Intelligent energy station scheduling method and device based on artificial intelligence, terminal and medium

Abstract

The invention relates to the field of energy dispatching, and particularly provides an intelligent energy station dispatching method and device based on artificial intelligence, a terminal and a medium, and the method comprises the steps: obtaining multi-source data of an energy station, and generating a short-term load prediction result and a medium-and-long-term load prediction result; a multi-objective optimization function is constructed with the minimum energy consumption, the minimum operation cost and the minimum carbon emission, a Pareto optimal solution set is generated through an NSGA-II algorithm, a final operation plan is selected as a scheduling scheme according to the real-time electricity price and carbon price weight, the scheduling scheme is disassembled into executable instructions, and the executable instructions are issued to a field control unit after safety certification; making a maintenance plan according to the medium-term load prediction result, and making expansion and reconstruction according to the long-term load prediction result; and continuously training and optimizing each model. According to the method, multi-source data is utilized, multi-time-scale accurate prediction is realized, multi-target collaborative optimization is carried out, and a scheduling plan is ensured to be safe and reliable.

Description

TECHNICAL FIELD

[0001] The application relates to the field of energy scheduling, in particular to an intelligent energy station scheduling method and device based on artificial intelligence, a terminal and a medium. BACKGROUND

[0002] The traditional energy station scheduling method mainly relies on simple statistical analysis of historical load data, which has limitations. In terms of load prediction, the traditional method often only uses a single time series prediction model (such as ARIMA), and fails to effectively fuse meteorological, occupancy planning, day type and other multi-source heterogeneous data, resulting in insufficient prediction accuracy, especially difficulty in accurately capturing load fluctuations caused by external factors such as sudden weather changes and special holiday activities. The prediction time scale is single, and there is a lack of coordination between short-term, medium-term and long-term prediction, which cannot meet the whole-cycle decision-making needs from real-time scheduling to strategic planning. In terms of optimization scheduling, the traditional method usually optimizes with a single objective (such as the lowest operating cost), ignoring the trade-off relationship between multiple key objectives such as energy consumption and carbon emissions. The multi-objective optimization model constructed has insufficient solving ability, and is often converted into a single-objective problem by using weighted summation and other methods, making it difficult to obtain a uniformly distributed Pareto optimal solution set, and the decision-making dimension is single, which cannot adapt to the flexible decision-making needs in complex market environments such as real-time electricity prices and dynamic carbon prices. In terms of scheme generation and execution, the traditional scheduling scheme often lacks close combination with the actual physical characteristics of the energy station, resulting in poor scheme executability. The instruction issuing process lacks strict security authentication and encryption mechanism, and there is a risk of tampering or incorrect execution, so the system reliability is difficult to guarantee. SUMMARY

[0003] To solve the above problems, the application provides an intelligent energy station scheduling method and device based on artificial intelligence, which uses multi-source data to achieve multi-time scale accurate prediction, multi-objective collaborative optimization, and ensures the safety and reliability of the scheduling plan.

[0004] In a first aspect, the technical solution of the application provides an intelligent energy station scheduling method based on artificial intelligence, comprising the following steps: Obtain multi-source data of the energy station, including drawing design load data, historical load data, meteorological data and occupancy feature data; Based on the multi-source data, use the exponential smoothing method to generate short-term load prediction results, and use a day type-based time series neural network model to generate medium and long-term load prediction results; Construct a multi-objective optimization function with minimum energy consumption, lowest operating cost and lowest carbon emission; In combination with the short-term load prediction result, multi-target strategy data and energy station design drawing data, a multi-target optimization function is used to generate a Pareto optimal solution set through an NSGA-II algorithm, and a final operation plan is selected as a scheduling scheme according to the real-time electricity price and carbon price weight; The scheduling scheme is disassembled into executable instructions and is issued to a field control unit after security authentication; A repair plan is made according to the medium-term load prediction result, and an expansion and reconstruction is made according to the long-term load prediction result; Based on system operation historical data, actual load data and multi-source data, machine learning methods are used to continuously train and optimize the load prediction algorithm model, the energy system model, the operation constraint model and the multi-target decision algorithm model.

[0005] In a second aspect, the technical solution of the present application provides an energy station intelligent scheduling device based on artificial intelligence, comprising: A multi-source data acquisition module is configured to acquire multi-source data of the energy station, including drawing design load data, historical load data, meteorological data and stationed feature data; A load prediction module is configured to generate a short-term load prediction result using an exponential smoothing method and a medium-long-term load prediction result using a time series neural network model based on the multi-source data; A multi-target optimization function construction module is configured to construct a multi-target optimization function with minimum energy consumption, lowest operation cost and lowest carbon emission; A scheduling scheme generation module is configured to combine the short-term load prediction result, multi-target strategy data and energy station design drawing data, use a multi-target optimization function, generate a Pareto optimal solution set through an NSGA-II algorithm, and select a final operation plan as a scheduling scheme according to the real-time electricity price and carbon price weight; An instruction issuing module is configured to disassemble the scheduling scheme into executable instructions and issue them to a field control unit after security authentication; A medium-long-term prediction result using module is configured to make a repair plan according to the medium-term load prediction result and make an expansion and reconstruction according to the long-term load prediction result; A continuous optimization module is configured to use machine learning methods to continuously train and optimize the load prediction algorithm model, the energy system model, the operation constraint model and the multi-target decision algorithm model based on system operation historical data, actual load data and multi-source data.

