitted back to the data acquisition module via the uplink channel to complete closed-loop control.

[0037] When the allocation algorithm module performs timestamp alignment, it extracts timestamps from regional endowment profile data and real-time data, and uses a global clock synchronization protocol to coordinate the time bases of different data sources. The timestamp alignment process involves parsing the time encoding in the packet header, converting Unix timestamps to a unified format, and calculating time offsets for compensation. The data association mapping table is constructed based on the site code as the primary key, with association fields including data source identifier, timestamp sequence, and data type classification. The mapping table uses a hash index structure to optimize query efficiency and supports fast matching of multi-source data items.

[0038] In the mixed-integer linear programming model construction phase, the decision variables are defined to include continuous allocation variables and binary authorization flag variables. The objective function is designed as a linear weighted combination, with weight coefficients dynamically adjusted based on energy waste cost factors and communication delay impact factors. Constraints include upper and lower bound constraints on real-time inventory capacity, equality constraints on demand forecasts, and inequality constraints on communication status weights. The model solver utilizes a mathematical programming solver, employing a branch and bound algorithm to process integer variables and obtain an initial optimal solution.

[0039] Q-learning reinforcement learning algorithms are integrated into the optimization model. The state space is defined as a discretized representation of network load level and inventory urgency. The action space corresponds to the data flow weight adjustment strategy, including increasing weights, decreasing weights, or maintaining the current weights. The reward function is calculated based on the improvement of the objective function after historical policy executions, and the reward signal is used to update the state-action value function table. The Q-learning algorithm employs an ε-greedy strategy to balance exploration and exploitation, progressively optimizing the weight allocation scheme.

[0040] When rehearsing the allocation strategy in a digital twin simulation environment, a virtual mapping model of the gas station network is established. Model elements include tank capacity, fuel dispenser flow characteristics, and pipeline transmission delay. Real-time data is imported into the simulation environment to drive the model's execution, simulating energy flow processes under different allocation strategies. Key performance indicators (KPIs) are evaluated, including energy efficiency ratio, system response time delay, and inventory balance variance. Simulation results are recorded in a performance log for subsequent analysis.

[0041] The Pareto optimal solution selection process employs a multi-objective optimization algorithm, with a non-dominated sorting algorithm identifying non-dominated solutions within the solution set. Solution set selection considers the uniformity of the objective function value distribution, and crowding degree calculation is used to maintain solution set diversity. The final selected Pareto optimal solution is encoded as a control command sequence. The command sequence format uses the Protocol Buffers serialization protocol, and the field definitions include floating-point values ​​for the oil allocation quantity, the enumerated type of the refueling authorization mode parameters, and a percentage representation of the price adjustment coefficient.

[0042] The physical model of the digital twin simulation environment is built based on the actual topology of the gas station network. Model parameters include tank geometry, pipe diameter, and pumping power. The physical model is discretized using the finite element method to simulate the fluid dynamics of energy flow. After importing real-time data and regional endowment profiles, the simulation environment initializes state variables and sets boundary conditions to reflect the actual operating scenario. The simulation process is driven by a time-stepping algorithm, updating the model state and recording system behavior at each time step.

[0043] When candidate rationing strategies are executed in a simulation environment, strategy parameters include ration allocation schemes, authorization mode settings, and price adjustment ranges. System behavior data monitoring is achieved through virtual sensors, collecting indicators such as energy flow rate, peak equipment utilization, and alarm trigger counts. Monitoring data is stored in a circular buffer, supporting real-time querying and historical backtracking.

[0044] In the performance index calculation phase, the energy efficiency index is derived by the ratio of actual consumption to the theoretical minimum consumption. The response time index measures the delay from command issuance to equipment response. The inventory balance index calculates the standard deviation of inventory levels at each site from the target value. The index calculations employ a sliding window statistical method to reflect trends over time.

[0045] When selecting a Pareto optimal solution set using a multi-objective optimization algorithm, the algorithm flow includes population initialization, fitness evaluation, selection operation, and crossover / mutation. Fitness evaluation is based on weighted sum or constraint methods to handle multi-objective problems. The selection operation employs a tournament selection mechanism, retaining superior individuals for the next iteration. The final Pareto optimal solution set undergoes redundancy removal to ensure its simplicity and effectiveness.

