Energy efficiency optimization scheduling method of charging module based on big data prediction

Abstract

This invention discloses a charging module energy efficiency optimization scheduling method based on big data prediction, comprising: collecting multi-source heterogeneous data related to charging load and grid status, performing preprocessing and feature fusion to generate a spatiotemporally correlated fused dataset; training and applying a big data prediction model based on the fused dataset to output regional charging demand prediction results and grid load prediction results within the future scheduling cycle; inputting the regional charging demand prediction results, grid load prediction results, and renewable energy output prediction data into a collaborative scheduling optimization model established for multiple charging modules; solving the collaborative scheduling optimization model to obtain an optimized scheduling instruction sequence for the start / stop status and output power of each charging module within the future scheduling cycle, so as to execute start / stop control and power allocation for each charging module. This invention can achieve improved charging energy efficiency, extended equipment life, and increased renewable energy consumption rate.

Description

Technical Field

[0001] This invention relates to the fields of charging energy efficiency optimization and smart grid scheduling technology, and in particular to a charging module energy efficiency optimization scheduling method based on big data prediction. Background Technology

[0002] With the rapid development of the new energy vehicle industry and the continuous expansion of charging infrastructure, the energy efficiency optimization and scheduling of charging modules has become a key technology for ensuring stable grid operation and improving energy utilization efficiency. Currently, charging module scheduling mainly adopts a passive response strategy based on real-time load, lacking the ability to accurately predict future charging demand. This makes it difficult to effectively address the challenges of randomness in electric vehicle user behavior, grid load fluctuations, and uncertainties brought about by renewable energy grid integration. Consequently, charging facilities experience high idle rates during off-peak hours and overload operation during peak hours, restricting energy utilization efficiency and equipment lifespan. Furthermore, existing scheduling methods often rely on fixed thresholds or simple rules for local optimization, failing to fully integrate multi-source heterogeneous data such as historical charging data, grid load information, meteorological conditions, and user behavior characteristics. They lack the ability for collaborative scheduling between charging modules and global energy efficiency optimization, resulting in limited effectiveness in peak shaving, valley filling, reducing grid losses, and improving renewable energy absorption rates. This makes it difficult to meet the refined and intelligent management needs of large-scale charging networks in the context of smart grids.

[0003] Therefore, there is an urgent need to provide a technical solution to address the above problems. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a charging module energy efficiency optimization scheduling method based on big data prediction. The technical solution of this method is as follows:

[0005] Collect multi-source heterogeneous data related to charging load and power grid status, including historical charging data, real-time power grid load data, weather forecast data, and user charging behavior characteristics;

[0006] The multi-source heterogeneous data is preprocessed and feature fused to generate a spatiotemporally correlated fused dataset;

[0007] Based on the fused dataset, a big data prediction model is trained and applied to output the regional charging demand prediction results and the power grid load prediction results within the future scheduling cycle.

[0008] The regional charging demand forecast, the power grid load forecast, and the renewable energy output forecast are input into a collaborative scheduling optimization model established for multiple charging modules. The collaborative scheduling optimization model aims to smooth out fluctuations in the total power grid load, minimize total operating costs, and maximize the renewable energy absorption rate.

[0009] Solving the cooperative scheduling optimization model yields an optimized scheduling instruction sequence for the start / stop status and output power of each charging module in the future scheduling cycle;

[0010] Based on the optimized scheduling instruction sequence for each charging module, start / stop control and power allocation are performed for each charging module.

[0011] The technical solution of this invention collects and integrates multi-source information such as historical charging data, real-time grid load, weather forecasts, and user behavior characteristics to train a big data prediction model to obtain future charging demand and grid load. It combines renewable energy output to construct a multi-charging module collaborative scheduling optimization model to solve start-stop and power commands. This solves the problem that traditional passive response strategies lack accurate prediction and global coordination capabilities, overcomes the peak-valley contradiction caused by the randomness of user behavior and the uncertainty of renewable energy, and achieves improved charging energy efficiency, extended equipment life, and increased renewable energy consumption rate.