[0006] In a third aspect, the technical solution of the present application provides a terminal, comprising: A memory is configured to store an energy station intelligent scheduling program based on artificial intelligence; A processor is configured to implement the steps of the AI-based intelligent scheduling method for energy stations as described above when executing the AI-based intelligent scheduling program for energy stations.

[0007] Fourthly, the present invention provides a computer-readable storage medium storing an artificial intelligence-based intelligent scheduling program for energy stations. When the artificial intelligence-based intelligent scheduling program for energy stations is executed by a processor, it implements the steps of the artificial intelligence-based intelligent scheduling method for energy stations as described in any of the above claims.

[0008] As can be seen from the above technical solutions, this application has the following advantages: 1. This application comprehensively utilizes multi-source information such as design load from drawings, historical load, meteorological data, and resident characteristic data, and adopts differentiated prediction models for different time scales. This enables the prediction results to respond quickly to short-term fluctuations and accurately grasp medium- and long-term trends and periodic patterns, providing more accurate and reliable data for subsequent optimized scheduling. 2. This application constructs a multi-objective optimization function with the goals of minimizing energy consumption, operating costs, and carbon emissions, and uses an advanced multi-objective evolutionary algorithm such as NSGA-II to generate a Pareto optimal solution set. This overcomes the subjective drawbacks of the traditional weighted summation method, effectively weighs multiple conflicting objectives, and dynamically selects the final solution by combining real-time electricity price and carbon price weights. This ensures that the dispatch decision meets the requirements of economy, energy efficiency, and environmental protection, and achieves the comprehensive operation goals of low-carbon, economical, and efficient energy stations. 3. This application integrates the optimized scheduling scheme with the energy station design drawings to construct a characteristic model of the energy supply system as an optimization constraint, ensuring that the scheduling scheme conforms to the actual operating conditions of the system and guaranteeing the executability of the scheme. Furthermore, after the scheme is broken down into executable instructions, it must undergo security authentication before being sent to the field control unit, effectively preventing malicious tampering or misoperation of instructions during transmission and reception, thus improving the reliability and security of the entire scheduling system. 4. This application uses the medium-term load forecast results to formulate maintenance plans and the long-term load forecast results to guide expansion and renovation plans, thereby realizing intelligent management of the entire life cycle of energy station planning, construction, operation and maintenance, and improving the long-term operating efficiency of energy stations. 5. Based on historical system operation data, actual load data, and multi-source data, this application uses machine learning methods to continuously train and optimize load forecasting, energy system, operational constraints, and multi-objective decision-making models, enabling the entire system to continuously adapt to changes in the external environment and internal state, possess the ability to self-evolve and self-improve, and maintain good scheduling performance. Attached Figure Description

[0009] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of an intelligent scheduling method for energy stations based on artificial intelligence, provided as an embodiment of the present invention.

[0011] Figure 2 This is a schematic block diagram of an intelligent dispatching system for energy stations based on artificial intelligence, provided as an embodiment of the present invention.

[0012] Figure 3 This is a schematic diagram of the structure of a terminal provided in an embodiment of the present invention. Detailed Implementation

[0013] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this application and in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0015] Figure 1 This is a schematic flowchart illustrating an intelligent energy station scheduling method based on artificial intelligence, provided as an embodiment of the present invention. Figure 1 The executing entity can be an AI-based intelligent energy station scheduling system. The AI-based intelligent energy station scheduling method provided in this embodiment is executed by computer equipment; correspondingly, the AI-based intelligent energy station scheduling system runs on the computer equipment. Depending on different requirements, the order of steps in this flowchart can be changed, and some steps can be omitted.

[0016] like Figure 1 As shown, the method includes the following steps.

[0017] S1 acquires multi-source data from the energy station, including design load data, historical load data, meteorological data, and occupancy characteristic data.

[0018] S2, Based on the multi-source data, use the exponential smoothing method to generate short-term load forecast results, and use a time series neural network model based on daily types to generate medium- and long-term load forecast results.

[0019] S3 constructs a multi-objective optimization function with the goals of minimizing energy consumption, operating costs, and carbon emissions.

[0020] S4 combines short-term load forecasting results, multi-objective strategy data, and energy station design drawings. Using a multi-objective optimization function, a Pareto optimal solution set is generated through the NSGA-II algorithm. The final operation plan is selected as the scheduling scheme based on the weights of real-time electricity price and carbon price.

[0021] S5, the scheduling scheme is decomposed into executable instructions, which are then sent to the field control unit after security authentication.

[0022] S6. Develop maintenance plans based on medium-term load forecasts and expansion / renovation plans based on long-term load forecasts.

[0023] S7 uses machine learning methods to continuously train and optimize the load forecasting algorithm model, energy system model, operating constraint model, and multi-objective decision-making algorithm model based on historical system operation data, actual load data, and multi-source data.

[0024] As a refinement and extension of the specific implementation of the above embodiments, in order to fully illustrate the specific implementation process of this embodiment, the following will provide possible embodiments to describe the specific implementation of the above steps in a non-limiting manner.

[0025] In some optional implementations, when acquiring multi-source data from the energy station, for the design load data in the drawings, a file parsing interface is deployed to access the building information model (BIM) files, HVAC system design drawings, and equipment list database of the energy station. Structured data is extracted from these databases, including: design cooling / heating load values, equipment rated parameters, and pipeline topology. The equipment rated parameters include the rated capacity, rated efficiency, and upper and lower operating limits of each major piece of equipment such as chiller units, boilers, water pumps, and cooling towers.