[0046] When the control execution module verifies the energy allocation command, the digital signature verification uses an asymmetric encryption algorithm to compare the command signature with the registered signature in the public key store. Parameter range checks involve upper and lower bound verification of numerical parameters and validity verification of enumerated parameters. Verification failure triggers an error handling process, logs the security information, and requests the regeneration of the command.

[0047] The instruction parsing process maps opcodes to device driver functions. The driver function library includes functions for setting fuel dispenser authorization, configuring preset thresholds, and triggering alarm rules. During the control signal generation phase, the signal format follows industrial communication protocol standards, such as the Modbus TCP frame structure. The signal content includes the target device address, opcode, and parameter list.

[0048] When handling uncertainty, the fuzzy logic controller fuzzifies the input variables, including the fuzzy sets of device status and environmental influences. The output variable is the fuzzy set of control signal strength. The fuzzy rule base defines conditional statements, such as adjusting the control strength to medium when the device status is warning. The defuzzification process uses the centroid method to calculate the precise control quantity.

[0049] Command execution status monitoring is achieved by polling the device status register, with the monitoring frequency dynamically adjusted according to the device type. Execution results are collected including pulse counts of the actual refueling volume, bitmask readings of the device operating status, and string parsing of alarm information. Timestamps and source device identifiers are added to the collected data.

[0050] The feedback data packet encoding adopts a structured data format, with fields including the UUID encoding of the instruction identifier, the ISO format of the execution timestamp, and an integer enumeration of the result status code. Uplink transmission uses an asynchronous communication mode; the data packets are compressed and transmitted to the data acquisition module via the HTTPS protocol. This feedback mechanism forms a closed-loop control, supporting continuous system optimization.

[0051] After receiving the raw data stream, the data processing module performs data cleaning, integrity verification, and format standardization. The module parses the data stream, verifies data integrity and numerical range compliance, and identifies missing or abnormal portions. To address missing data, multiple simulation optimization techniques are employed for data filling, including the application of the Kriging interpolation algorithm to handle geographically distributed data gaps. The Kriging interpolation algorithm is based on a semi-variogram model, and its formula is as follows: ; in, Point to be estimated The predicted value, It is a spatial location coordinate vector (such as latitude and longitude), representing the point to be estimated; It is the summation index, indicating the index of the summation index. There are 1 known points, with values ​​ranging from 1 to 1. ; It is the number of known points; These are weighting coefficients, calculated using the Kriging equations. It is a known point The observed values, It is a spatial position coordinate vector, representing the first... Known points. Weighting coefficients. It satisfies the unbiasedness and minimum variance constraints. The semi-variogram is used to calculate the weights, as shown in the following formula: ; in, It is distance The semi-variance value at the location; It is a spatial distance vector; The distance is The number of point pairs; It is a point Observed values; It is a point Offset distance Subsequent observations typically represent another point. Weighting coefficients are determined by solving the Kriging equations. Simultaneously, the ARIMA model is used to predict missing values ​​in the time series. The ARIMA model is represented as ARIMA(p,d,q), where p is the autoregressive order, d is the differencing order, and q is the moving average order. The formula is as follows: ; in, It is a time series in time The value; It is a backoff operator, defined as That is, to go back one time unit; It is the autoregressive order; It is the summation index, indicating the index of the summation index. One lagging term; It is the autoregressive coefficient; It is the difference order; It is the order of the moving average; It is the summation index, indicating the index of the summation index. One moving average item; It is the moving average coefficient; The white noise error term in time The model parameters are fitted to historical data using maximum likelihood estimation or least squares method. Furthermore, data uncertainty is assessed through Monte Carlo simulation, and statistical distributions are calculated using multiple random scenarios. In the anomaly detection phase, the isolated forest algorithm is applied to construct isolated trees and calculate anomaly scores. ; in, It is a sample Abnormal scores; It is an input data sample; It is the sample size; It is a sample The expected path length in an isolated tree; Is for The average path length of each sample is used for standardization. Scores close to 1 indicate anomalies. After processing, the data is converted to a uniform JSON format for field standardization and data type normalization.

