A booster station plant data acquisition automatic control energy-saving management system

Abstract

The application relates to the technical field of booster station plant management, and discloses a booster station plant data acquisition self-control energy-saving management system. Through standardized collection and deep feature extraction of full-factor operation data of the booster station plant, an innovative combination of physical information neural networks and a Transformer architecture is adopted to establish a precise prediction system covering load changes, equipment efficiency attenuation and environmental coupling influence. Meanwhile, through data envelopment analysis and life cycle assessment methods, a dynamic evaluation mechanism in the double dimensions of energy efficiency and carbon emission is constructed. Based on the fusion of multi-objective optimization algorithms and model predictive control, a real-time control strategy considering economic operation and low-carbon constraints is formed, so that the dual goals of improving the overall energy efficiency level and the management accuracy of carbon emission of the system are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of booster station plant management, and particularly relates to a booster station plant data acquisition self-control energy-saving management system. BACKGROUND

[0002] The booster station plant self-control management refers to real-time data acquisition, state monitoring and automatic control of main electrical equipment such as transformers, circuit breakers, capacitors and auxiliary systems (such as lighting and ventilation) in the station through integrated sensors, control units and communication networks. The core goal is to realize automatic regulation of grid active / reactive power (AGC / AVC), device safety warning and energy efficiency optimization, so as to ensure grid stability, reduce plant energy consumption and improve operation efficiency.

[0003] The existing booster station self-control system generally lacks energy efficiency prediction and adaptive optimization capabilities based on artificial intelligence. The system relies on static models and preset values for control, and cannot accurately predict future energy efficiency trends according to real-time load, weather, electricity price and other dynamic factors, nor can it autonomously learn and adjust operation strategies (such as transformer switching and auxiliary machine start / stop). This results in energy-saving control remaining at the passive response level, failing to achieve forward-looking energy efficiency maximization and optimal control of operating costs, limiting the intelligent level of the system and the potential of energy-saving exploration. At the same time, the current booster station monitoring system generally lacks comprehensive energy efficiency evaluation and carbon emission management functions for the plant itself.

[0004] The optimization method for realizing AGC and AVC in the power plant booster station monitoring system with publication number CN103401247B specifically discloses receiving dispatching instructions through a communication management machine, and calculating the reactive power increment or decrement of each unit based on system impedance identification and linear interpolation algorithm, to realize optimal allocation of active and reactive power, and improve grid regulation speed and reliability. However, this method relies on fixed calculation formulas and historical operation experience, lacks deep mining and prediction of massive operation data by AI algorithms such as machine learning, and cannot realize adaptive learning and forward-looking optimization control of future energy efficiency trends.

[0005] The booster station monitoring and operation system with publication number CN119891561A specifically discloses collecting data using multi-source heterogeneous sensors, and processing data drift caused by electromagnetic interference through a data anomaly detection and correction module, to restore the accuracy of the data and provide a reliable basis for equipment state evaluation. However, this system focuses on data cleaning and fault warning, and does not introduce an artificial intelligence model to perform energy efficiency analysis and intelligent decision-making on the corrected high-quality data, and cannot realize adaptive adjustment of energy-saving strategies based on prediction. SUMMARY

[0006] The technical problems to be solved by the present application are that in the prior art, the automatic control system of the booster station generally lacks energy efficiency prediction and adaptive optimization capabilities based on artificial intelligence, and the current booster station monitoring system generally lacks evaluation of the comprehensive energy efficiency of the plant and carbon emission management functions.

[0007] In order to achieve the above-mentioned purpose, the following technical scheme is adopted in the present application: a booster station plant data acquisition automatic control energy-saving management system, comprising: a multi-source data acquisition and preprocessing module: collecting heterogeneous data of various intelligent sensors in the booster station plant, preprocessing the original heterogeneous data to form a multi-source feature data set; an AI energy efficiency prediction module: based on the feature data set, using an attention mechanism enhanced long short-term memory network to perform multi-time scale load prediction, realizing device efficiency decay curve and health state early warning through a Transformer architecture, coupling environmental factors and new energy output using a physical information neural network, analyzing the influence mechanism of the main transformer cooling efficiency, load capacity and net load, and outputting load prediction, efficiency warning, coupling coefficient and model parameters; an energy efficiency and carbon emission evaluation module: based on the real-time data and prediction results, establishing an energy efficiency-carbon emission dual-dimensional evaluation system, dynamically calculating the comprehensive energy efficiency index of the plant and identifying the energy efficiency bottleneck, while accounting for carbon footprint and green power emission reduction, generating a comprehensive evaluation report through benchmark comparison and compliance check; an adaptive control optimization module: according to the prediction and evaluation results, constructing an optimization model, rolling solving the control strategy of the transformer tap, capacitor bank, energy storage system and auxiliary machine, and outputting the instruction sequence.