[0026] For historical load data, deploy data acquisition interfaces to connect to the energy station's Supervisory Control and Data Acquisition (SCADA) system, smart meters, and the historical database of the Energy Management System (EMS). Extract historical time-series data according to a preset time granularity, including: total system load, sub-item load, and equipment operating status sequence.

[0027] For meteorological data, an application programming interface (API) is deployed to subscribe to real-time and forecast meteorological data provided by authoritative meteorological service providers, and data from micro-meteorological stations deployed on-site at the energy station is also integrated as a supplement. Time-series data of meteorological elements strongly correlated with energy load are extracted, including: temperature, humidity, solar radiation intensity, wind speed, and wind direction.

[0028] For occupancy characteristic data, deploy data exchange interfaces to connect to the enterprise's building automation system (BAS), access control system, and meeting room reservation management system. Acquire multi-source data from the energy station, including design load data, historical load data, meteorological data, and occupancy characteristic data.

[0029] Align all features with historical payloads by timestamp to construct a fusion dataset with one row of samples and multiple columns of features. , where the feature vector It contains all the multi-source data information at time t. This represents the actual load value at that moment, which is used for supervised learning training of the subsequent model.

[0030] In some optional implementations, step S2 involves generating short-term load forecast results based on the multi-source data using exponential smoothing, specifically including the following steps.

[0031] S2.11 Cleans, aligns, and fuses the acquired multi-source data to form a unified input feature set.

[0032] Extract historical load data ,in , indicating a historical point in time. The actual load value at time t.

[0033] Process meteorological data to extract dry-bulb temperature at the forecast time. wet-bulb temperature relative humidity Solar radiation intensity Features, and calculate the comprehensive meteorological index. (e.g., perceived temperature).

[0034] Process the occupancy feature data and extract the occupancy rate corresponding to the predicted time. population density and special event logos (Such as holidays, large conferences), this data comes from the building automation system or reservation system.

[0035] Extract the rated design load from the design load data in the drawings. and system capacity limit Physical constraints as output of the model.

[0036] S2.12 introduces meteorological and settlement characteristics as exogenous variables into the exponential smoothing framework to construct a dynamic parameter model, expressed as follows:

[0037] in, This is the load forecast value for time t+1 based on information at time t. The actual load value at time t. This represents the predicted value from the previous time step to the current time step. Let be the predicted or planned value of the j-th exogenous variable at time t+1. Exogenous variables refer to all input features extracted from multi-source data, excluding the historical load series itself. Let be the regression coefficient of the j-th exogenous variable, representing the weight of this feature on the load. Introducing exogenous variables allows the prediction model to account for sudden changes, improves prediction accuracy, and enables more accurate predictions of future loads.

[0038] The dynamic smoothing parameter is determined by the eigenvector at time t. The feature vector is obtained through mapping using the sigmoid function. It refers to the multi-source feature vector at time t, such as those containing... , wait. S2.13, using a dynamic parameter model to generate a short-term load forecast sequence for the next 24 to 72 hours. .

[0039] S2.14, The predicted value is corrected using the data from the energy station design drawings, as shown below.

[0040] in, To contribute the minimum technical effort to the system, This is the final, short-term load forecast that conforms to the system characteristics.

[0041] S2.15, the final corrected prediction sequence As a result of short-term load forecasting.

[0042] When training a dynamic parameter model, the mean squared error (MSE) is used as the loss function, and the model parameters are jointly optimized using either gradient descent or maximum likelihood estimation.

[0043] In some alternative implementations, step S2 uses a day-type time series neural network model to generate medium- to long-term load forecast results, specifically including the following steps.

[0044] S2.21 Clean and align the acquired multi-source data, and define day type labels.

[0045] Extract historical load data ,in , indicating a historical point in time. The actual load value at time t.

[0046] Based on onboarding characteristic data, holiday calendars, and historical load patterns, a day type label is assigned to each historical date. The label is a One-Hot encoded vector, for example... ∈{regular weekdays, weekends, public holidays, special event days}.

[0047] Extracting historical daily average temperatures from meteorological data Daily maximum temperature Daily minimum temperature and sunshine duration As a daily-level feature.

[0048] Extract the design daily load curve from the design load data in the drawings. Reference baseline features are used as input to the model.

[0049] S2.22, Define the model input feature matrix For the target predicted date d, it is represented as,

[0050] in, This is a historical daily load sequence over the past N days. This is a sequence of daily type labels corresponding to the past N days. This is a historical daily meteorological feature sequence over the past N days. This is a sequence of historical occupancy rates over the past N days. The daily load characteristics are designed as inputs for the system's static attributes.

[0051] Data is organized on a daily basis to construct supervised learning samples suitable for medium- to long-term forecasting. The input of each sample is multi-source time series features from the past N days, and the output is the forecast target for the next H days (medium-term) or from the Hth month to the H+Mth month (long-term).

[0052] S2.23, Define the prediction target Including medium-term forecasts and long-term forecasts ,in This represents the average monthly load or total monthly load for the m-th month in the future.

[0053] S2.24, Construct and train a neural network model with an encoder-decoder structure. This model uses day type labels as keys and values ​​in the attention mechanism to capture differences in load patterns under different day types.

[0054] Encoder: Composed of a Long Short-Term Memory (LSTM) network or a Gated Recurrent Unit (GRU), used to encode the hidden state of the input sequence.

[0055] The day type attention layer calculates the correlation weights between the current state of the decoder and the historical day type encodings, and generates a context vector.