[0052] The endowment analysis module extracts regional energy endowment features based on structured datasets and uses principal component analysis (PCA) to reduce the dimensionality of multidimensional feature vectors. PCA is implemented by calculating the eigenvalues ​​and eigenvectors of the covariance matrix. ; in, It is the covariance matrix; It is the sample size; It is a standardized feature matrix, with rows representing samples and columns representing features; yes The transpose of .

[0053] The eigenvalue decomposition formula is ; in It is the eigenvector matrix; It is an eigenvalue diagonal matrix; yes The transpose of . Principal components with a cumulative variance contribution rate exceeding 85% are retained, and the eigenvectors after dimensionality reduction are calculated as follows: ; in, It is the feature matrix after dimensionality reduction; It was before A matrix composed of eigenvectors This represents the number of principal components retained. Subsequently, the K-means++ clustering algorithm is used to perform cluster analysis on the dimensionality-reduced feature vectors. The K-means++ algorithm selects initial cluster centers based on the maximum-minimum distance principle and iteratively optimizes the sum of squares within each cluster: ; in, It is the cluster sum of squares objective function; It is the number of clusters; It is a clustering index; It is the first A sample set of each cluster; It is an eigenvector; It is the first The center vector of each cluster; It is the square of the Euclidean distance. The iterative process reassigns site labels until the clustering results stabilize.

[0054] After aligning the regional endowment profile with the timestamps of real-time data, the allocation algorithm module constructs a mixed-integer linear programming model to dynamically calculate the energy allocation strategy. The model defines decision variables including allocation quantity (continuous variable) and binary authorization flag (integer variable), and the objective function is to minimize the weighted cost. ; in, This represents minimizing the objective function; It is the summation index, indicating the index of the summation index. One continuous variable; It is the first Energy waste cost coefficients for several continuous variables; It is the first One continuous decision variable (allocation quantity); It is the summation index, indicating the index of the summation index. One integer variable; It is the first The communication delay impact coefficient of an integer variable; It is the first There are 1 binary decision variable (authorization flag). Constraints include real-time inventory capacity, demand forecast, and communication status weights, expressed as: ; in, It is a constraint matrix of continuous variables; It is a continuous variable vector; It is a constraint matrix of integer variables; It is a vector of integer variables; It is the right-hand vector of the constraint condition; The condition "less than or equal to" indicates that the constraint is met. After the model is solved, the data flow weights are dynamically adjusted using the Q-learning reinforcement learning algorithm. Q-learning updates the state-action value function: ; in, It is a state Next action Q value; It is a status defined as network load level and inventory urgency level; This refers to the action, which corresponds to the data stream weight adjustment strategy; This indicates an assignment operation; It is the learning rate, which controls the update step size; It is a reward signal, based on the degree of improvement of the objective function after policy execution; It is a discount factor that weighs current and future rewards; Indicates the next state All possible actions Take the maximum Q value; The next state Possible actions below; The next state Next action Q value; This is the next state. The strategy is rehearsed in a digital twin simulation environment to assess energy efficiency, response time, and inventory balance.

[0055] The energy allocation optimization control device, adapted to regional conditions, achieves dynamic optimization of energy allocation across a gas station network through modular design, addressing control difficulties caused by differences in regional communication infrastructure and uneven energy distribution. The data acquisition module obtains real-time data from smart communication boxes and sensor devices at distributed gas stations. This real-time data includes tank level, refueling transaction records, energy inventory levels, station geographical location information, environmental parameters, and equipment operating status. The data acquisition module initializes its communication connection with the gas station, automatically selecting either RS485 wired communication or LoRa wireless communication protocol based on communication infrastructure conditions, adapting to varying network environments in different regions. The system employs a polling scheduling algorithm to allocate the acquisition task sequence, sending data request commands to each station at preset time intervals. When communication delays or interruptions occur, a redundant transmission mechanism initiates a data retransmission strategy. Successfully received data is timestamped and includes a station identifier, temporarily stored in a local buffer, and then output as a raw data stream to the data processing module. This design effectively handles communication fluctuations, providing continuous data input for downstream processing.