[0008] The technical effects and advantages of the present application are: in the present application, through standardized collection and deep feature extraction of full-factor operation data of the booster station plant, the innovative combination of physical information neural network and Transformer architecture breaks through the prediction limitations of traditional static models under complex working conditions, establishes a precise prediction system covering load changes, device efficiency decay and environmental coupling effects, and through data envelopment analysis and life cycle assessment methods, a dynamic evaluation mechanism of energy efficiency-carbon emission dual-dimension is established, solving the problem of lack of management of plant energy consumption and carbon emissions in traditional systems. Finally, based on the fusion of multi-objective optimization algorithm and model predictive control, a real-time control strategy considering economic operation and low-carbon constraints is formed, thereby achieving the dual goals of improving the overall energy efficiency level and carbon emission management precision of the system. Through the closed-loop control of forward-looking prediction and dynamic optimization, the plant operation cost and carbon emission intensity are significantly reduced under the premise of ensuring the stable operation of the power grid, realizing the synergistic optimization of energy saving and carbon reduction and economic benefits, and enhancing the adaptive ability and intelligent level of the system in complex operating environments, providing a complete technical support system for the green and low-carbon transformation of the booster station. BRIEF DESCRIPTION OF DRAWINGS

[0009] The disclosure of the present application will be described with reference to the accompanying drawings. It should be appreciated that the drawings are for purposes of illustration only and are not intended to limit the scope of the present application. In the drawings, the same reference numerals are used to refer to the same components:

[0010] Figure 1 is a whole logic diagram of the present application; Figure 2 is a multi-source data acquisition and preprocessing module logic diagram of the present application; Figure 3 is an AI energy efficiency prediction module logic diagram of the present application; Figure 4 is an energy efficiency and carbon emission evaluation module logic diagram of the present application. DETAILED DESCRIPTION

[0011] It is easy to understand that, according to the technical solution of the present application, those skilled in the art can propose a plurality of structures and implementation modes which can be replaced with each other without changing the essential spirit of the present application. Therefore, the following specific embodiments and drawings are only exemplary descriptions of the technical solution of the present application, and should not be regarded as the whole or as a limitation or restriction on the technical solution of the present application.

[0012] Reference Figure 1 It is shown that the present application provides a technical solution: a booster station plant data acquisition and automatic control energy saving management system, comprising: a multi-source data acquisition and preprocessing module: collecting heterogeneous data of various intelligent sensors in the booster station plant, preprocessing the original heterogeneous data to form a multi-source feature data set. AI energy efficiency prediction module: based on the feature data set, using attention mechanism enhanced long short-term memory network for multi-time scale load prediction, realizing device efficiency decay curve and health state early warning through Transformer architecture, using physical information neural network to couple environmental factors and new energy output, analyzing its influence mechanism on main transformer cooling efficiency, load capacity and net load, outputting load prediction, efficiency warning, coupling coefficient and model parameters. Energy efficiency and carbon emission evaluation module: based on real-time data and prediction results, establishing an energy efficiency-carbon emission dual-dimensional evaluation system, dynamically calculating the comprehensive energy efficiency index of the plant and identifying the energy efficiency bottleneck, and calculating the carbon footprint and green power emission reduction amount, generating a comprehensive evaluation report through benchmark comparison and compliance check. Adaptive control optimization module: according to the prediction and evaluation results, an optimization model is constructed, and the control strategies of transformer taps, capacitor banks, energy storage systems and auxiliary machines are solved in a rolling manner, and the instruction sequence is output.

[0013] Reference Figure 2As shown, in the present embodiment, the multi-source data acquisition and preprocessing module specifically includes: a heterogeneous data interface and acquisition module: for receiving the original data stream transmitted by various intelligent sensors in the substation through multi-protocol (such as IEC 61850, Modbus-TCP), such as intelligent electric meters, protection devices, and environmental sensors, etc., through protocol parser to convert non-standard signals into unified engineering values, and based on time stamp to perform preliminary data alignment and invalid data filtering, to ensure the real-time and consistency of the data source, and output the preliminary standardized real-time data sequence to the data quality management and synchronization module for deep processing, through the unified protocol analysis and preliminary alignment technology in this section to solve the problem of non-uniform data format of multi-source heterogeneous devices, which is difficult to be directly utilized, to achieve the effect of providing stable and standardized data source for the system.

[0014] The data quality management and synchronization module: for receiving standardized data, which is the deep processing version of the real-time data sequence, containing engineering values and time stamps, which can be directly used for management, implementing integrity check, rationality verification and consistency analysis, using wavelet transform to remove impulse noise, and using high-precision Beidou timing signal as reference for time stamp correction and interpolation synchronization, eliminating data conflicts and redundancies, outputting clean, time strictly aligned multi-source data collection to the operation situation awareness and feature extraction module, for state evaluation and feature mining, through the multi-dimensional data verification and synchronization technology in this section to solve the problem of analysis distortion caused by noise and asynchronization in original data, to achieve the effect of improving data quality and subsequent analysis accuracy.

