Multi-source data fusion airport energy optimization operation method

Abstract

The invention discloses an airport energy optimization operation method based on multi-source data fusion, and belongs to the field of airport energy management. The method comprises the following steps: collecting airport real-time operation, flight scheduling and meteorological data; fusion and feature extraction are carried out on multi-source data based on a deep learning model, and future electric, thermal and cold loads are accurately predicted; by taking the prediction result as a boundary, constructing an optimal scheduling model with the lowest total operation cost as a target, and solving to generate a first scheduling instruction containing each equipment plan parameter; after the instruction is executed, an energy efficiency deviation coefficient is calculated by comparing actual and planned operation parameters of the equipment, the instruction is subjected to graded correction according to the coefficient, a second scheduling instruction is obtained, and closed-loop optimization control over the energy equipment is achieved through the industrial Ethernet. According to the method, the whole-process closed-loop management of prediction, optimization, execution and feedback correction is realized, and the economical efficiency, the self-adaptability and the intelligent level of an airport energy system are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of airport energy management, and in particular to an airport energy optimization operation method based on multi-source data fusion. BACKGROUND

[0002] In the field of civil aviation transportation, as a high energy consumption hub, the safe, economic and green operation of the energy system of the airport is of great importance. Most of the existing airport energy management systems follow the basic framework of "collection prediction dispatch", aiming to reduce operating costs through limited optimization means.

[0003] However, with the expansion of the airport scale, the increasing complexity of energy consumption structure and the access of renewable energy, the traditional framework gradually exposes obvious deficiencies in real-time, accuracy and adaptability, mainly in: Firstly, in terms of data perception and prediction, the existing method has weak fusion processing ability for multi-source dynamic information. Energy consumption prediction relies mainly on single historical data or simplified linear models, which is difficult to effectively couple real-time factors such as flight flow fluctuations and weather condition changes, resulting in large prediction deviation and inability to provide reliable basis for fine scheduling.

[0004] Secondly, in terms of optimization scheduling, the existing model mostly adopts static or offline optimization strategy, which fails to deeply link with real-time prediction. The scheduling scheme is often based on fixed rules or periodic plans, lacking the ability of forward rolling optimization, and is difficult to respond to load mutation and equipment state changes in time, which easily causes energy supply and demand imbalance or cost rise.

[0005] Finally, in terms of closed-loop control, the existing system generally lacks a continuous optimization mechanism based on operation feedback. The deviation between the actual energy efficiency of the energy equipment and the scheduling expectation is not effectively monitored, evaluated and utilized after the energy equipment executes the predetermined instruction, which leads to the inability of dynamic correction of the scheduling strategy and the bottleneck of the overall energy efficiency improvement of the system. SUMMARY

[0006] The purpose of the present application is to provide an airport energy optimization operation method based on multi-source data fusion, to overcome the technical defects of inaccurate energy consumption prediction, static and rigid scheduling strategy and lack of closed-loop feedback regulation ability in the prior art, so as to realize the advanced and accurate perception, dynamic collaborative optimization and continuous adaptive operation of the airport energy system, effectively improve the energy utilization efficiency and reduce the overall operating cost.

[0007] The embodiment of the present application provides an airport energy optimization operation method based on multi-source data fusion, comprising: Collect real-time operation data based on sensing devices deployed in each energy-using unit of the airport to obtain a current operation state data set of the airport, and obtain flight scheduling plan data and weather forecast data of the airport; Fuse and predict the current operation state data set, the flight scheduling plan data and the weather forecast data based on a pre-trained energy consumption prediction model to obtain predicted energy consumption demand of each region of the airport in a future set period; Build an airport energy optimization scheduling model with the goal of minimizing the total operation cost of the airport, input the predicted energy consumption demand as a core boundary condition into the airport energy optimization scheduling model for solving to obtain a first scheduling instruction of the energy equipment of the airport, the first scheduling instruction including planned operation parameters of each energy equipment in the future set period; Control the operation of the energy equipment based on the first scheduling instruction, obtain actual operation feedback data of each energy equipment, and for each energy equipment, calculate an energy efficiency deviation coefficient based on its actual operation parameters and the corresponding planned operation parameters in the first scheduling instruction; Optimize and correct the first scheduling instruction based on the energy efficiency deviation coefficient to obtain a second scheduling instruction, and perform real-time optimization control on the energy equipment based on the second scheduling instruction.

[0008] Optionally, the real-time operation data includes water, electricity, cold and warm energy production and consumption data and device operation state data of the airport, and the energy equipment includes at least one of a gas turbine, an energy storage battery, a photovoltaic inverter and a water chiller.

[0009] Optionally, the energy consumption prediction model includes a time series feature extraction module, a multi-source data fusion module and a fully connected layer; The time series feature extraction module is constructed based on a long short-term memory network and is used for extracting time series features in the current operation state data; The multi-source data fusion module is constructed based on an attention mechanism and is used for dynamic weight allocation and feature fusion of the time series features, flight features extracted from the flight scheduling plan data, and weather features extracted from the weather forecast data; The fully connected layer is used to map the fused features to the predicted energy consumption demand.

[0010] Optionally, the fusion and prediction of the current operation state data set, the flight scheduling plan data and the weather forecast data based on the pre-trained energy consumption prediction model to obtain the predicted energy consumption demand of each region of the airport in the future set period includes: Extract time series features from the current operation state data set based on the time series feature extraction module to obtain a time series feature vector of the current operation state data set; structurally parsing the flight scheduling plan data to obtain a flight feature vector in a future set period, wherein the flight features include flight peak landing and taking-off period, and passenger flow characteristics of the terminal; converting the environmental parameters of the weather forecast data into a weather feature vector, wherein the environmental parameters include temperature, humidity, and light intensity; inputting the time series feature vector, the flight feature vector, and the weather feature vector into the multi-source data fusion module, calculating contribution weights of different data source features to energy consumption prediction based on an attention mechanism, and performing weighted fusion on the time series feature vector, the flight feature vector, and the weather feature vector based on the contribution weights to obtain a fusion feature vector; inputting the fusion feature vector into a full connection layer of the energy consumption prediction model, and performing deep nonlinear fitting on the fusion feature vector based on multiple layers of neurons of the full connection layer to obtain energy consumption demand prediction values of each region of the airport in the future set period, wherein the energy consumption demand prediction values include power load, heat load, and cold load.