[0056] After receiving the raw data stream, the data processing module performs data cleaning, integrity verification, and format standardization. The data cleaning phase verifies data integrity and numerical range compliance, identifying missing or outlier data. For missing data, multiple simulation optimization techniques are used to fill in gaps, including applying Kriging interpolation to handle geographically distributed data gaps, using the ARIMA model to predict missing time-series values, and evaluating the range of data uncertainty through Monte Carlo simulation. In the anomaly detection phase, the Isolation Forest algorithm is used to identify and remove noisy data points. The processed data is converted to a unified JSON format, and field standardization and data type normalization are performed to generate a structured dataset, which is then sent to the endowment analysis module. Simultaneously, the data processing module calculates data quality assessment metrics, such as data completeness and accuracy, and feeds this data back to the data acquisition module to dynamically adjust the acquisition frequency, forming an initial feedback loop to optimize data stream quality.

[0057] The endowment analysis module extracts regional energy endowment features based on a structured dataset. The feature vector includes the distribution ratio of energy types, historical consumption pattern trends, climate condition influence coefficients, and energy consumption habit parameters. The module uses principal component analysis (PCA) to reduce the dimensionality of the feature vector, retaining principal components with a cumulative variance contribution rate exceeding a threshold to form the dimensionality-reduced feature vector, thus eliminating redundant information between features. Subsequently, the K-means++ clustering algorithm is applied to divide the gas station sites into regions. Initial cluster centers are selected based on the maximum-minimum distance principle, and site labels are assigned iteratively by calculating Euclidean distance until the clustering results stabilize. The endowment analysis module generates a regional endowment profile, including static endowment indicators and dynamic trend predictions, and calculates the energy demand priority weight and elasticity coefficient for each cluster region. The endowment analysis module receives strategy execution effect data from the allocation algorithm module. When the deviation between the actual regional consumption data and the predicted values ​​continuously exceeds a preset threshold, incremental learning is used to update the clustering model parameters, recalculate the site classification labels, and optimize the accuracy of the regional endowment profile.

[0058] The allocation algorithm module integrates regional endowment profiles and real-time data to dynamically calculate energy allocation strategies. It aligns data timestamps and establishes a data association mapping table to ensure consistency across multiple data sources. The module constructs a mixed-integer linear programming model, defining decision variables including allocation quantity and binary authorization flags. The objective function is to minimize the weighted sum of energy waste costs and communication delay impacts, with constraints including real-time inventory capacity, demand forecasts, and communication state weights. The module employs Q-learning reinforcement learning to construct a state-action value function, dynamically adjusting data flow weights for priority stations based on historical reward signals to enhance model adaptability. The module pre-simulates different allocation strategies in a digital twin simulation environment, evaluating energy utilization efficiency, response time, and inventory balance indicators. It selects the Pareto optimal solution set and encodes it into a control command sequence, including oil allocation quantities, refueling authorization mode parameters, and price adjustment coefficients.

[0059] The control execution module receives energy allocation instructions, verifies the digital signature and parameter range, and then parses the instruction content. It maps the instruction opcode to the device driver function, generating control signals to adjust the fuel dispenser authorization mode, set preset refueling thresholds, or trigger tank alarm rules. The control execution module uses a fuzzy logic controller to handle uncertainties in instruction execution, achieving smooth control and improving operational stability through a fuzzy rule base. It monitors the instruction execution status of the gas station's on-site equipment, collecting actual refueling volume, equipment operating status, and alarm information as execution results. The execution results, along with the instruction identifier, execution timestamp, and result status code, are encoded into a feedback data packet and transmitted to the data acquisition module via the uplink channel, forming a closed-loop feedback mechanism. The entire data flow, from acquisition to execution and feedback, with modules interacting bidirectionally, enhances the system's adaptive capabilities, effectively addressing control challenges arising from regional communication differences and uneven energy distribution.