[0015] The operation situation awareness and feature extraction module: for receiving clean aligned data, which is the result of the multi-source data collection after management, emphasizing data cleanliness and time synchronization, applying clustering analysis and anomaly detection algorithm to realize real-time awareness of the running state of the whole station equipment, normal, warning or abnormal of the equipment running, and using principal component analysis to extract key feature quantities from high-dimensional data, such as load rate, temperature gradient, harmonic distortion rate, to form a feature data set for advanced analysis, outputting the operation situation evaluation result and the feature data set to the multi-time scale load forecasting module of the AI energy efficiency prediction module, for the operation situation evaluation result, to the equipment efficiency decay prediction module, for the operation situation evaluation result for equipment state analysis, and to the energy efficiency index dynamic calculation module of the energy efficiency and carbon emission evaluation module, for the feature data set for energy efficiency calculation, through real-time situation awareness and feature dimension reduction technology to solve the problems of data redundancy, information island and non-obvious key features, to achieve the effect of providing deep data support for intelligent decision-making.

[0016] Reference Figure 3As shown, in the present embodiment, the AI energy efficiency prediction module specifically comprises: a multi-time scale load prediction module: for receiving the feature data set from the operation situation awareness and feature extraction module, combining historical load curve, weather forecast data and holiday marker, using an Attention-LSTM model enhanced by attention mechanism, respectively performing short-term and ultra-short-term plant total active load and key auxiliary machine load prediction, and quantifying prediction uncertainty, outputting multi-time scale load prediction results with confidence interval, and outputting to the equipment efficiency attenuation prediction module and the environment and new energy coupling analysis module for efficiency analysis and coupling calculation, through the Attention-LSTM multi-time scale prediction model to solve the problem of being unable to accurately predict future energy efficiency trends according to real-time load, weather and other dynamic factors, and achieve the effect of providing accurate and reliable load forward-looking information.

[0017] The equipment efficiency attenuation prediction module: for receiving the prediction results and real-time operation data, the real-time operation data coming from the feature data set output by the operation situation awareness and feature extraction module of the multi-source data acquisition and preprocessing module, such as transformer oil temperature, winding current, capacitor dielectric loss value, based on the Transformer architecture, constructing an equipment health state evaluation model, through analyzing the deviation trend of the operation parameters and the benchmark efficiency, predicting the efficiency attenuation curve and the remaining high-efficiency operation life of the key equipment, including main transformer, circuit breaker, capacitor bank, electric reactor, GIS equipment, current transformer, voltage transformer, surge arrester, relay protection device, auxiliary power system, etc., identifying early signs of performance degradation, outputting the efficiency attenuation rate prediction and health state early warning of the key equipment, and outputting to the environment and new energy coupling analysis module and the model online self-learning and performance monitoring module for coupling analysis and model optimization, through the Transformer equipment health state prediction model to solve the problem of being unable to autonomously learn and adjust the operation strategy, and achieve the effect of realizing the preventive maintenance of equipment and the adaptive adjustment of operation strategy.

[0018] Further, as the optimal embodiment, the device health state evaluation model is constructed based on the Transformer architecture, by analyzing the deviation trend of the operating parameters and the benchmark efficiency, the efficiency decay curve and the remaining high-efficiency operation life of the key equipment are predicted, and the following operations are performed: collect the historical operating parameter data of the key equipment, including the time series signals of current, voltage, temperature, vibration frequency, etc., and the corresponding benchmark efficiency value, standardize the data for pretreatment to eliminate the dimension effect, and divide it into training set and test set, construct the Transformer encoder model, which includes multi-head self-attention mechanism and feedforward neural network layer, which is used to capture the long-term dependence and local features in the operating parameter sequence, in the training stage, use the mean square error loss function and Adam optimizer, minimize the deviation between the predicted efficiency and the real efficiency through the back propagation algorithm, and introduce the early stopping mechanism to prevent overfitting, in the prediction stage, input the real-time operating parameter sequence, the model outputs the key parameters of the efficiency decay curve, such as decay coefficient and initial efficiency value, and then calculates the remaining high-efficiency operation life through integral, such as remaining life through numerical integral efficiency decay curve from the current time to the efficiency threshold, evaluate the model robustness through cross-validation and confidence interval, and deploy it to the edge computing device to realize online monitoring and early warning.

[0019] The formula of the Transformer encoder model is: wherein, represents the device efficiency value predicted at time , is the output of the Transformer multi-head self-attention mechanism, query, key and value matrices respectively, obtained by standardizing the input sequence by linear transformation, is the running parameter sequence (such as current, voltage, temperature) from time to , with dimension , is the sequence length, is the feature dimension, is the dimension of the key vector, used to scale the dot product, and are the weight and bias parameters of the output layer, obtained by training, and the final output is used to calculate the efficiency decay curve wherein is the initial efficiency, is the decay coefficient, and the remaining high-efficiency operation life is , is the efficiency threshold, and the parameter is an indicator function, which is used to determine whether the current efficiency of the device is higher than the preset efficiency threshold.