[0056] Decoder: Composed of another RNN network, its inputs are the final state of the encoder, the daily type context vector, and known features (such as weather forecasts and planned occupancy rates), which gradually generate a prediction sequence.

[0057] S2.25, Input the latest sequence Based on future daily load plans, the neural network model outputs daily load forecasts for the next few weeks. .

[0058] S2.26, using recursive prediction, input the latest sequence Assuming the exogenous variable is the historical average for the same period, the neural network model outputs the monthly average load forecast for the next few months to several years. .

[0059] In some alternative implementations, step S3 constructs a multi-objective optimization function with the goals of minimizing energy consumption, operating costs, and carbon emissions, specifically including the following steps.

[0060] S3.1 Define the operational decision variables for each device within the scheduling cycle. ,in, Let T be the output vector of each traditional generator during time period T. This is the unit start-up and shutdown status vector within time period T. This represents the output vector of the CHP unit during time period T. Let T be the energy storage charge / discharge state vector during time period T. This represents the power vector purchased from the grid during time period T.

[0061] S3.2, Construct the objective of minimizing the total energy consumption of the system, expressed as,

[0062] in, , , Let be the energy consumption coefficient of the i-th conventional generator. , Let be the energy consumption coefficient of the j-th CHP unit. For the number of traditional generators, This refers to the number of CHP units. Let be the active power output of the i-th conventional generator during time period t. Let represent the start / stop status of the i-th conventional generator during time period t. The active power output of the j-th CHP unit during time period t.

[0063] S3.3, Construct the objective of minimizing the total operating cost of the system, expressed as,

[0064]

[0065] in, The operating and maintenance cost of the i-th traditional generator during time period t is... The operating and maintenance cost of the j-th CHP unit during time period t is... Let t be the cost of purchasing electricity from an external power grid during time period t. The starting cost of all conventional generators during time period t. Let be the unit power operation and maintenance cost coefficient of the i-th device. The real-time electricity price for time period t. For demand-based electricity pricing, Let $\frac{i}{i}$ be the startup cost of the $i$-th generating unit.

[0066] S3.4, Construct the objective of minimizing the total carbon emissions of the system, expressed as,

[0067] in, The carbon emission intensity of the i-th conventional generator, The carbon emission intensity of the j-th CHP unit is... Carbon emission intensity of purchasing electricity from the grid Let t represent the power purchased from the grid during time period t.

[0068] S3.5, Construct a unified multi-objective optimization function, expressed as follows: .

[0069] in, Let be the j-th inequality constraint function, and m be the total number of inequality constraints. Let be the k-th equality constraint function, and p be the total number of equality constraints.

[0070] Inequality constraints include: Equipment output upper and lower limits , , respectively, indicate that the output force must not be lower than the minimum value and the output force must not be higher than the maximum value; Slope rate constraint This means that the change in output between adjacent time periods must not exceed the maximum gradient rate; Energy storage SOC constraints , , respectively, indicate that the energy storage state of charge must not be lower than the lower limit and the energy storage state of charge must not be higher than the upper limit; Power grid interaction constraints This means that the power purchased from the power grid must not exceed the line capacity.

[0071] Equality constraints include: Power balance constraints,

[0072] In the formula, The discharge power of the stored energy during time period t. This represents the predicted combined load for time period t, i.e., the total load demand that needs to be met. The charging power of energy storage during time period t. The load power reduction during time period t.

[0073] This means that the total power output by all power sources must be equal to the total power consumed by all loads; Energy storage dynamic equations

[0074] In the formula, For energy storage charging efficiency, For the discharge efficiency of energy storage, This represents the state of charge (remaining energy) at the end of time period t.

[0075] This means that the energy storage status in this period must be equal to the status in the previous period plus the net energy input in this period.

[0076] It should be noted that the operating constraints and efficiency parameters of the equipment are provided by the energy supply system characteristic model constructed based on the energy station design drawings.

[0077] In some optional implementations, step S4 combines short-term load forecast results, multi-objective strategy data, and energy station design drawing data, uses a multi-objective optimization function, generates a Pareto optimal solution set through the NSGA-II algorithm, and selects the final operation plan as the scheduling scheme based on the weights of real-time electricity price and carbon price. Specifically, it includes the following steps.

[0078] Multi-objective strategy data includes real-time electricity price data, carbon price data, and demand response incentive signals. Energy station design drawings are used to construct a characteristic model of the energy supply system, including equipment capacity constraints, pipeline topology, and energy conversion efficiency characteristics.

[0079] S4.1, acquire multi-objective strategy data, including real-time electricity price series, real-time carbon price series, and demand response incentive signals.

[0080] S4.2 Analyze the energy station design drawing data, extract and construct optimization constraints, including equipment capacity constraints, pipeline topology, and energy conversion efficiency characteristics.

[0081] S4.3, Define the decision variable vector .

[0082] S4.4, Construct core system operation constraints, including power balance constraints that depend on short-term load forecast sequences.

[0083] S4.5 sets the algorithm parameters, including population size, maximum number of iterations, crossover probability, and mutation probability.

[0084] S4.6, randomly generate the initial population, where each individual represents the encoding of a decision variable vector.