[0060] Embodiment 1 of the present invention: Optimized control of energy distribution in gas station networks in remote areas; When implementing this solution in gas station networks in remote areas with poor communication infrastructure, the data acquisition module obtains real-time operational data from each station via a smart communication box. Due to the unstable 4G signal in remote areas, the data acquisition module automatically switches to the LoRa wireless communication protocol to establish a connection, employing an adaptive polling mechanism to collect data by geographical region. When a station times out, the system immediately activates a redundant transmission mechanism, prioritizing the retransmission of that station's data in the next acquisition cycle, and temporarily storing successfully acquired data in a local cache after adding a high-precision timestamp.

[0061] The data processing module performs multiple checks on the collected raw data stream and discovers missing tank level data at some sites. The system uses the Kriging interpolation algorithm, based on the geographical coordinate distribution characteristics of neighboring sites, to generate estimated tank level values ​​for the missing points. Simultaneously, the ARIMA model is applied to compensate for time-series data, predicting missing values ​​caused by communication interruptions. After anomaly detection and format standardization, a structured dataset including complete tank levels, transaction records, and inventory levels is generated.

[0062] The endowment analysis module extracts regional features from the processed data, including the distribution ratio of diesel and gasoline, historical consumption fluctuation patterns, and the influence coefficient of diurnal temperature range. After dimensionality reduction through principal component analysis, the K-means++ algorithm is used to divide the 17 stations into three endowment regions: a high-altitude mountainous station group, a highway trunk line station group, and a rural low-traffic station group. A differentiated endowment profile is generated for each region, with the mountainous station group marked as a high-priority energy supply area and its elasticity coefficient set to 0.7.

[0063] The rationing algorithm module combines real-time data and endowment profiles to construct a mixed-integer programming model. For mountain stations with unstable communication, the algorithm automatically increases the weight of data collection and rehearses multiple rationing schemes in a digital twin environment. The final optimized instruction increases the diesel ration at mountain stations by 15% while decreasing the gasoline price coefficient by 0.2. The control execution module translates the instruction into specific control signals and sends them to the fuel dispensers at each station via the LoRa network to adjust the authorization mode and price parameters. The execution results are fed back to the data acquisition end through the same channel, forming a complete closed loop.

[0064] Embodiment 2 of the present invention: Multi-energy coordinated distribution control of gas station networks in urban agglomerations; When implementing this solution in urban clusters with significant differences in energy endowments, the system faces the challenge of coordinating the allocation of diverse energy sources. The data acquisition module establishes a high-speed connection with gas stations in various urban areas via a wired network, collecting multi-source information every 5 minutes, including charging pile status, hydrogen energy inventory, and current fuel data. The data processing module uses Monte Carlo simulation to assess data uncertainty and performs probabilistic completion on the charging pile usage data.

[0065] The endowment analysis module identifies three types of endowment characteristics: central business districts, residential areas, and transportation hubs. Central business districts exhibit a dual-peak pattern of midday charging peaks and evening fuel demand; residential areas show concentrated energy demand during commuting hours; and transportation hubs require balanced energy distribution 24 hours a day. The system dynamically adjusts cluster centers through incremental learning, promptly incorporating newly built high-speed rail station areas into hub area management.

[0066] The allocation algorithm module constructs an urban energy flow model within a digital twin environment, simulating multi-energy allocation schemes at different times. The algorithm detects idle hydrogen energy at transportation hubs during the early morning hours and generates instructions to allocate some hydrogen to logistics parks operating at night. The control execution module simultaneously adjusts the energy allocation schemes at eight stations, dynamically allocating 20% ​​of the hydrogen energy inventory at the hubs to the logistics stations, while also optimizing the time-of-use pricing strategy for fast-charging stations. The system monitors the operational status of equipment at each station in real time and dynamically adjusts control parameters to maximize the overall efficiency of multi-energy coordinated allocation.