[0011] Optionally, the deep nonlinear fitting of the fusion feature vector based on the multiple layers of neurons of the full connection layer to obtain the energy consumption demand prediction values of each region of the airport in the future set period comprises: mapping the fusion feature vector to a high-dimensional feature space based on an input layer of the full connection layer to obtain a high-dimensional fusion feature vector; performing nonlinear transformation on the high-dimensional fusion feature vector based on a ReLU activation function to obtain a nonlinear activation feature vector, and performing layer-by-layer feature mapping and dimension adjustment on the nonlinear activation feature vector to obtain abstract features highly related to energy consumption demand prediction targets; mapping the abstract features to a specific numerical space of energy consumption demand of each region of the airport to obtain power load, heat load, and cold load prediction values of each region in the future set period.

[0012] Optionally, the airport energy optimization scheduling model takes the minimum total operating cost of the airport as an objective function, and constraint conditions of the objective function are energy production constraints, energy storage constraints, and load balancing constraints, wherein the energy production constraints include upper and lower limits of output of the energy equipment and climbing rate limits; the energy storage constraints include charge and discharge power limits and state of charge upper and lower limits of the energy storage battery, and the load balancing constraints require that the energy supply amount and the predicted energy consumption demand amount of each region of the airport remain in dynamic balance at any time.

[0013] Optionally, the predicted energy consumption demand is input as a core boundary condition into the airport energy optimization scheduling model to obtain a first scheduling instruction of the energy equipment of the airport, comprising: the predicted power load, heat load and cold load are taken as energy demand constraints of the airport energy optimization scheduling model; the upper and lower limits of the output of the energy equipment, the ramp rate and the start-stop frequency limit are taken as operation constraints; the charge and discharge power and capacity limit of the energy storage equipment are taken as energy storage constraints; the total operation cost of the airport is taken as an objective function, and the objective function is solved based on a mixed integer linear programming algorithm to obtain an optimal output plan of the energy equipment and an optimal charge and discharge strategy of the energy storage equipment; the optimal output plan and the optimal charge and discharge strategy are integrated into a first scheduling instruction, wherein the first scheduling instruction includes the start-stop time and output curve of the gas turbine, the active power output value of the photovoltaic inverter, the charge and discharge time period and power value of the energy storage battery, and the operation number and load distribution ratio of the water chilling unit.

[0014] Optionally, for each energy equipment, based on the actual operation parameter and the corresponding planned operation parameter in the first scheduling instruction, an energy efficiency deviation coefficient of the energy equipment is calculated, comprising: For each energy equipment, the following steps are performed: the actual operation parameter of the energy equipment is extracted from the actual operation feedback data of the energy equipment; the planned operation parameter corresponding to the energy equipment is read from the first scheduling instruction; the difference between the actual operation parameter and the planned operation parameter is calculated to obtain an energy consumption deviation of the energy equipment; based on the energy consumption deviation, an energy efficiency deviation coefficient of the energy equipment is calculated; wherein when the energy consumption deviation is positive, the energy efficiency deviation coefficient is positive, and when the energy consumption deviation is negative, the energy efficiency deviation coefficient is negative.

[0015] Optionally, the first scheduling instruction is optimized and corrected based on the energy efficiency deviation coefficient to obtain a second scheduling instruction, comprising: for the energy efficiency deviation coefficient of each energy equipment, the absolute value of the energy efficiency deviation coefficient is divided into correction levels; for the energy equipment whose absolute value of the energy efficiency deviation coefficient is less than or equal to a preset threshold, it is determined as a slight deviation, and the predicted energy consumption value of the energy equipment in the first scheduling instruction is fine-tuned based on a proportional adjustment algorithm, wherein the fine-tuning amplitude and the energy efficiency deviation coefficient are linearly and positively correlated; For the energy efficiency deviation coefficient of the energy equipment, if the absolute value of the energy efficiency deviation coefficient is greater than a preset threshold value, it is determined that there is a significant deviation, the energy efficiency deviation coefficient of each energy equipment is introduced as a penalty term into a target function of the airport energy optimization scheduling model, and the target function is re-solved to obtain a second scheduling instruction.

[0016] Optionally, the second scheduling instruction is used to perform real-time optimization control on the energy equipment, and the real-time optimization control comprises the following steps. The second scheduling instruction is converted into a control signal based on an industrial Ethernet. The control signal is sent to each sub-control system of an airport energy management system, wherein a gas turbine subsystem adjusts a fuel valve opening degree and an air intake amount based on a modified start-stop time and an output curve, and a chiller unit group control system adjusts a cooling tower fan rotating speed and a water pump flow based on a modified running number and a load distribution ratio. Execution feedback signals of the energy equipment are collected in real time, wherein the execution feedback signals comprise an actual rotating speed of the gas turbine, an output voltage and current of the photovoltaic array, a state of charge value of the energy storage battery and an outlet water temperature of the chiller unit. The execution feedback signals and set values of the second scheduling instruction are compared in real time to obtain deviation values, and an alarm mechanism is triggered when the deviation values exceed a deviation threshold value to notify an operation and maintenance personnel to intervene.

[0017] The technical scheme provided by the embodiment of the application deeply integrates multi-source heterogeneous data such as flight scheduling, weather forecast and real-time running state, performs self-adaptive feature fusion and extraction through a deep learning model based on an attention mechanism, and overcomes the defect that the traditional method does not consider dynamic factors. The method can accurately predict the electric, thermal and cold multi-element load demand of each area of the airport in the future period, provides a high-credibility data basis for subsequent optimization scheduling, and improves the accuracy of energy system planning.

[0018] Moreover, the airport energy optimization scheduling model taking the minimum total running cost as the target is constructed based on the accurate prediction result, the model uniformly represents the dynamic balance relationship between energy production, storage and load and the physical constraints of equipment. By solving the model, the start-stop, output and charging-discharging strategies of the multi-element equipment such as the gas turbine, the energy storage system, the photovoltaic array and the chiller unit are cooperatively optimized, the economic optimal day-ahead or day-in scheduling plan is formed, the waste of energy is effectively avoided, and the global optimal allocation of resources is realized.