[0085] S4.7, repeat the following steps until the maximum number of iterations is reached: For each individual in the population, i.e., the scheduling scheme, calculate its three objective function values. During the calculation, check whether the current scheme satisfies all constraints. Individuals that violate the constraints will be penalized or directly eliminated. For all individuals in the current population, perform non-dominated sorting based on their objective function values ​​and divide them into multiple non-dominated layers F1, F2, ... For individuals in the same non-dominated layer, calculate the crowding distance of each individual in each objective dimension to measure its distribution density in the objective space. Use the binary tournament selection operator, simulated binary crossover operator, and polynomial mutation operator to generate the offspring population. Merge the parent population and the offspring population into Rt. Perform non-dominated sorting and crowding calculation on Rt. Select individuals from F1, F2, ... in sequence until the new population is filled.

[0086] S4.8 After the algorithm converges, it outputs all individuals in the first non-dominated layer F1, forming the Pareto optimal solution set. .

[0087] S4.9, Calculate each Pareto solution The corresponding three objective function values .

[0088] S4.10, Based on real-time electricity price and carbon price data, calculate the dynamic weight of each objective for the current period, expressed as follows:

[0089]

[0090] in, and These are the average electricity price and average carbon price during the scheduling period, respectively, with δ being a positive number used to prevent the denominator from being zero.

[0091] S4.11, Construct each Pareto solution Comprehensive utility function ,

[0092] in , and , These represent the minimum and maximum values ​​of the Pareto solution set's operating cost and carbon emission target, respectively.

[0093] S4.12, Select the value with the highest overall utility. individual As the final operational plan, the final decision variables will be... This is interpreted as a specific scheduling scheme.

[0094] The above steps unify the three major categories of factors—forecasting, market, and physics—within an optimization framework, making decision-making more comprehensive. By solving the Pareto front, the competitive relationship between different objectives is revealed, and decision-makers are given a choice space rather than a single outcome. By dynamically adjusting the decision weights through real-time electricity prices and carbon prices, the dispatch scheme can flexibly adapt to market changes and maximize economic or environmental benefits.

[0095] Step S5 involves breaking down the scheduling scheme into executable instructions, which are then sent to the field control unit after security authentication. This step includes the following steps.

[0096] S5.1 decomposes the finalized scheduling scheme into hourly sub-plans according to the time series.

[0097] S5.2 breaks down the operation instructions for each sub-plan to generate an executable instruction set, which includes start / stop instructions for each device, power setting values ​​for each device, and charging / discharging power instructions for the energy storage device.

[0098] S5.3 categorizes the executable instruction set according to device type and control unit to form instruction subsets for field control units.

[0099] S5.4. Perform security authentication on each instruction subset. After successful authentication, perform digital signature to generate signature information. The signature information is used to encrypt the instruction subset using a preset encryption algorithm and key.

[0100] Specifically, a safety check is first performed on each instruction subset, including: checking whether the device start-stop instructions comply with the device operation constraints (such as minimum start-stop time, device status, etc.); checking whether the power set value is within the device's allowable operating range; and checking whether the energy storage device's charging and discharging power instructions comply with the energy storage device's state of charge and charging / discharging rate limits.

[0101] If the security check passes, the subset of instructions is digitally signed.

[0102] S5.5 packages a subset of instructions with signature information into an authentication instruction package.

[0103] S5.6 sends the authentication instruction packet to the corresponding field control unit through a preset communication protocol and network.

[0104] In some alternative implementations, step S6 involves developing a maintenance plan based on the medium-term load forecast results, specifically including the following steps.

[0105] S6.11 identifies valleys and analyzes stability of the medium-term load forecast curve, and uses a peak-valley detection algorithm to find the time window when the load is low and the load changes are gradual.

[0106] The following conditions must be met during periods of low load:

[0107] in, This is the load threshold coefficient. This is the average value of the medium-term forecast load. This is the load fluctuation threshold.

[0108] S6.12, Establish a maintenance plan optimization model with the main objectives of maximizing system reliability and minimizing load loss during maintenance. The constraints of the maintenance plan optimization model include equipment maintenance time constraints, maintenance resource constraints, and system reliability constraints.

[0109] The objective function is expressed as follows:

[0110] in, Let t be the system's reserve capacity for time period t. This represents the minimum required standby capacity for the system.

[0111] S6.13 uses a heuristic algorithm or integer programming solver to solve the maintenance plan optimization model and outputs the optimal maintenance plan.

[0112] The maintenance plan is expressed as .

[0113] Step S6 involves formulating expansion and renovation plans based on long-term load forecast results, specifically including the following steps.

[0114] S6.21 compares the long-term load forecast results with the existing system capacity of the energy station to identify capacity gaps.

[0115] S6.22, when the time-series load forecast value continuously exceeds the system capacity safety threshold, triggers a capacity expansion demand signal.

[0116] S6.23, after recognizing the expansion requirement, provides prompts for the expansion and renovation plan formulation process.

[0117] The development of expansion and renovation plans can include multi-scenario prediction, technical and economic comparison, generation of recommended solutions, and development of implementation roadmaps.

[0118] Multi-scenario forecasting considers different future development scenarios and uses long-term load forecasting models to generate load growth paths under different scenarios. Technical and economic comparisons are conducted based on the forecast results of different scenarios to construct a set of expansion options, and life-cycle cost (LCC) analysis and payback period indicators are used to evaluate each option. Recommended options are generated using a multi-criteria decision-making method, comprehensively considering investment costs, operating costs, carbon emission reduction benefits, and reliability indicators to select recommended options from the set. An implementation roadmap is developed based on the load growth sequence revealed by long-term load forecasts, breaking down the recommended expansion and renovation options into phased milestone plans.