Claims

1. An energy distribution optimization control device adapted to regional endowments, characterized in that, include: The data acquisition module is used to acquire real-time data from the smart communication box and sensor devices of the distributed gas station. The real-time data includes oil tank level, refueling transaction records, energy inventory level, site geographical location information, environmental parameters and equipment operating status, and sends the real-time data as a raw data stream to the data processing module. The data processing module is used to receive the raw data stream, perform data cleaning, integrity verification and format standardization on the raw data stream, generate a structured dataset, and send the structured dataset to the endowment analysis module. The endowment analysis module is used to receive the structured dataset, extract regional energy endowment features from the structured dataset, use machine learning clustering algorithms to divide gas station sites into regions, generate regional endowment profiles, and send the regional endowment profiles to the allocation algorithm module. The allocation algorithm module is used to receive the regional endowment profile and the real-time data output by the data processing module, dynamically calculate the energy allocation strategy using an optimization model, generate energy allocation instructions, and send the energy allocation instructions to the control execution module. The control execution module is used to receive the energy allocation instruction, convert the energy allocation instruction into a control signal and send it to the gas station field equipment, and collect the instruction execution results and feed them back to the data acquisition module.

2. The energy allocation optimization control device adapted to regional endowments according to claim 1, characterized in that, The data acquisition module is also used for: Initialize the communication connection with the distributed gas stations, and select RS485 wired communication or LORA wireless communication protocol according to the communication infrastructure conditions; Data request instructions are sent to each gas station at preset time intervals, and a polling scheduling algorithm is used to allocate the sequence of data collection tasks. Receive station response data and initiate a redundant transmission mechanism to retransmit data in response to communication delays or interruptions; The successfully received data is added with a timestamp and a site identifier, temporarily stored in a local cache, and the raw data stream is output to the data processing module. The data processing module is also used to generate data quality indicators and send them to the data acquisition module; the data acquisition module is also used to dynamically adjust the sending frequency of the data request command according to the data quality indicators; The allocation algorithm module is also used to generate strategy execution effect data and send it to the endowment analysis module; the endowment analysis module is also used to update the clustering model parameters based on the strategy execution effect data.

3. The energy allocation optimization control device adapted to regional endowments according to claim 2, characterized in that, The data processing module is also used for: Receive and parse the raw data stream to verify data integrity and compliance with numerical range; For the missing parts in the validated data, multiple simulation optimizations were used to fill them in, including: using the Kriging interpolation algorithm to fill the gaps in the geographically distributed data, using the ARIMA model to predict the missing values ​​in the time series, and using Monte Carlo simulation to assess the uncertainty of the data; The isolated forest algorithm is used to detect and remove outlier data points after the data is filled in. The data after removing outliers is converted into a unified JSON format, and then the fields are standardized and the data types are normalized to generate the structured dataset and output to the endowment analysis module.

4. The energy allocation optimization control device adapted to regional endowments according to claim 3, characterized in that, The data processing module is also used for: Calculate the quality assessment metrics for the structured dataset, including data completeness, data accuracy, and data timeliness; The quality assessment indicators are sent to the data acquisition module; The data acquisition module is also used to dynamically adjust the data acquisition frequency based on the received quality assessment indicators: when the data integrity rate or the data accuracy rate is lower than a preset threshold, the frequency of sending data request instructions is increased; when both the data integrity rate and the data accuracy rate are higher than the preset threshold, the frequency of sending data request instructions is decreased.

5. The energy allocation optimization control device adapted to regional endowments according to claim 4, characterized in that, The endowment analysis module is also used for: Extract multidimensional feature vectors from the structured dataset. These multidimensional feature vectors include the distribution ratio of energy types, historical consumption pattern trends, climate condition influence coefficients, and energy consumption habit parameters. Principal component analysis algorithm is used to reduce the dimensionality of the multidimensional feature vector, calculate the variance contribution rate of each principal component, and retain the principal components with a cumulative variance contribution rate of more than 85% to form the dimensionality-reduced feature vector. The K-means++ clustering algorithm is used to perform cluster analysis on the dimensionality-reduced feature vectors. Initial cluster centers are selected based on the maximum-minimum distance principle. The Euclidean distance from the stations to the cluster centers is calculated iteratively and the station labels are reassigned until the clustering results are stable. A regional endowment profile is generated based on the stabilized clustering results. The regional endowment profile includes static endowment indicators and dynamic trend predictions, and the energy demand priority weight and elasticity coefficient of each cluster region are calculated.