[0020] The environment and new energy coupling analysis module is used to receive load data, device state data and real-time environment data. The device state data includes efficiency decay rate prediction and health state warning, and is used to quantify the response of device performance to the environment. The real-time environment data comes from the environment sensor data of the multi-source data acquisition and preprocessing module, such as environmental temperature, humidity, wind speed and photovoltaic output. The physical information neural network is used to combine the device heat dissipation model, the aerodynamics model and the data-driven model, to quantitatively analyze the coupling influence mechanism of environmental factors and new energy output on the cooling efficiency of the main transformer, the load flow of overhead lines and the system net load, to predict the energy efficiency changes under different conditions, and to output the environmental and new energy coupling influence coefficient and the net load power curve. The output is given to the model online self-learning and performance monitoring module and the energy efficiency and carbon emission evaluation module, which are used for model optimization and evaluation calculation. The physical information neural network coupling modeling technology is used to solve the problem of accurately predicting the future energy efficiency trend according to real-time load, weather, electricity price and other dynamic factors, so as to improve the generalization ability and accuracy of the energy efficiency prediction model under complex environment.

[0021] Further, as the optimal embodiment, the physical information neural network is used to combine the device heat dissipation model, the aerodynamics model and the data-driven model, to quantitatively analyze the coupling influence mechanism of environmental factors and new energy output on the cooling efficiency of the main transformer, the load flow of overhead lines and the system net load, to predict the energy efficiency changes under different conditions. The following operations are performed: the physical equation of device heat dissipation is defined based on the Fourier heat conduction law, and the simplified form of the Navier-Stokes equation of aerodynamics, a physical information neural network is constructed, the input layer of which receives environmental temperature, humidity, wind speed, photovoltaic output and other variables, the hidden layer of which is composed of multiple fully connected layers, and the output layer of which predicts the cooling efficiency of the main transformer, the load flow of overhead lines and the system net load. In the training process, the loss function is weighted sum of data-driven loss and physical constraint loss. The data-driven loss uses mean square error to calculate the deviation between the predicted value and the real monitoring value. The physical constraint loss calculates the partial derivative of the neural network output to the input by automatic differentiation, and forces it to satisfy the physical equation residual. For example, the physical constraint loss ensures that the network output meets the laws of thermodynamics and fluid mechanics through the PDE residual term. The total loss is optimized by gradient descent algorithm. The network learns the nonlinear coupling relationship between environmental factors and new energy output, reasons under different boundary conditions, predicts the energy efficiency change trend, and provides data support for real-time control.

[0022] The specific formula of the coupling influence mechanism is: wherein, is the total loss function, and These are hyperparameter weighting coefficients. and standards They are the first The true and predicted values ​​(such as cooling efficiency or flow rate) of each sample. It refers to the number of data samples. It is the partial derivative of the predicted output with respect to time. It is a convective term. It is a wind speed vector. It is the thermal diffusivity. The Laplace operator represents the diffusion effect. It represents the number of physical constraint points, obtained through random sampling within the domain.

[0023] The online self-learning and performance monitoring module continuously receives real feedback data from the system's operation, including efficiency decay rate prediction, health status warnings, environmental and renewable energy coupling influence coefficients, and net load power curves. Employing an incremental learning and federated learning framework, it periodically updates parameters and fine-tunes the structure of load forecasting, efficiency forecasting, and environmental coupling analysis. Simultaneously, it monitors model performance indicators (such as MAPE and RMSE), establishes a performance degradation warning mechanism, and ensures the prediction model adapts to changes in the power grid structure and equipment aging. It outputs updated AI model parameters and performance reports, which are then shared with other sub-modules within this module and the adaptive control optimization module for real-time optimization of prediction accuracy and control decisions. Through incremental learning and online performance monitoring optimization mechanisms, it addresses the problem of the system's inability to learn and adjust its operating strategies autonomously, achieving the goal of enabling the prediction system to continuously evolve and maintain long-term prediction accuracy.

[0024] Reference Figure 4As shown, in the present embodiment, the energy efficiency and carbon emission evaluation module specifically comprises: an energy efficiency index dynamic calculation module: for receiving real-time data of the operation situation awareness and feature extraction module and the prediction results of the AI energy efficiency prediction module, the prediction results specifically come from the load prediction results of the multi-time scale load prediction module and the health state warning of the equipment efficiency attenuation prediction module, based on the standard dynamic calculation of the comprehensive energy efficiency index of the plant area, and introducing the data envelopment analysis (DEA) method, the relative energy efficiency level of the operation unit of the booster station plant area is evaluated online, the energy efficiency bottleneck is identified, and a multi-dimensional dynamic energy efficiency index set and energy efficiency bottleneck analysis report are output, which are output to the carbon footprint precise accounting module and the green power consumption and emission reduction accounting module for carbon emission correlation analysis and emission reduction calculation. Through the DEA online energy efficiency evaluation method, the problem of lack of monitoring, analysis and optimization of key indicators such as transformer loss and auxiliary machine energy consumption in the station is solved, and the effect of fine and benchmarking management of energy efficiency is achieved. It needs to be supplemented that: each operation unit refers to those electrical equipment or auxiliary systems in the booster station plant area that can be independently operated, have clear energy consumption boundaries and can be individually evaluated for energy efficiency. The main electrical equipment includes: for example, transformers, circuit breakers, capacitor banks, reactors, GIS combined electrical apparatus, etc. The auxiliary systems include: for example, lighting systems, ventilation systems, cooling systems, air conditioning systems, DC power supply systems, etc.