[0119] Continuous training and optimization are performed on the load forecasting algorithm model, energy system model, operational constraint model, and multi-objective decision-making algorithm model. The load forecasting algorithm model includes exponential smoothing and a day-type-based neural network model. The energy system model refers to the characteristic model of the energy supply system constructed from energy station design drawings. The operational constraint model refers to the implicit constraints on equipment, pipelines, and balance when constructing the multi-objective function. The multi-objective decision-making algorithm model refers to the NSGA-II algorithm and the scheme selection decision rule based on real-time electricity price / carbon weight.

[0120] The above text describes in detail an embodiment of an AI-based intelligent scheduling method for energy stations. Based on the AI-based intelligent scheduling method for energy stations described in the above embodiment, this invention also provides an AI-based intelligent scheduling device for energy stations corresponding to the method.

[0121] Figure 2This is a schematic block diagram of an intelligent energy station dispatching device based on artificial intelligence, provided as an embodiment of the present invention. In this embodiment, the intelligent energy station dispatching device 200 based on artificial intelligence can be divided into multiple functional modules according to the functions it performs. A module, as referred to in this invention, is a series of computer program segments that can be executed by at least one processor and perform a fixed function, and is stored in memory.

[0122] The multi-source data acquisition module 210 is used to acquire multi-source data of the energy station, including design load data, historical load data, meteorological data and resident characteristic data.

[0123] The load forecasting module 220 is used to generate short-term load forecasting results using exponential smoothing based on the multi-source data, and to generate medium- and long-term load forecasting results using a time series neural network model based on daily data types.

[0124] The multi-objective optimization function construction module 230 is used to construct multi-objective optimization functions with the goals of minimizing energy consumption, operating costs, and carbon emissions.

[0125] The scheduling scheme generation module 240 is used to combine short-term load forecast results, multi-objective strategy data and energy station design drawing data, use a multi-objective optimization function, generate a Pareto optimal solution set through the NSGA-II algorithm, and select the final operation plan as the scheduling scheme based on the weight of real-time electricity price and carbon price.

[0126] The instruction issuing module 250 is used to decompose the scheduling scheme into executable instructions, which are then sent to the field control unit after security authentication.

[0127] The medium- and long-term forecast results are used in module 260 to develop maintenance plans based on medium-term load forecast results and to develop expansion and renovation plans based on long-term load forecast results.

[0128] The continuous optimization module 270 is used to continuously train and optimize the load forecasting algorithm model, energy system model, operating constraint model and multi-objective decision-making algorithm model based on historical system operation data, actual load data and multi-source data, using machine learning methods.

[0129] The AI-based intelligent energy station scheduling device of this embodiment is used to implement the aforementioned AI-based intelligent energy station scheduling method. Therefore, the specific implementation of this device can be found in the embodiment section of the AI-based intelligent energy station scheduling method above. Thus, the specific implementation can be referred to the description of the corresponding embodiments, and will not be elaborated here.

[0130] Furthermore, since the AI-based intelligent energy station scheduling device in this embodiment is used to implement the aforementioned AI-based intelligent energy station scheduling method, its function corresponds to the function of the above method, and will not be repeated here.

[0131] Figure 3 This is a schematic diagram of the structure of a terminal 300 provided in an embodiment of the present invention, including: a processor 310, a memory 320, and a communication unit 330. The processor 310 is used to implement the flow steps of an embodiment of an AI-based intelligent energy station scheduling method when implementing the AI-based intelligent energy station scheduling program stored in the memory 320.

[0132] The terminal 300 includes a processor 310, a memory 320, and a communication unit 330. These components communicate via one or more buses. Those skilled in the art will understand that the server structure shown in the figure does not constitute a limitation of the present invention. It can be a bus topology or a star topology, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0133] The memory 320 can be used to store the execution instructions of the processor 310. The memory 320 can be implemented by any type of volatile or non-volatile memory terminal or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. When the execution instructions in the memory 320 are executed by the processor 310, the terminal 300 is able to perform some or all of the steps in the above method embodiments.

[0134] The processor 310 serves as the control center of the storage terminal, connecting various parts of the electronic terminal via various interfaces and lines. It executes software programs and / or modules stored in the memory 320, and calls data stored in the memory to perform various functions of the electronic terminal and / or process data. The processor can be composed of integrated circuits (ICs), such as a single packaged IC or multiple packaged ICs with the same or different functions connected together. For example, the processor 310 may consist only of a central processing unit (CPU). In this embodiment of the invention, the CPU may have a single processing core or include multiple processing cores.

[0135] The communication unit 330 is used to establish a communication channel, enabling the storage terminal to communicate with other terminals. It can receive user data sent by other terminals or send user data to other terminals.

[0136] The present invention also provides a computer storage medium, which may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0137] The present invention also provides a computer storage medium, which may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0138] The computer storage medium stores an AI-based intelligent scheduling program for energy stations. When the AI-based intelligent scheduling program is executed by the processor, it implements the process steps of an AI-based intelligent scheduling method embodiment for energy stations.