6. The energy allocation optimization control device adapted to regional endowments according to claim 5, characterized in that, The endowment analysis module is also used for: Receive strategy execution effect data sent by the allocation algorithm module, wherein the strategy execution effect data includes the deviation between the actual consumption data and the predicted value in the region; When the deviation continues to exceed the preset threshold, the cluster center parameters are updated using an incremental learning method. Based on the updated cluster center parameters, the distance from the site to the cluster center is recalculated and new site labels are assigned; Based on the new site labels, update the energy demand priority weights and resilience coefficients in the regional endowment profile.

7. The energy allocation optimization control device adapted to regional endowments according to claim 6, characterized in that, The allocation algorithm module is also used for: Align the regional endowment profile with the timestamps of the real-time data to establish a data association mapping table; Based on the data association mapping table, a mixed integer linear programming model is constructed. The decision variables are defined as the allocation quantity and the binary authorization flag. The objective function is set as minimizing the weighted sum of energy waste cost and communication delay impact. The constraints include real-time inventory capacity, demand forecast value and communication status weight. After constructing the mixed-integer linear programming model, the Q-learning reinforcement learning algorithm is used to construct the state-action value function, and the data flow weights of priority stations in the model are dynamically adjusted according to historical reward signals. Using the weighted mixed-integer linear programming model, different allocation strategies are pre-run in a digital twin simulation environment to evaluate key performance indicators and select the Pareto optimal solution set. The selected Pareto optimal solution is encoded into a control command sequence, which includes the quantity of oil to be allocated, refueling authorization mode parameters, and price adjustment coefficient.

8. The energy allocation optimization control device adapted to regional endowments according to claim 7, characterized in that, The allocation algorithm module is also used for: A physical model of the gas station network is established for the digital twin simulation environment, and the real-time data and the regional endowment profile are imported to drive the simulation. In the simulation environment after the driver is activated, multiple candidate allocation strategies are executed, and system behavior data under each strategy is monitored and recorded. Based on the monitored system behavior data, calculate the energy utilization efficiency, response time and inventory balance index corresponding to each candidate allocation strategy; Based on the calculated indicators, a multi-objective optimization algorithm is used to select the Pareto optimal solution set from all candidate strategies.

9. The energy allocation optimization control device adapted to regional endowments according to claim 8, characterized in that, The control execution module is also used for: Verify the digital signature and parameter range of the energy allocation instruction; The verified energy distribution command is parsed, the command opcode is mapped to the corresponding device driver function, and a control signal is generated. The control signal is used to adjust the fuel dispenser authorization mode, set a preset fuel amount threshold, or trigger the fuel tank alarm rule. After generating the control signal, a fuzzy logic controller is used to handle the uncertainty in instruction execution, and smooth control is achieved through a fuzzy rule base. After the control signal is issued, the command execution status of the equipment at the gas station is monitored, and the actual refueling volume, equipment operating status and alarm information are collected as the execution results. The execution result, along with the corresponding instruction identifier, execution timestamp, and result status code, is encoded into a feedback data packet and transmitted to the data acquisition module via the uplink channel.

10. An energy allocation optimization control system adapted to regional endowments, applied to the energy allocation optimization control device adapted to regional endowments as described in any one of claims 1 to 9, characterized in that, include: The data acquisition module is used to acquire real-time data from the smart communication box and sensor devices of the distributed gas station. The real-time data includes oil tank level, refueling transaction records, energy inventory level, site geographical location information, environmental parameters and equipment operating status, and sends the real-time data as a raw data stream to the data processing module. The data processing module is used to receive the raw data stream, perform data cleaning, integrity verification and format standardization on the raw data stream, generate a structured dataset, and send the structured dataset to the endowment analysis module. The endowment analysis module is used to receive the structured dataset, extract regional energy endowment features from the structured dataset, use machine learning clustering algorithms to divide gas station sites into regions, generate regional endowment profiles, and send the regional endowment profiles to the allocation algorithm module. The allocation algorithm module is used to receive the regional endowment profile and the real-time data output by the data processing module, dynamically calculate the energy allocation strategy using an optimization model, generate energy allocation instructions, and send the energy allocation instructions to the control execution module. The control execution module is used to receive the energy allocation command, convert the energy allocation command into a control signal and send it to the gas station field equipment, and collect the command execution results and feed them back to the data acquisition module.

Images