[0012] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0014] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0015] Figure 1 This is a flowchart illustrating an embodiment of the charging module energy efficiency optimization scheduling method based on big data prediction according to the present invention. Detailed Implementation

[0016] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0017] Figure 1 The diagram illustrates a flowchart of an embodiment of a charging module energy efficiency optimization scheduling method based on big data prediction provided by the present invention, executed by a control terminal. Figure 1As shown, it includes the following steps:

[0018] S1. Collect multi-source heterogeneous data related to charging load and power grid status. The multi-source heterogeneous data includes historical charging data, real-time power grid load data, weather forecast data, and user charging behavior characteristics.

[0019] S2. Preprocess and feature fusion of the multi-source heterogeneous data to generate a spatiotemporally correlated fusion dataset;

[0020] S3. Based on the fused dataset, train and apply the big data prediction model to output the regional charging demand prediction results and power grid load prediction results within the future scheduling cycle.

[0021] S4. Input the regional charging demand forecast, the power grid load forecast, and the renewable energy output forecast into the collaborative scheduling optimization model established for multiple charging modules; wherein, the collaborative scheduling optimization model aims to smooth power grid load fluctuations, minimize total operating costs, and maximize renewable energy absorption rate.

[0022] S5. Solve the cooperative scheduling optimization model to obtain the optimized scheduling instruction sequence of the start / stop state and output power of each charging module in the future scheduling cycle;

[0023] S6. Based on the optimized scheduling instruction sequence for each charging module, perform start-stop control and power allocation for each charging module.

[0024] The technical solution of this embodiment collects and integrates multi-source information such as historical charging data, real-time grid load, weather forecasts and user behavior characteristics to train a big data prediction model to obtain future charging demand and grid load. Combined with renewable energy output, a multi-charging module collaborative scheduling optimization model is constructed to solve the start-stop and power commands. This solves the problem that traditional passive response strategies lack accurate prediction and global coordination capabilities, overcomes the peak-valley contradiction caused by the randomness of user behavior and the uncertainty of renewable energy, and achieves improved charging energy efficiency, extended equipment life and increased renewable energy consumption rate.

[0025] In one alternative approach, S1 specifically includes:

[0026] The historical charging data and user charging behavior characteristics related to the charging load are obtained from the charging pile data platform.

[0027] The charging pile data platform refers to a software system used to access, manage, and store operational data and user interaction data of charging infrastructure. For example, all DC charging piles in a given area upload their charging records, power curves, and user identification information to the operator's cloud data platform. Charging load refers to the total electrical energy demand generated by electric vehicle charging behavior, expressed as a power-time curve. For example, the total power required by all vehicles charging in a given area during the evening peak hours reaches a certain value. Historical charging data refers to a collection of detailed information about each charging process recorded in the charging pile data platform over a past period. For example, the platform stores the charging start time, end time, electricity consumption, and charging power change curves for each charging pile in a given area over the past few months. User charging behavior characteristics refer to quantitative indicators extracted from historical charging data that reflect users' charging habits and preferences. For example, analysis of historical data reveals that a significant proportion of users in a given area tend to start charging immediately upon arrival, with an average charging time of several minutes.

[0028] The real-time grid load data related to the grid status is obtained from the grid dispatching system.

[0029] The power grid dispatching system refers to the automated system used by power companies to monitor, control, and optimize power grid operation; for example, a local power grid dispatching system collects and displays real-time voltage, current, and power data at the outlets of substations in a certain area. Power grid status refers to the set of electrical quantities describing the operation of the power grid at a specific moment; for example, the status of a regional power grid at a certain moment includes the bus voltage at a specific voltage level, system frequency, and total load. Real-time power grid load data refers to time-series data provided by the power grid dispatching system that reflects the latest changes in the total power consumption of the power grid; for example, current total power consumption data for a certain area obtained from the dispatching system at fixed time intervals.

[0030] Obtain the meteorological forecast data for the corresponding geographical area and future scheduling cycle from the meteorological data interface.