[0019] Furthermore, this invention innovatively introduces an energy efficiency deviation coefficient as a feedback signal. By comparing the deviation between the actual operating parameters of the equipment and the scheduling plan values, the system can evaluate the execution efficiency of scheduling instructions online. For different deviation levels, it adaptively employs strategies such as proportional fine-tuning or re-optimization to dynamically correct the initial scheduling instructions. This enables the energy management system to adaptively adjust and continuously optimize in response to uncertain disturbances, improving the system's robustness and operational economy. Attached Figure Description

[0020] Figure 1 A flowchart illustrating an airport energy optimization operation method based on multi-source data fusion, provided as an embodiment of the present invention; Figure 2 for Figure 1 A flowchart of a specific implementation of S120; Figure 3 for Figure 1 A flowchart of a specific implementation of S130; Figure 4 for Figure 1 A flowchart of a specific implementation of S150. Detailed Implementation To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0021] like Figure 1 As shown in the figure, an airport energy optimization operation method based on multi-source data fusion is provided by an embodiment of the present invention. The method may include the following steps: S110: Based on the real-time operation data collected by the sensing devices deployed in each energy-consuming unit of the airport, the current operation status dataset of the airport is obtained, and the flight scheduling plan data and weather forecast data of the airport are acquired.

[0022] This step aims to provide a comprehensive and real-time data foundation for subsequent prediction and optimization. Specifically, firstly, by widely deploying sensing devices (such as smart meters, flow meters, temperature and humidity sensors, and equipment status monitoring units) in various energy-consuming units such as airport terminals, runways, lighting, and air conditioning systems, data on the production and consumption of various energy sources such as water, electricity, heating, and cooling, as well as the operating status data (such as start-up and shutdown status, operating frequency, and valve opening) of key equipment (such as chillers and gas turbines), a current operating status dataset reflecting the real-time operational status of the airport is formed.

[0023] Simultaneously, flight scheduling data for a future time period (e.g., the next 24 hours) is obtained from the airport operations information system. This data includes detailed information such as flight takeoff and landing times, aircraft types, and expected passenger volume. In addition, weather forecast data for the same period is obtained from the meteorological service interface. Key environmental parameters include temperature, humidity, and light intensity, which directly affect air conditioning load and photovoltaic power generation.

[0024] S120, based on a pre-trained energy consumption prediction model, the current operating status dataset, flight scheduling plan data and weather forecast data are fused and predicted to obtain the predicted energy consumption demand of each area of ​​the airport within a future set time period.

[0025] This step utilizes a pre-trained dedicated model to perform deep fusion and high-precision prediction on the three types of data acquired by S110. The energy consumption prediction model structurally integrates time-series learning and feature-focusing capabilities.

[0026] First, a time-series feature extraction module in the model (built based on the principle of long short-term memory network) processes the current operating status dataset, aiming to capture the long-term dependencies and periodic changes such as daily and weekly changes hidden in the energy consumption data, and output a time-series feature vector representing the historical operating mode.

[0027] Secondly, a multi-source data fusion module (built based on an attention mechanism) is responsible for integrating information from different sources. It receives the aforementioned time-series feature vectors, as well as flight feature vectors extracted from flight scheduling data through structured parsing (such as peak flight arrival and departure times and terminal passenger flow fluctuations), and meteorological feature vectors obtained after parametric transformation of weather forecast data (such as temperature, humidity, and light intensity sequences). This module dynamically calculates the contribution weights of each feature from different data sources to the final prediction target using an attention mechanism, and performs weighted fusion accordingly to generate a fused feature vector that adaptively focuses on key information.

[0028] Finally, the fused feature vector is subjected to deep nonlinear fitting and transformation through the model's fully connected layers (typically containing an input layer, multiple hidden layers, and an output layer, using a nonlinear activation function such as ReLU). This process first maps the features to a high-dimensional space for abstraction, then adjusts the dimensions layer by layer, ultimately mapping them to a specific numerical space that matches the energy consumption demands of various functional areas of the airport (such as different terminals, cargo areas, and apron areas), outputting the predicted energy consumption demands of each area within a future set time period. This prediction result is specifically quantified into three demand curves: electricity load, heat load, and cooling load.

[0029] S130, construct an airport energy optimization scheduling model with the goal of minimizing the total operating cost of the airport, input the predicted energy consumption demand as the core boundary condition into the airport energy optimization scheduling model for solution, and obtain the first scheduling instruction of the airport's energy equipment.

[0030] The first dispatch instruction contains the planned operating parameters for each energy device within a specified future time period.

[0031] This step involves developing a coordinated scheduling plan for energy equipment based on accurate demand forecasting, with the goal of achieving optimal economic efficiency.

[0032] First, an airport energy optimization scheduling model is constructed. This model is essentially a mathematical programming problem considering the coupling of multiple time scales and various energy forms. Its objective function is set to minimize the total operating cost of the airport energy system within the scheduling cycle. The model's constraint system aims to ensure the technical feasibility of the solution, mainly including: load balance constraints to ensure an accurate match between the total energy supply at any given time and the predicted energy demand obtained from S120; energy production constraints reflecting the physical operating limits of equipment such as gas turbines and photovoltaic inverters (e.g., minimum technical output, maximum output, and power change rate); energy storage constraints limiting the charging and discharging power range and state-of-charge safety boundary of energy storage batteries; and equipment operation constraints considering the start-up and shutdown logic and operating characteristics of equipment such as chillers.

[0033] Then, the predicted energy demand output by S120 (i.e., electricity, heat, and cooling load curves) is used as the core boundary condition input for model solving. An optimization algorithm suitable for handling mixed discrete and continuous variable problems (such as mixed-integer linear programming) is employed to solve the model. The solution results are the optimal power output plan and the optimal charging and discharging strategy. Integrating these strategies forms the first scheduling instruction sent to each energy device. This instruction specifically includes the planned operating parameters of equipment such as gas turbines, energy storage batteries, photovoltaic inverters, and chillers for various future time periods, such as planned start-up and shutdown times, planned power output, planned charging and discharging power and time periods, planned number of units in operation, and load allocation ratio.