[0025] Further, as the optimal embodiment, based on the standard dynamic calculation of the comprehensive energy efficiency index of the plant area, and introducing the data envelopment analysis method, the relative energy efficiency level of each operation unit is evaluated online, and the energy efficiency bottleneck is identified, which specifically performs the following operations:

[0026] According to the standard ISO, the comprehensive energy efficiency index of the plant area is defined, including the plant power rate, transformer operation efficiency, auxiliary unit energy consumption, etc. The input data (such as power consumption, fuel consumption) and output data (such as power generation, processing capacity) of each operation unit are collected in real time. The CCR model in the data envelopment analysis is applied, each unit is regarded as a decision unit, and the relative efficiency value is calculated by linear programming. In online evaluation, the data is updated using a sliding window, and the efficiency score is calculated by solving the optimization problem. The units with efficiency values less than 1 are identified as potential bottlenecks. For example, the key improvement factors of the bottleneck unit are determined through efficiency sorting and sensitivity analysis. The improvement target value of the non-effective unit is calculated through projection analysis, thereby providing a quantitative basis for operation and maintenance decisions. The evaluation results are integrated into a database and visualized.

[0027] Carbon footprint precise accounting module: used for receiving energy consumption data of energy efficiency index dynamic calculation module and power grid carbon emission factor, energy consumption data is the core component of dynamic energy efficiency index set, used for quantifying energy consumption, power grid carbon emission factor is an external dynamic input parameter, used for converting energy consumption into carbon emission, adopting life cycle assessment method, accounting for direct and indirect carbon emissions within the operating boundary of the booster station, and incorporating potential emission sources such as SF6 gas leakage, establishing carbon flow model to track the spatial and temporal distribution of carbon emissions, outputting carbon footprint list of each device and time period, outputting to green power consumption and emission reduction accounting module and benchmark comparison and compliance checking module, used for emission reduction calculation and compliance checking, through the precise accounting model based on life cycle assessment method and dynamic carbon factor to solve the problem that the system cannot convert energy consumption data into carbon footprint index, achieving the effect of panoramic and traceable accounting of carbon emissions.

[0028] Green power consumption and emission reduction accounting module: used for receiving carbon emission data from carbon footprint precise accounting module and distributed photovoltaic output data, carbon emission data comes from carbon footprint list, used for calculating emission reduction benchmark, distributed photovoltaic output data comes from net load power curve output by AI energy efficiency prediction module, accurately accounting for the proportion of green power generated and used by the plant, calculating the carbon dioxide emission reduction brought by this, outputting green power consumption and emission reduction accounting report, outputting to benchmark comparison and compliance checking module and comprehensive evaluation report generation module, used for compliance checking and report integration, through green power precise accounting and emission reduction development technology to solve the problem that the system cannot convert energy consumption data into carbon footprint index, and to achieve the effect of mining the environmental value of green power.

[0029] Benchmark comparison and compliance checking module: used for receiving energy efficiency data, carbon emission data and emission reduction data, dynamic energy efficiency index set and energy efficiency bottleneck analysis report output by energy efficiency index dynamic calculation module, carbon emission data comes from carbon footprint list output by carbon footprint precise accounting module, emission reduction data comes from green power consumption and emission reduction accounting report output by green power consumption and emission reduction accounting module, automatically comparing the dynamically calculated energy efficiency and carbon emission indicators with historical benchmarks, industry benchmark values and industry threshold values, identifying the main driving factors of indicator deviation through regression analysis, and checking carbon emission compliance risks in real time through rule engine, outputting gap analysis report and compliance warning information, outputting to comprehensive evaluation report generation module and user interface module, user interface module is the human-computer interaction component of the system, used for visualizing data, report and warning, supporting operator monitoring and decision-making, such as displaying energy efficiency and carbon emission status in real time through Web or mobile interface, used for report generation and warning display, through multi-benchmark comparison and rule engine driven checking mechanism to solve the problem of difficult to meet the fine management and policy compliance requirements of modern power system for green and low-carbon operation, achieving the effect of automatically identifying energy efficiency gap and compliance risk.

[0030] The comprehensive evaluation report generation module integrates data from the dynamic energy efficiency index set output by the dynamic energy efficiency index calculation module, the energy efficiency bottleneck analysis report, the carbon footprint list output by the carbon footprint precise accounting module, the green power consumption and emission reduction accounting report output by the green power consumption and emission reduction accounting module, and the gap analysis report and compliance warning information output by the benchmark comparison and compliance checking module. Based on the preset report template, a comprehensive energy efficiency and carbon emission evaluation report with pictures and text is automatically generated. The report content includes trend analysis, benchmarking results, potential problems and summary suggestions, and supports one-key export to a standard format (such as PDF, Excel). The final comprehensive evaluation report is output to the user interface module for visual display and decision support. Through the automatic report generation and multi-source information integration technology, the problem of lack of monitoring, analysis and optimization of key indicators such as transformer loss and auxiliary machine energy consumption is solved, and the effect of providing intuitive decision support information for management is achieved.