[0139] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium such as a USB flash drive, mobile hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, or other media capable of storing program code. It includes several instructions to cause a computer terminal (which may be a personal computer, server, or a second terminal, network terminal, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0140] In the embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0141] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0142] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0143] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An artificial intelligence-based energy station intelligent scheduling method, characterized in that, The method comprises the following steps: Obtaining multi-source data of the energy station, including design load data, historical load data, meteorological data and occupancy feature data; Based on the multi-source data, using exponential smoothing method to generate short-term load prediction results, using day type based time series neural network model to generate medium and long term load prediction results; Constructing a multi-objective optimization function with minimum energy consumption, lowest operation cost and lowest carbon emission; Combining short-term load prediction results, multi-objective strategy data and energy station design drawing data, using multi-objective optimization function, generating Pareto optimal solution set through NSGA-II algorithm, selecting final operation plan as scheduling scheme according to real-time electricity price and carbon price weight; Decomposing the scheduling scheme into executable instructions and issuing them to the on-site control unit after security authentication; Formulating maintenance plan according to medium-term load prediction results and expansion and reconstruction according to long-term load prediction results; Based on system operation history data, actual load data and multi-source data, using machine learning method to continuously train and optimize load prediction algorithm model, energy system model, operation constraint model and multi-objective decision algorithm model. 2.The AI-based energy station intelligent scheduling method according to claim 1, characterized in that, Based on the multi-source data, using exponential smoothing method to generate short-term load prediction results, specifically including: Cleaning, aligning and fusing the obtained multi-source data to form a unified input feature set; Introducing meteorological and occupancy features as exogenous variables into the exponential smoothing framework to construct a dynamic parameter model, represented as, wherein, is the load forecast value at time t+1 based on the information at time t, is the actual load value at time t, is the forecast value at the previous time for the current time, is the forecast value or planned value of the jth exogenous variable at time t+1, the exogenous variable refers to all input features extracted from multi-source data in addition to the historical load sequence itself, is the regression coefficient of the jth exogenous variable, indicating the weight of the feature on the load. is a dynamic smoothing parameter, whose value is determined by the eigenvector of time t is obtained by mapping through a sigmoid function, and the eigenvector is a multi-source eigenvector at time t Generating short-term load forecast sequences for the next 24 to 72 hours using dynamic parameter models ; Using energy station design drawing data to correct the predicted value, represented as, wherein, is the minimum technical power of the system, is the final published short-term load forecast value, consistent with the system characteristics; The final revised predicted value sequence as a short-term load prediction result. 3.The AI-based energy station intelligent scheduling method of claim 1, wherein, Using day type based time series neural network model to generate medium and long term load prediction results, specifically including: Cleaning and aligning the obtained multi-source data and defining day type labels; Defining model input feature matrix For a target prediction day d, denoted as wherein, is the historical daily load sequence of the past N days, is the corresponding daily type label sequence of the past N days, is the historical daily level meteorological feature sequence of the past N days, is the historical occupancy rate sequence of the past N days, is the design daily load feature as the system static attribute input; Defining the prediction target , including medium-term prediction and long-term prediction , wherein is the monthly average load or monthly total of the future mth month; Building and training an encoder-decoder structure neural network model, which takes day type labels as keys and values in attention mechanism to capture load mode differences under different day types; Input latest sequence and future day type plan, the neural network model outputs daily load prediction values within future weeks ; Recursive prediction is used, with the latest sequence as input And assuming the exogenous variables are the historical same-period average, the neural network model outputs the monthly average load prediction value for the next few months to a few years . 4.The AI-based energy station intelligent scheduling method according to claim 1, characterized in that, Constructing a multi-objective optimization function with minimum energy consumption, lowest operation cost and lowest carbon emission, specifically including: Defining the operating decision variables of each device in the scheduling period wherein, is the output vector of each conventional generator in the T period, is the unit start-stop state vector in the T period, is the output vector of each CHP unit in the T period, is the charge-discharge state vector of each energy storage in the T period, is the power purchase vector from the grid in the T period; Constructing a system total energy consumption minimization objective, represented as, wherein, , , is the energy consumption coefficient of the i-th conventional generator, , is the energy consumption coefficient of the j-th CHP unit, is the number of conventional generators, is the number of CHP units, is the active power output of the i-th conventional generator at time period t, is the start-stop state of the i-th conventional generator at time period t, is the active power output of the j-th CHP unit at time period t; Constructing a system total operation cost minimization objective, represented as, wherein, is the operation and maintenance cost of the ith conventional generator at time period t, is the operation and maintenance cost of the jth CHP unit at time period t, is the cost of purchasing electricity from the external grid at time period t, is the start-up cost of all conventional generators at time period t, is the operation and maintenance cost coefficient of the ith unit per unit power, is the real-time electricity price at time period t, is the demand charge, is the start-up cost of the ith unit. Constructing a system total carbon emission minimization objective, represented as, wherein, is the carbon emission intensity of the ith conventional generator, is the carbon emission intensity of the jth CHP unit, is the carbon emission intensity of grid electricity purchase, is the power purchased from the grid at time t. Constructing a unified multi-objective optimization function, represented as, wherein, is the jth inequality constraint function, and m is the total number of inequality constraints, is the kth equality constraint function, and p is the total number of equality constraints. 5.The AI-based energy station intelligent scheduling method according to claim 1, characterized in that, Combining short-term load prediction results, multi-objective strategy data and energy station design drawing data, using multi-objective optimization function, generating Pareto optimal solution set through NSGA-II algorithm, selecting final operation plan as scheduling scheme according to real-time electricity price and carbon price weight, specifically including: Obtaining multi-objective strategy data, including real-time electricity price sequence, real-time carbon price sequence and demand response incentive signal; Analyzing energy station design drawing data, extracting and constructing optimization constraint conditions, including device capacity constraint, pipe network topology structure and energy conversion efficiency characteristics; Defining the decision variable vector ; Constructing core system operation constraints, including power balance constraint depending on short-term load prediction sequence; Setting algorithm parameters, including population size, maximum iteration number, crossover probability and mutation probability; Randomly generating initial population, each individual representing a code of decision variable vector; The following steps are executed in cycles until the maximum number of iterations is reached: for each individual in the population, i.e., the scheduling scheme, calculate its three objective function values, and in the process, check whether the current scheme satisfies all constraints, and individuals that violate the constraints will be penalized or directly eliminated; for all individuals in the current population, perform non-dominated sorting according to their objective function values, and divide them into multiple non-dominated layers F1, F2,...; for individuals in the same non-dominated layer, calculate the crowding distance of each individual in each objective dimension to measure its distribution density in the objective space, generate a child population using a binary tournament selection operator, a simulated binary crossover operator and a polynomial mutation operator, and combine the parent population and the child population into Rt, and perform non-dominated sorting and crowding calculation on Rt, and sequentially select individuals from F1, F2,... until the new population is filled; After the algorithm converges, output all individuals in the first non-dominated layer F1 to form the Pareto optimal solution set ; Compute each Pareto solution corresponding three objective function values ; According to the real-time electricity price and carbon price data, the dynamic weight of each target in the current period is calculated, which is represented as, wherein, and are the average electricity price and average carbon price within the dispatch period, respectively, and δ is a positive number used to prevent the denominator from being zero; constructing a comprehensive utility function for each pareto solution ,​ wherein , and , are the minimum and maximum values of the operating cost and carbon emission target in the Pareto solution set, respectively; selecting the individual having the highest overall utility value as the final operational plan resolving the final decision variables into a specific scheduling scheme. 6.The AI-based energy station intelligent scheduling method according to claim 1, characterized in that, The scheduling scheme is disassembled into executable instructions, and after security authentication, it is issued to the field control unit, specifically including: The final determined scheduling scheme is decomposed into hourly sub-plans according to the time sequence; For each sub-plan, operation instruction disassembly is performed to generate an executable instruction set, which includes start-stop instructions of each device, power set values of each device, and charge-discharge power instructions of energy storage devices; The executable instruction set is classified according to device type and control unit to form an instruction subset for the field control unit; Each instruction subset is subjected to security authentication, and after authentication, it is digitally signed to generate signature information, which is encrypted by a preset encryption algorithm and key to process the instruction subset; The instruction subset with signature information is packaged into an authentication instruction package; The authentication instruction package is sent to the corresponding field control unit through a preset communication protocol and network. 7.The AI-based energy station intelligent scheduling method according to claim 1, characterized in that, According to the medium-term load forecasting result, a maintenance plan is formulated, specifically including: Valley identification and stationarity analysis are performed on the medium-term load forecasting curve, and a peak-valley detection algorithm is used to find the load valley period and the time window with smooth load change; A maintenance plan optimization model is established, with the highest system reliability and the smallest load loss during maintenance as the main targets; the constraint conditions of the maintenance plan optimization model include device maintenance time constraints, maintenance resource constraints, and system reliability constraints; A heuristic algorithm or integer programming solver is used to solve the maintenance plan optimization model, and the optimal maintenance plan is output; According to the long-term load forecasting result, an expansion and reconstruction is formulated, specifically including: The long-term load forecasting result is compared with the existing system capacity of the energy station to identify the capacity gap; When the time-series load forecasting value continuously exceeds the system capacity safety threshold, an expansion demand signal is triggered; When the expansion demand is identified, an expansion and reconstruction scheme formulation process prompt is given.