[0034] S140, based on the first scheduling instruction, control the operation of the energy equipment, obtain the actual operation feedback data of each energy equipment, and for each energy equipment, calculate the energy efficiency deviation coefficient of the energy equipment based on its actual operating parameters and the planned operating parameters corresponding to the first scheduling instruction.

[0035] This step activates the system's closed-loop feedback mechanism, quantifying operational efficiency by comparing the differences between the plan and actual execution.

[0036] Based on the first scheduling command generated by S130, the energy management system drives energy equipment such as gas turbines and chillers to operation via the control network. During the execution of commands by the equipment, the system simultaneously collects its actual operational feedback data.

[0037] Next, for each dispatched energy device, the system performs the following calculations to evaluate its operational energy efficiency: Key actual operating parameters (such as actual power generation and actual cooling capacity) are extracted from the device's actual operating feedback data; the planned operating parameters for the device in the same time period are located and read from the first dispatch instruction. The difference between the actual parameters and the planned parameters is calculated to obtain the energy consumption deviation. Based on this deviation, a quantified energy efficiency deviation coefficient is calculated. A positive coefficient indicates that actual energy consumption is higher than planned (lower energy efficiency), while a negative coefficient indicates that actual energy consumption is lower than planned (higher energy efficiency or reduced demand). The absolute value reflects the significance of the deviation.

[0038] This step, by establishing a refined "plan-actual" comparison mechanism, has for the first time achieved online quantitative assessment of equipment-level energy efficiency within the closed loop of airport energy dispatch. It transforms the abstract question of "whether the operation is efficient" into a precisely calculable "energy efficiency deviation coefficient," providing an objective and quantitative basis for subsequent dynamic corrections. This fundamentally overcomes the technical shortcomings of traditional systems that rely solely on total statistics and cannot pinpoint the energy efficiency bottlenecks of specific equipment.

[0039] S150, the first scheduling instruction is optimized and corrected based on the energy efficiency deviation coefficient to obtain a second scheduling instruction, and the energy equipment is optimized and controlled in real time based on the second scheduling instruction.

[0040] This step involves dynamically adjusting the scheduling strategy based on the energy efficiency assessment results to achieve system self-adaptation and continuous optimization.

[0041] The system optimizes and corrects the first scheduling command based on the energy efficiency deviation coefficients of each device calculated in S140. The correction strategy is implemented in stages according to the severity of the deviation: For equipment with a small absolute value of the energy efficiency deviation coefficient (e.g., less than or equal to a certain preset threshold), it is determined that its operating status deviates slightly from the plan. The system uses a proportional adjustment algorithm to fine-tune the planned operating parameters of the equipment in subsequent time periods according to the magnitude of the deviation coefficient.

[0042] For equipment with a large absolute value of energy efficiency deviation coefficient (e.g., greater than a preset threshold), it is determined that there is a significant deviation. In this case, simple fine-tuning is insufficient to correct it. The system incorporates the energy efficiency deviation coefficient of each device as a penalty term into the objective function of the airport energy optimization scheduling model described in S130, constructing a new optimization problem that considers actual operational efficiency. By resolving this model, a fully revised optimization scheme that better fits the current actual operating state is obtained, namely the second scheduling instruction.

[0043] Ultimately, based on the second scheduling command, the system implements real-time optimized control of energy equipment through the control network. Simultaneously, the system continuously monitors execution feedback signals in real time, including the actual speed of the gas turbine, the real-time output voltage and current of the photovoltaic array, the real-time state of charge of the energy storage battery, and the real-time outlet water temperature of the chiller unit, comparing these signals with the set values ​​in the second scheduling command. If a deviation exceeds the permissible safety threshold, the system immediately triggers an alarm mechanism, notifying on-site maintenance personnel to inspect and intervene, thus ensuring the safety and reliability of the entire energy system while pursuing optimization goals.

[0044] The technical solution provided by this invention collects real-time operational data from multiple airport sources (including flight scheduling and weather forecasts), and performs high-precision fusion prediction using a deep learning-based energy consumption prediction model. Then, using the predicted demand as the boundary, an airport energy optimization scheduling model generates the most economically optimal initial scheduling instructions. During execution, the energy efficiency deviation coefficient is further calculated based on the deviation between the actual equipment operating data and the planned values, and this coefficient is used to dynamically feed back and correct the scheduling instructions, forming a closed-loop control system of "prediction-optimization-execution-evaluation-correction". This method overcomes the limitations of traditional static scheduling modes in terms of prediction accuracy, response speed, and adaptive capability, realizing the transformation of airport energy systems from passive response to active optimization, and from open-loop execution to closed-loop control, significantly improving energy utilization efficiency, operational economy, and system intelligence.

[0045] As one implementation of this invention, in step S120, the current operating status dataset, flight scheduling plan data, and weather forecast data are fused and predicted based on a pre-trained energy consumption prediction model to obtain the predicted energy consumption demand of each area of ​​the airport within a future set time period, such as... Figure 2 As shown, it may include the following steps: S121, Based on the time-series feature extraction module, perform time-series feature extraction on the current running state dataset to obtain the time-series feature vector of the current running state dataset.

[0046] The current operational status dataset is processed using a time-series feature extraction module within the energy consumption prediction model. This module, built upon a Long Short-Term Memory (LSTM) network, effectively captures the time-dependent and periodic patterns (such as intraday fluctuations and weekend effects) in operational data related to water, electricity, and heating / cooling energy consumption. Through learning and encoding historical sequences, this module outputs a time-series feature vector that characterizes the system's recent operational status and potential trends.

[0047] S122 performs structured analysis on the flight scheduling plan data to obtain flight feature vectors for a future set time period. The flight features include peak take-off and landing times and passenger flow characteristics of the terminal.

[0048] S123 converts the environmental parameters of the weather forecast data into a weather feature vector, where the environmental parameters include temperature, humidity and light intensity.

[0049] In parallel, the system performs feature processing on two other types of data. On the one hand, it performs structured parsing on the acquired flight scheduling plan data to identify the flight takeoff and landing distribution within a future set time period (such as the next 24 hours), extracting key flight features, such as peak takeoff and landing times and terminal passenger flow fluctuation curves estimated based on flight times and aircraft types, and encodes them into flight feature vectors. On the other hand, it normalizes and serializes key environmental parameters (including temperature, humidity, and light intensity) in the weather forecast data, converting them into meteorological feature vectors that can reflect dynamic changes in the external environment.