[0031] Reference Figure 1 As shown in the figure, in the embodiment, the adaptive control optimization module specifically includes:

[0032] The multi-objective real-time optimization decision module is used to receive the prediction results from the multi-time scale load prediction module, the equipment efficiency attenuation prediction module, and the environment and new energy coupling analysis module, and the evaluation data from the energy efficiency index dynamic calculation module, the carbon footprint precise accounting module, and the benchmark comparison and compliance checking module. The system total operation cost is minimized and the carbon emission is minimized as the target. A multi-objective optimization model is constructed. An improved non-dominated sorting genetic algorithm (NSGA-II) is combined with a model predictive control (MPC) framework to solve the optimal control problem in a limited time domain. A refined control strategy including transformer tap position, capacitor bank switching, energy storage charging and discharging plan, and auxiliary machine start-stop is generated. A real-time rolling optimization control instruction sequence is output. The safety execution and closed-loop feedback module is output for final execution. Through the multi-objective optimization algorithm and the MPC rolling optimization framework, the problem of unable to realize the optimal control of energy efficiency maximization and operation cost is solved. The effect of realizing the coordinated optimization of economic and environmental benefits under the premise of safety is achieved.

[0033] Further, as the optimal embodiment, the improved non-dominated sorting genetic algorithm is combined with the model predictive control framework to solve the optimal control problem in a limited time domain. A refined control strategy including transformer tap position, capacitor bank switching, energy storage charging and discharging plan, and auxiliary machine start-stop is generated. The following operations are specifically performed:

[0034] A multi-objective optimization function is defined, including minimizing operation cost and carbon emission, and system constraints are set, such as voltage stability, equipment capacity and power balance, an improved non-dominated sorting genetic algorithm is adopted, control variables are coded as real and binary mixed strings, representing transformer tap position, capacitor group state, energy storage charging and discharging power and auxiliary machine start-stop plan, in the rolling time domain, a model predictive control framework is used, the state evolution is predicted based on the system dynamic model, and the multi-objective optimization problem is solved, for example, the rolling optimization is realized by repeatedly solving the current time domain problem and implementing the first step control, in NSGA-II, the crowdedness calculation and elite reservation strategy are introduced to maintain the diversity of the solution set, while the crossover and mutation operations are used to generate new individuals, the Pareto optimal frontier is output, and the satisfactory solution is selected according to the preference, and is sent to the actuator to realize real-time adjustment.

[0035] The safe execution and closed-loop feedback module is used for receiving a control instruction sequence, performing safety checking before execution, such as N-1 checking and device operation sequence logic checking, preventing misoperation by using a software and hardware interlocking mechanism, collecting device state feedback in real time after the instruction is issued, comparing with the expected state, triggering strategy re-planning if the deviation is out of limit, recording the execution success rate and effect, forming a complete control closed loop, outputting safe and reliable and verified execution results, and outputting to the existing control system or actuator of the booster station for actual device adjustment.

[0036] The technical scope of the present application is not limited to the above description, and those skilled in the art can make various modifications and changes to the above embodiments without departing from the technical idea of the present application, and these modifications and changes should belong to the protection scope of the present application.