8. An artificial intelligence-based energy station intelligent scheduling device, characterized in that, It includes: A multi-source data acquisition module is used to acquire multi-source data of the energy station, including design load data, historical load data, meteorological data and resident feature data; a load prediction module configured to generate short-term load prediction results using an exponential smoothing method and medium and long-term load prediction results using a time series neural network model based on the multi-source data; a multi-objective optimization function construction module configured to construct a multi-objective optimization function with the minimum energy consumption, the lowest operation cost, and the lowest carbon emission; a scheduling scheme generation module configured to generate a Pareto optimal solution set by using the multi-objective optimization function and the NSGA-II algorithm based on the short-term load prediction results, multi-objective strategy data, and energy station design drawing data, and select a final operation plan as a scheduling scheme according to real-time electricity price and carbon price weights; an instruction issuing module configured to decompose the scheduling scheme into executable instructions and issue the executable instructions to a field control unit after security authentication; a medium and long-term prediction result using module configured to develop a maintenance plan according to the medium-term load prediction results and develop an expansion and reconstruction plan according to the long-term load prediction results; a continuous optimization module configured to continuously train and optimize a load prediction algorithm model, an energy system model, an operation constraint model, and a multi-objective decision algorithm model based on system operation history data, actual load data, and multi-source data by using a machine learning method.

9. A terminal, characterized by comprising: comprise: a memory configured to store an artificial intelligence-based energy station intelligent scheduling program; a processor configured to implement the steps of the artificial intelligence-based energy station intelligent scheduling method according to any one of claims 1 to 7 when the artificial intelligence-based energy station intelligent scheduling program is executed.

10. A computer-readable storage medium, characterized in that, The readable storage medium has an artificial intelligence-based energy station intelligent scheduling program stored thereon, and the steps of the artificial intelligence-based energy station intelligent scheduling method according to any one of claims 1 to 7 are implemented when the artificial intelligence-based energy station intelligent scheduling program is executed by a processor.

Images