[0050] S124. Input the time series feature vector, flight feature vector and meteorological feature vector into the multi-source data fusion module. Calculate the contribution weight of different data source features to energy consumption prediction based on the attention mechanism. Then, perform weighted fusion of the time series feature vector, flight feature vector and meteorological feature vector based on the contribution weight to obtain the fused feature vector.

[0051] The time-series feature vector, flight feature vector, and meteorological feature vector obtained from the above steps are input together into the multi-source data fusion module of the energy consumption prediction model. The core of this module is an attention mechanism. This mechanism first calculates a contribution weight for each feature from different data sources, dynamically reflecting the feature's importance in the current prediction context. For example, in the summer afternoon, temperature and light intensity features may receive higher weights; while during peak flight hours, passenger flow features will have a significantly increased weight. Subsequently, the module performs a weighted summation of the time-series feature vector, flight feature vector, and meteorological feature vector based on the calculated weights, thereby generating a deeply fused feature vector that focuses on key prediction information. This process achieves adaptive focusing and efficient fusion among multi-source heterogeneous data.

[0052] S125 inputs the fused feature vector into the fully connected layer of the energy consumption prediction model. Based on the multi-layer neurons of the fully connected layer, the fused feature vector is subjected to deep nonlinear fitting to obtain the predicted energy consumption demand of each area of ​​the airport within a future set time period.

[0053] The energy demand forecast includes electricity load, heat load and cooling load.

[0054] As one implementation of this invention, deep nonlinear fitting of the fused feature vector based on multi-layer neurons of fully connected layers to obtain the predicted energy consumption demand of each area of ​​the airport within a future set time period may include the following steps: The first step is to map the fused feature vector to a high-dimensional feature space based on the fully connected input layer, thus obtaining the high-dimensional fused feature vector.

[0055] The second step involves performing a nonlinear transformation on the high-dimensional fused feature vector based on the ReLU activation function to obtain a nonlinear activated feature vector. This nonlinear activated feature vector is then subjected to layer-by-layer feature mapping and dimensional adjustment to obtain abstract features that are highly correlated with the energy consumption demand prediction target.

[0056] The third step is to map the abstract features to the specific numerical space of energy consumption demand in each area of ​​the airport, so as to obtain the predicted values ​​of power load, heat load and cooling load in each area in the future set time period.

[0057] Specifically, the fused feature vector output from the above steps is input into the fully connected layer of the energy consumption prediction model. The fully connected layer typically consists of an input layer, one or more hidden layers, and an output layer. The hidden layers often use activation functions such as ReLU to introduce nonlinearity. The fused feature vector is first mapped to a high-dimensional feature space through the input layer. Subsequently, through successive nonlinear transformations and feature abstraction in multiple hidden layers, the model gradually extracts deep patterns highly correlated with the final energy consumption demand. Finally, the output layer maps these abstract features to a specific numerical space, generating predicted energy consumption demands for various areas of the airport (such as Terminals 1 and 2, the cargo area, etc.) within a specified future time period. This predicted value is output in quantified form, specifically including three prediction curves: electricity load (in kW), heat load (in kW), and cooling load (in RT or kW), providing accurate and fine-grained demand input for subsequent optimized scheduling.

[0058] Based on the above embodiments, as one implementation of the present invention, the airport energy optimization scheduling model takes the minimum total operating cost of the airport as the objective function. The constraints of the objective function are energy production constraints, energy storage constraints, and load balance constraints. Among them, energy production constraints include the upper and lower limits of the output of energy equipment and the limit of the ramp rate; energy storage constraints include the charging and discharging power limit of energy storage batteries and the upper and lower limits of the state of charge; and load balance constraints require that the energy supply and the predicted energy demand of each area of ​​the airport remain in dynamic balance at any time.

[0059] Specifically, the model is a mathematical optimization model involving multiple time periods and multiple devices. Its core objective is to find the scheduling scheme that minimizes the total operating cost of the airport energy system within a scheduling cycle (such as the next 24 hours, with time intervals of 15 minutes or 1 hour).

[0060] The objective function is to minimize the total operating cost, which mainly consists of the following components: 1. Fuel cost: The cost incurred by the gas turbine in consuming natural gas during operation.

[0061] 2. Cost of purchased electricity: The cost of electricity purchased from an external power grid, usually at a time-of-use price.

[0062] 3. Equipment operation and maintenance and depreciation costs: This mainly considers the depreciation costs of key equipment (such as gas turbines and energy storage batteries) due to start-up, shutdown, and charge-discharge cycles.

[0063] To ensure the technical feasibility and system security of the scheduling scheme, the model must meet the following core constraints: 1. Energy Production Constraints: These constraints reflect the physical operating limits of controllable power generation equipment. For gas turbines, constraints include upper and lower limits of output (i.e., minimum technical output and rated maximum output) and ramp rate limits (i.e., the maximum increase or decrease in power between adjacent time periods). For photovoltaic inverters, their active power output is constrained between zero and the maximum power that the photovoltaic array can generate under the current illumination conditions.

[0064] 2. Energy Storage Constraints: These constraints are used to manage the charging and discharging behavior of energy storage batteries (such as lithium-ion batteries). Specifically, they include: charging and discharging power limits (maximum charging power and maximum discharging power); upper and lower limits of state of charge, meaning that the remaining capacity of the battery must always be kept within a preset safe range (such as 20% to 90%) to prevent overcharging or over-discharging; and energy conservation constraints, meaning that the state of charge at the end of the current period is equal to the state of charge at the end of the previous period plus the net charge amount in the current period (considering charging and discharging efficiency).

[0065] 3. Load Balance Constraint (Core Boundary): This constraint is crucial for connecting forecasting and scheduling, ensuring real-time matching of energy supply and demand. It requires that at any given moment within the scheduling cycle, the total energy supply provided by all energy equipment (gas turbines, photovoltaic systems, energy storage battery discharge, purchased electricity, etc.) must be precisely equal to the total predicted energy demand of each area of ​​the airport at the corresponding moment, as predicted in step S120. This constraint applies to the electricity, heat, and cooling systems respectively, constituting the rigid boundary conditions for model solving.