Claims

1. A data acquisition, automatic control, and energy-saving management system for a booster substation, characterized in that: include: Multi-source data acquisition and preprocessing module: Collects heterogeneous data from various intelligent sensors within the substation area, preprocesses the raw heterogeneous data to form a multi-source feature dataset; AI energy efficiency prediction module: Based on the feature dataset, uses an attention-enhanced long short-term memory network for multi-timescale load prediction, implements equipment efficiency degradation curves and health status early warnings through a Transformer architecture, and utilizes a physical information neural network to couple environmental factors and new energy output to analyze their impact mechanism on main transformer cooling efficiency, current carrying capacity, and net load, outputting load predictions, efficiency warnings, coupling coefficients, and model parameters; wherein, the implementation of equipment efficiency degradation curves and health status early warnings through a Transformer architecture includes: constructing an equipment health status assessment model based on the Transformer architecture, and predicting the efficiency of key equipment by analyzing the deviation trend between operating parameters and baseline efficiency. The efficiency decay curve and remaining efficient operating life are determined by the following steps: Historical operating parameter data of key equipment are collected, the data is standardized and preprocessed, and split into training and testing sets. A Transformer encoder model is constructed, which includes a multi-head self-attention mechanism and a feedforward neural network layer to capture long-term dependencies and local features in the operating parameter sequence. During training, a mean squared error loss function and an Adam optimizer are used to minimize the deviation between predicted and actual efficiency through backpropagation, and an early stopping mechanism is introduced to prevent overfitting. During prediction, the real-time operating parameter sequence is input, and the model outputs the key parameters of the efficiency decay curve. The remaining efficient operating life is then calculated through integration. The robustness of the model is evaluated through cross-validation and confidence intervals, and it is deployed to edge computing devices for online monitoring and early warning. The Transformer encoder model formula is: ,in, Indicates time Predicted equipment efficiency values, It is the output of the Transformer's multi-head self-attention mechanism. These are query, key, and value matrices, respectively. It is the dimension of the key vector. and standards The weights and bias parameters of the output layer are learned through training, and the final output is used to calculate the efficiency decay curve. ,in It is the initial efficiency. It is the attenuation coefficient, the standard for remaining high-efficiency operating life. , It is the efficiency threshold, parameter This is an indicator function that determines whether the current efficiency of the equipment is higher than a preset efficiency threshold. The method of using a physical information neural network to couple environmental factors and renewable energy output to analyze their impact on the main transformer cooling efficiency, current carrying capacity, and net load includes: combining the equipment heat dissipation model, aerodynamic model, and data-driven model using a physical information neural network to quantitatively analyze the coupling influence mechanism of environmental factors and renewable energy output on the main transformer cooling efficiency, overhead line current carrying capacity, and system net load, and predicting energy efficiency changes under different conditions; specifically: defining the physical equation for equipment heat dissipation, constructing a physical information neural network based on the simplified form of Fourier's heat conduction law and the Navier-Stokes equations of aerodynamics, and during training, the loss function is a weighted sum of data-driven loss and physical constraint loss, with the data-driven loss using mean square... The error calculation determines the deviation between the predicted and actual monitored values. Physical constraint losses are calculated by automatically differentiating the partial derivatives of the neural network output with respect to the input and forcing the residuals to satisfy the physical equations. The total loss is optimized using a gradient descent algorithm, enabling the network to learn the nonlinear coupling relationship between environmental factors and new energy output. Under different boundary conditions, it infers and predicts the trend of energy efficiency changes. The energy efficiency and carbon emission assessment module establishes a dual-dimensional assessment system of energy efficiency and carbon emissions based on real-time data and prediction results. It dynamically calculates the comprehensive energy efficiency index of the plant area and identifies energy efficiency bottlenecks. At the same time, it calculates the carbon footprint and green electricity emission reduction. A comprehensive assessment report is generated through benchmark comparison and compliance checks. The adaptive control optimization module constructs an optimization model based on the prediction and assessment results, continuously solves the control strategies of transformer tap changers, capacitor banks, energy storage systems, and auxiliary equipment, and outputs a command sequence.

2. The data acquisition, automatic control, and energy-saving management system for booster station areas according to claim 1, characterized in that: The multi-source data acquisition and preprocessing module specifically includes: a heterogeneous data interface and acquisition module: used to receive raw data streams transmitted from various intelligent sensors within the substation area via multiple protocols, convert non-standard signals into unified engineering values ​​through a protocol parser, perform preliminary data alignment and invalid data filtering based on timestamps, and output real-time data sequences; a data quality governance and synchronization module: used to receive standardized data, use wavelet transform to remove impulse noise, and perform time-scale correction and interpolation synchronization based on high-precision BeiDou time signals, outputting a multi-source data set; and an operational status awareness and feature extraction module: used to receive clean and aligned data, apply cluster analysis and anomaly detection algorithms to perceive the operational status of all equipment in the station in real time, and use principal component analysis to extract key features from high-dimensional data to form a feature dataset.

3. The data acquisition, automatic control, and energy-saving management system for booster station areas according to claim 1, characterized in that: The AI ​​energy efficiency prediction module specifically includes: a multi-timescale load prediction module, which receives feature datasets, combines historical load curves, weather forecast data, and holiday markers, and uses an attention-enhanced long short-term memory network model to predict the total active power load and key auxiliary equipment load of the plant in the short and ultra-short term, respectively, quantifies the prediction uncertainty, and outputs multi-timescale load prediction results; and an equipment efficiency degradation prediction module, which receives prediction results and real-time operating data, builds an equipment health status assessment model based on the Transformer architecture, analyzes the deviation trend between operating parameters and baseline efficiency, predicts the efficiency degradation curve and remaining high-efficiency operating life of key equipment, identifies early signs of performance degradation, and outputs predictions of the efficiency degradation rate and health status warnings of key equipment. The Environment and New Energy Coupling Analysis Module receives load data, equipment status data, and real-time environmental data. It utilizes a physical information neural network to combine equipment heat dissipation models, aerodynamic models, and data-driven models to quantitatively analyze the coupling impact mechanism of environmental factors and new energy output on the main transformer cooling efficiency, overhead line current carrying capacity, and system net load. It predicts energy efficiency changes under different conditions and outputs the environment and new energy coupling impact coefficient and net load power curve. The Online Self-Learning and Performance Monitoring Module continuously receives real-time feedback data from system operation. Employing incremental learning and federated learning frameworks, it periodically updates parameters and fine-tunes the structure of load forecasting, efficiency forecasting, and environmental coupling analysis. Simultaneously, it monitors model performance indicators, establishes a performance degradation early warning mechanism, and outputs AI model parameters and performance reports.