[0066] In addition, the model usually includes other necessary technical constraints, such as: equipment start-up and shutdown logic and minimum operation / downtime constraints (to prevent frequent start-up and shutdown of gas turbines and chillers); regional energy supply capacity constraints (such as chilled water pipeline transmission capacity limitations); and system safety operation constraints (such as voltage and frequency stability requirements).

[0067] By setting the above objective function and constraint system, this optimal scheduling model transforms a complex energy management problem into a mathematically accurate and efficiently solvable (mixed-integer linear) programming problem. The predicted energy demand output from step S120 is used as the specific numerical input to the model for load balancing constraints. Equipment parameters, cost parameters, etc., are also input, and the solver can then solve the problem, ultimately outputting the optimal scheduling scheme that satisfies all constraints and has the lowest total cost, i.e., the first scheduling instruction.

[0068] In one implementation of this invention, S130, the predicted energy consumption demand is input as a core boundary condition into the airport energy optimization scheduling model to obtain the first scheduling instruction for the airport's energy equipment, such as... Figure 3 As shown, it may include the following steps: S131 uses the predicted power load, heat load, and cooling load as energy demand constraints for the airport energy optimization scheduling model.

[0069] The predicted energy consumption demand for each area of ​​the airport within the future set time period, obtained from step S120, is specifically deconstructed into its core components: the electricity load prediction curve, the heat load prediction curve, and the cooling load prediction curve. These three prediction curves are used as precise numerical inputs to the energy demand constraints and are incorporated into the airport energy optimization scheduling model. This constraint requires that, in each scheduling period (e.g., every 15 minutes) solved by the model, the total output of all energy production and supply units must be strictly equal to the total predicted load value (electricity, heat, and cooling are balanced separately) for the corresponding period, thus transforming the prediction results into rigid boundaries that the optimization problem must satisfy.

[0070] S132 sets the upper and lower limits of energy equipment output, ramp rate, and start / stop frequency as operating constraints.

[0071] Based on the physical characteristics and operating procedures of the main controllable energy equipment within the airport, operational constraints are set in the model. These mainly include: 1. Output upper and lower limits: Set the minimum allowable technical output and maximum safe output for each gas turbine and each chiller unit.

[0072] 2. Climbing rate: Set the maximum allowable rate of power increase and decrease between adjacent scheduling periods for equipment such as gas turbines to ensure stable operation of the equipment.

[0073] 3. Start-up and shutdown frequency and minimum start-up and shutdown time limits: To prevent damage caused by frequent equipment start-ups and shutdowns, the maximum number of start-ups and shutdowns allowed within the scheduling cycle is set for gas turbines, chillers, etc., as well as the minimum continuous operation time after each start-up and the minimum shutdown time after shutdown.

[0074] S133 uses the charging and discharging power and capacity limits of energy storage devices as energy storage constraints.

[0075] For energy storage systems (such as lithium-ion battery energy storage stations), energy storage constraints are set in the model to regulate their charging and discharging behavior and ensure safety. Key constraints include: 1. Charge and discharge power limits: Set the maximum charging power and maximum discharging power of the energy storage battery.

[0076] 2. Capacity and State of Charge Limitations: Set the total usable capacity of the energy storage battery and stipulate that its state of charge must operate within a set safe range (e.g., 20% to 95%) to avoid overcharging and over-discharging.

[0077] 3. Energy conservation and efficiency: Establish dynamic equations for the state of charge between adjacent time periods, taking into account charging efficiency and discharging efficiency, to ensure that the model accurately reflects the actual energy flow of the energy storage system.

[0078] S134 uses the minimum total operating cost of the airport as the objective function and solves the objective function based on a mixed integer linear programming algorithm to obtain the optimal output plan of energy equipment and the optimal charging and discharging strategy of energy storage equipment.

[0079] After setting the above constraints, the objective function of the model is to minimize the total airport operating cost. This objective function is the optimization objective under the constraints defined in steps S131-S133. Subsequently, a mixed-integer linear programming algorithm suitable for handling integer variables (such as equipment start-up and shutdown status) and continuous variables (such as output value) is used to solve the model using the corresponding mathematical optimization solver. The solver's output is the optimal output plan for energy equipment and the optimal charging and discharging strategy for energy storage equipment that minimizes the total operating cost while satisfying all safety and technical constraints. This strategy specifies in detail the status and power values ​​of various types of equipment in each future scheduling period.

[0080] S135 integrates the optimal power output plan and the optimal charging and discharging strategy into a first dispatch instruction. The first dispatch instruction includes the start-up and shutdown time and output curve of the gas turbine, the active power output value of the photovoltaic inverter, the charging and discharging period and power value of the energy storage battery, and the number of operating chillers and the load allocation ratio.

[0081] The mathematically optimal strategy obtained in step S134 is converted into a first scheduling instruction that the energy management system can recognize and issue. This conversion process integrates abstract optimization variable values ​​into specific equipment control parameters. The generated first scheduling instruction is a structured instruction set, whose key contents include: 1. Gas turbine: Specific start-up and shutdown schedules and output curves (planned power generation) for each operating period.

[0082] 2. Photovoltaic inverter: Planned active power output values ​​for various future time periods (usually based on solar irradiance forecasts).

[0083] 3. Energy storage battery: Detailed charging and discharging schedule, and planned charging and discharging power values ​​for each charging and discharging period.

[0084] 4. Chillers: A list of the number of chillers to be put into operation, and the load distribution ratio among the chillers.

[0085] Through steps S131 to S135, this embodiment of the invention completes the transformation from accurate demand forecasting to specific, economical, and feasible equipment-level scheduling instructions for controlling various energy devices.

[0086] In one implementation of this invention, S150, real-time optimization control of energy equipment is performed based on the second scheduling instruction, such as... Figure 4 As shown, it includes: S151 converts the second scheduling command into a control signal based on industrial Ethernet.

[0087] The central controller of the energy management system receives the second scheduling instruction generated by the optimization and correction in step S150 via a highly reliable, low-latency industrial Ethernet communication network. This instruction is a digital instruction set containing specific control parameters for each device. The protocol conversion module within the central controller is responsible for encapsulating and converting these instruction parameters according to the communication protocols of various device subsystems, generating standard control signals that the devices can recognize and execute.

[0088] S152 sends control signals to the various sub-control systems of the airport energy management system. The gas turbine subsystem adjusts the fuel valve opening and air intake based on the corrected start-up and shutdown times and output curves. The chiller group control system adjusts the cooling tower fan speed and water pump flow based on the corrected number of operating units and load distribution ratio.

[0089] The converted control signals are transmitted via industrial Ethernet to the corresponding sub-control systems of the airport energy management system. Upon receiving the instructions, each subsystem drives the actuators to make precise adjustments. 1. Gas turbine subsystem: Based on the corrected start-stop time and output curve in the command, its controller adjusts the opening of the fuel regulating valve and the air intake in a coordinated manner to precisely control the combustion process and ensure that the actual output power of the generator set stably tracks the command requirements.

[0090] 2. Chiller Group Control System: Based on the corrected number of operating units and load distribution ratio in the instructions, the main controller first coordinates the start-up and shutdown status of each chiller unit, and then dynamically controls the speed of the cooling tower fan and the flow rate of the chilled water / cooling water circulating pump by adjusting the frequency converter, thereby accurately matching the total cooling capacity and distribution ratio of the system to the corrected cooling load demand.

[0091] S153 collects the execution feedback signals of energy equipment in real time. The execution feedback signals include the actual speed of the gas turbine, the output voltage and current of the photovoltaic array, the state of charge value of the energy storage battery, and the outlet water temperature of the chiller unit.

[0092] While the equipment executes control commands, sensors and monitoring units deployed on each energy device continuously collect execution feedback signals reflecting the actual operating status of the equipment and upload them to the central monitoring system via the same industrial Ethernet. Key feedback signals include: the actual speed and power generation of the gas turbine; the real-time output voltage and current of the photovoltaic array; the real-time state of charge, charging and discharging power, and battery pack temperature of the energy storage battery; and the chilled water outlet temperature, compressor operating current, and condensing pressure of the chiller unit.

[0093] S154 compares the set values ​​of the execution feedback signal and the second scheduling instruction in real time to obtain the deviation value. When the deviation value exceeds the deviation threshold, an alarm mechanism is triggered to notify the operation and maintenance personnel to intervene.

[0094] The central monitoring system compares and calculates the execution feedback signal collected in real time in step S153 with the set value corresponding to the second scheduling instruction issued in step S151 in the monitoring screen and background logic in real time to obtain the real-time deviation value of key operating parameters (e.g., the difference between actual power and set power, the difference between actual temperature and set temperature).

[0095] The system presets safety deviation thresholds for various parameters. When the real-time calculated deviation value exceeds its corresponding threshold, it indicates that the equipment operation has significantly deviated from the optimized instructions and may be in an abnormal or inefficient state. At this time, the system immediately triggers the alarm mechanism: a prominent alarm window pops up on the monitoring screen, detailed alarm logs are recorded, and on-duty maintenance personnel are immediately notified via sound and light, SMS, or mobile application push notifications. Based on the alarm information, maintenance personnel can quickly locate the problematic equipment, analyze the cause, and implement necessary manual intervention, thereby ensuring that the entire optimization control closed loop operates within a safe and reliable range.

[0096] Through steps S151 to S154, this embodiment of the invention completes a full control closed loop, from optimization decision-making to physical execution, and then to state feedback and safety monitoring. This not only enables refined, automated, and real-time control of energy equipment, but also provides a guarantee for the long-term stable and efficient operation of the system through an active alarm mechanism.

[0097] This invention also provides an electronic device, comprising: At least one processor; Memory for storing the at least one processor-executable instruction; The at least one processor is configured to execute the instructions to implement the method provided in the above embodiments.

[0098] This invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method provided in the above embodiments.

Claims

1. A method for optimizing airport energy operation through multi-source data fusion, characterized in that, include: Based on the real-time operation data collected by the sensing devices deployed in each energy-consuming unit of the airport, the current operation status dataset of the airport is obtained, and the flight scheduling plan data and weather forecast data of the airport are acquired. Based on a pre-trained energy consumption prediction model, the current operating status dataset, flight scheduling plan data, and weather forecast data are fused and predicted to obtain the predicted energy consumption demand of each area of ​​the airport within a future set time period. An airport energy optimization scheduling model is constructed with the goal of minimizing the total operating cost of the airport. The predicted energy consumption demand is input into the airport energy optimization scheduling model as the core boundary condition for solving, and the first scheduling instruction of the airport's energy equipment is obtained. The first scheduling instruction contains the planned operating parameters of each energy equipment in a future set time period. Based on the first scheduling instruction, control the operation of the energy equipment, obtain the actual operation feedback data of each energy equipment, and for each energy equipment, calculate the energy efficiency deviation coefficient of the energy equipment based on its actual operating parameters and the planned operating parameters corresponding to the first scheduling instruction. The first scheduling instruction is optimized and corrected based on the energy efficiency deviation coefficient to obtain a second scheduling instruction, and the energy equipment is then subjected to real-time optimized control based on the second scheduling instruction.

2. The method according to claim 1, characterized in that, The real-time operational data includes the airport's water, electricity, heating and cooling energy production and consumption data, as well as equipment operating status data. The energy equipment includes at least one of gas turbines, energy storage batteries, photovoltaic inverters, and chiller units.

3. The method according to claim 1, characterized in that, The energy consumption prediction model includes a time-series feature extraction module, a multi-source data fusion module, and a fully connected layer; The temporal feature extraction module is built based on a long short-term memory network and is used to extract temporal features from the current running state data; The multi-source data fusion module is built based on an attention mechanism and is used to dynamically assign weights and fuse features of the time-series features, flight features extracted from the flight scheduling plan data, and meteorological features extracted from the weather forecast data. The fully connected layer is used to map the fused features to the predicted energy consumption requirement.