4. The automatic data acquisition and energy-saving management system for substation areas according to claim 1 or 3, characterized in that: The specific formula for the coupling effect mechanism is as follows: ,in, It is the total loss function. and These are hyperparameter weighting coefficients. and standards They are the first The true and predicted values ​​of each sample It refers to the number of data samples. It is the partial derivative of the predicted output with respect to time. It is a convective term. It is a wind speed vector. It is the thermal diffusivity. The Laplace operator represents the diffusion effect. It refers to the number of physical constraint points.

5. The data acquisition, automatic control, and energy-saving management system for booster station areas according to claim 1, characterized in that: The energy efficiency and carbon emission assessment module specifically includes: a dynamic energy efficiency index calculation module: used to receive real-time data and prediction results, dynamically calculate the comprehensive energy efficiency index of the plant area based on standards, and introduce data envelopment analysis to evaluate the energy efficiency level of the substation's operating units online, identify energy efficiency bottlenecks, and output a dynamic energy efficiency index set and an energy efficiency bottleneck analysis report; a precise carbon footprint accounting module: used for energy consumption data and grid carbon emission factors, employing life cycle assessment methods to calculate direct and indirect carbon emissions within the substation's operating boundary, incorporating potential emission sources, establishing a carbon flow model to track the spatiotemporal distribution of carbon emissions, and outputting a carbon footprint inventory; and a green electricity consumption and emission reduction accounting module: used for carbon emission data and distributed photovoltaic output data to accurately calculate the proportion of green electricity generated and consumed by the plant area, and calculate the resulting benefits. The system includes: a carbon dioxide emission reduction module, which outputs a green electricity consumption and emission reduction accounting report; a benchmark comparison and compliance check module, which receives energy efficiency data, carbon emission data, and emission reduction data, automatically compares dynamically calculated energy efficiency and carbon emission indicators with historical benchmarks, industry benchmark values, and industry thresholds, uses regression analysis to identify the main driving factors of indicator deviations, and uses a rule engine to check carbon emission compliance risks in real time, outputting a gap analysis report and compliance warning information; and a comprehensive assessment report generation module, which integrates dynamic energy efficiency indicator sets, energy efficiency bottleneck analysis reports, carbon footprint lists, green electricity consumption and emission reduction accounting reports, gap analysis reports, and compliance warning information, and automatically generates a comprehensive energy efficiency and carbon emission assessment report with graphics and text based on preset report templates, outputting the final comprehensive assessment report.

6. The data acquisition, automatic control, and energy-saving management system for booster station areas according to claim 5, characterized in that: The method described above is based on standard dynamic calculation of the plant's comprehensive energy efficiency index and incorporates data envelopment analysis (DEA) to evaluate the energy efficiency level of the substation's operating units online and identify energy efficiency bottlenecks. Specifically, the following operations are performed: The plant's comprehensive energy efficiency index is defined according to the standard; input and output data of each operating unit are collected in real time; the CCR model in DEA is applied, treating each unit as a decision-making unit; relative efficiency values ​​are solved using linear programming; during online evaluation, a sliding window is used to update data, and an optimization problem is solved to calculate efficiency scores; units with efficiency values ​​below 1 are identified as potential bottlenecks; improvement target values ​​for inefficient units are calculated using projection analysis; and the evaluation results are integrated into a database and visualized.

7. The data acquisition, automatic control, and energy-saving management system for booster station areas according to claim 1, characterized in that: The adaptive control optimization module specifically includes: a multi-objective real-time optimization decision module: used to receive prediction results and evaluation data, construct a multi-objective optimization model, and combine an improved non-dominated sorting genetic algorithm with a model predictive control framework to solve the optimal control problem in the finite time domain in a rolling manner, generating a refined control strategy that includes transformer tap position, capacitor bank switching, energy storage charging and discharging plan, and auxiliary machine start-up and shutdown, and outputting a real-time rolling optimization control command sequence; and a safety execution and closed-loop feedback module: used to receive the control command sequence, perform safety checks before execution, collect equipment status feedback in real time after the command is issued, compare it with the expected status, and if the deviation exceeds the limit, trigger strategy replanning, and record the execution success rate and effect.

8. The data acquisition, automatic control, and energy-saving management system for booster station areas according to claim 7, characterized in that: The proposed method combines an improved non-dominated sorting genetic algorithm with a model predictive control framework to solve the optimal control problem in the finite-time domain on a rolling basis. This generates a refined control strategy that includes transformer tap positions, capacitor bank switching, energy storage charging and discharging plans, and auxiliary machine start-up and shutdown. Specifically, the following operations are performed: a multi-objective optimization function is defined, and system constraints are set. An improved non-dominated sorting genetic algorithm is used to encode control variables as a mixed string of real numbers and binary numbers. In the rolling time domain, the model predictive control framework is used to predict state evolution based on the system dynamic model and solve the multi-objective optimization problem. In NSGA-II, congestion calculation and an elite retention strategy are introduced to maintain the diversity of the solution set. At the same time, crossover and mutation operations are used to generate new individuals, output the Pareto optimal frontier, and select a satisfactory solution according to preferences, which is then sent to the actuator for real-time adjustment.

Images