4. The method according to claim 3, characterized in that, Based on a pre-trained energy consumption prediction model, the current operational status dataset, flight scheduling plan data, and weather forecast data are fused and predicted to obtain the predicted energy consumption demand for each area of ​​the airport within a future set time period, including: Based on the time-series feature extraction module, time-series features are extracted from the current running state dataset to obtain the time-series feature vector of the current running state dataset; The flight scheduling plan data is structured and parsed to obtain flight feature vectors for a future set time period, wherein the flight features include peak flight take-off and landing times and terminal passenger flow characteristics; The environmental parameters of the weather forecast data are converted into a weather feature vector, wherein the environmental parameters include temperature, humidity and light intensity; The time-series feature vector, flight feature vector, and meteorological feature vector are input into the multi-source data fusion module. The contribution weights of different data source features to energy consumption prediction are calculated based on the attention mechanism. The time-series feature vector, flight feature vector, and meteorological feature vector are then weighted and fused based on the contribution weights to obtain the fused feature vector. The fused feature vector is input into the fully connected layer of the energy consumption prediction model. Based on the multi-layer neurons of the fully connected layer, the fused feature vector is subjected to deep nonlinear fitting to obtain the predicted energy consumption demand values ​​for each area of ​​the airport within a future set time period. The predicted energy consumption demand values ​​include electricity load, heat load, and cooling load.

5. The method according to claim 4, characterized in that, Based on the multi-layer neurons of the fully connected layer, a deep nonlinear fitting is performed on the fused feature vector to obtain the predicted energy consumption demand of each area of ​​the airport within a future set time period, including: The input layer based on the fully connected layer maps the fused feature vector to a high-dimensional feature space to obtain a high-dimensional fused feature vector; The high-dimensional fused feature vector is nonlinearly transformed based on the ReLU activation function to obtain a nonlinear activation feature vector. The nonlinear activation feature vector is then subjected to layer-by-layer feature mapping and dimension adjustment to obtain abstract features that are highly correlated with the energy consumption demand prediction target. The abstract features are mapped to the specific numerical space of energy consumption demand in each area of ​​the airport to obtain the predicted values ​​of power load, heat load and cooling load in each area within a future set time period.

6. The method according to claim 1, characterized in that, The airport energy optimization scheduling model takes the minimum total operating cost of the airport as its objective function. The constraints of the objective function are energy production constraints, energy storage constraints, and load balance constraints. The energy production constraints include the upper and lower limits of the output of the energy equipment and the ramp rate limit. The energy storage constraints include the charging and discharging power limit of the energy storage battery and the upper and lower limits of the state of charge. The load balance constraint requires that the energy supply and the predicted energy demand in each area of ​​the airport remain in dynamic balance at any time.

7. The method according to claim 6, characterized in that, The predicted energy consumption demand is input as a core boundary condition into the airport energy optimization scheduling model to obtain the first scheduling instruction for the airport's energy equipment, including: The predicted energy demand for electricity, heat, and cooling is used as the energy demand constraint for the airport energy optimization scheduling model. The upper and lower limits of the output of the energy equipment, the ramp rate, and the number of start-stop cycles are used as operating constraints. The charging and discharging power and capacity limits of energy storage devices are used as constraints for energy storage. Using the minimum total operating cost of the airport as the objective function, the objective function is solved using a mixed-integer linear programming algorithm to obtain the optimal output plan of the energy equipment and the optimal charging and discharging strategy of the energy storage equipment. The optimal output plan and optimal charging and discharging strategy are integrated into a first scheduling instruction, wherein the first scheduling instruction includes the start-up and shutdown time and output curve of the gas turbine, the active power output value of the photovoltaic inverter, the charging and discharging period and power value of the energy storage battery, and the number of operating chillers and the load allocation ratio.

8. The method according to claim 7, characterized in that, For each energy device, the energy efficiency deviation coefficient is calculated based on its actual operating parameters and the planned operating parameters corresponding to the first scheduling instruction, including: For each energy device, perform the following steps: Extract the actual operating parameters of the energy equipment from the actual operation feedback data of the energy equipment; Read the planned operating parameters corresponding to the energy equipment from the first scheduling instruction; The difference between the actual operating parameters and the planned operating parameters is calculated to obtain the energy consumption deviation of the energy equipment; Based on the energy consumption deviation, the energy efficiency deviation coefficient of the energy equipment is calculated; Specifically, when the energy consumption deviation is positive, the energy efficiency deviation coefficient is positive; when the energy consumption deviation is negative, the energy efficiency deviation coefficient is negative.

9. The method according to claim 8, characterized in that, The first scheduling instruction is optimized and corrected based on the energy efficiency deviation coefficient to obtain a second scheduling instruction, including: For each energy device, the correction level is determined based on the absolute value of the energy efficiency deviation coefficient. For energy devices whose absolute value of the energy efficiency deviation coefficient is less than or equal to a preset threshold, it is determined to be a slight deviation. The predicted energy consumption value of the energy device in the first scheduling instruction is fine-tuned based on the proportional adjustment algorithm, wherein the fine-tuning magnitude is linearly positively correlated with the energy efficiency deviation coefficient. For energy equipment whose absolute value of the energy efficiency deviation coefficient is greater than a preset threshold, it is determined to be a significant deviation. The energy efficiency deviation coefficient of each energy equipment is introduced as a penalty term into the objective function of the airport energy optimization scheduling model, and the objective function is re-solved to obtain the corrected second scheduling instruction.

10. The method according to claim 9, characterized in that, Real-time optimized control of the energy equipment based on the second scheduling instruction includes: The second scheduling command is converted into a control signal based on industrial Ethernet; The control signals are sent to the respective sub-control systems of the airport energy management system. The gas turbine subsystem adjusts the fuel valve opening and air intake based on the corrected start-up and shutdown time and output curve. The chiller group control system adjusts the cooling tower fan speed and water pump flow based on the corrected number of operating units and load distribution ratio. The execution feedback signals of the energy equipment are collected in real time, including the actual speed of the gas turbine, the output voltage and current of the photovoltaic array, the state of charge value of the energy storage battery, and the outlet water temperature of the chiller unit. The execution feedback signal and the set value of the second scheduling instruction are compared in real time to obtain the deviation value. When the deviation value exceeds the deviation threshold, an alarm mechanism is triggered to notify the operation and maintenance personnel to intervene.

Images