[0031] The meteorological data interface refers to an application programming interface (API) that provides programmatic access to meteorological forecast data services; for example, by calling the API of a standard meteorological data publishing agency, it obtains temperature, humidity, and weather conditions data for a specific future period in the target area. The geographical scope refers to a specific area with clearly defined spatial boundaries that serves as the object of scheduling analysis; for example, the geographical scope targeted by this scheduling analysis is a specific area with an area of ​​approximately several square kilometers. The future scheduling cycle refers to a time interval from the current moment onwards where optimized scheduling of charging modules is planned; for example, the future scheduling cycle is set to run from midnight to midnight the following day, totaling 24 hours. Meteorological forecast data refers to meteorological condition prediction information about a specific future period and geographical scope obtained through the meteorological data interface; for example, the obtained meteorological forecast data shows that the weather in the target area will be sunny turning cloudy at a certain time in the future, with a maximum temperature of a certain degree Celsius.

[0032] Among the above-mentioned optional methods, the specific sources of multi-source heterogeneous data collection are further clarified. By standardizing the acquisition of data from charging pile platforms, power grid dispatching, and meteorological interfaces, the data collection process is standardized and engineered, thereby improving the reliability of data sources and the convenience of solution implementation.

[0033] In one alternative approach, S2 specifically includes:

[0034] The historical charging data, user charging behavior characteristics, real-time power grid load data, and weather forecast data are cleaned and format standardized.

[0035] Among them, data cleaning and format standardization refers to the process of correcting, completing, and denoising the original multi-source heterogeneous data and unifying it into a predetermined format; for example, correcting the abnormal termination time of individual records in the charging pile platform and unifying all timestamps and power values ​​into a standard format and unit.

[0036] The historical charging data, user charging behavior characteristics, real-time power grid load data, and weather forecast data, after being cleaned and formatted, are time-aligned based on a unified time series benchmark.

[0037] Time alignment refers to the operation of adjusting data sequences from different sources with different timestamps to a unified time base; for example, aligning the time series of charging data, grid load data and meteorological data to the same set of time points with the same time interval unit.

[0038] The various types of data that have completed time alignment are spatially correlated and feature-stitched according to a preset geospatial grid to generate the spatiotemporally correlated fused dataset.

[0039] In this context, a geospatial grid refers to a spatial indexing system that divides a target geographic area into regular geographic units; for example, dividing the target area into square grids with fixed side lengths to form multiple grid units. Spatial association and feature stitching refers to the operation of combining different data features within the same geospatial grid to form a comprehensive feature vector for that grid; for example, combining the power of all charging piles in a grid, the weather forecast for that area, and the power grid load value into a multi-dimensional feature vector. A fused dataset refers to a structured data set that integrates multi-source information and possesses both temporal and spatial dimensions, generated after preprocessing, temporal alignment, and spatial association and feature stitching; for example, a dataset used for training a prediction model contains fused feature vectors sampled at fixed intervals for each grid over a specific past period, along with their corresponding actual charging power labels.

[0040] In the above-mentioned optional methods, the heterogeneous data is further cleaned, standardized, time-aligned, and spatially correlated and stitched together to transform the scattered data into a spatiotemporally correlated fused dataset, providing a high-quality, structured input foundation for subsequent model training.

[0041] In one alternative approach, S3 specifically includes:

[0042] The fused dataset is divided into a training set and a test set in chronological order.

[0043] The training set refers to a portion of the fused dataset specifically used for training machine learning models; for example, the first 80% of the time span in the fused dataset might be used as the training set. The test set refers to a portion of the fused dataset used to evaluate the performance of the trained model; for example, the remaining time span in the fused dataset might be used as the test set.

[0044] The big data prediction model is trained using the training set, and the trained big data prediction model is validated and evaluated using the test set.

[0045] Among them, big data prediction models refer to: a machine learning model structure that can process large-scale spatiotemporal fusion datasets and learn patterns from them to make future predictions; for example, using a deep learning model architecture that combines convolutional neural networks and recurrent neural networks to predict charging demand.

[0046] The big data prediction model that achieves the preset accuracy target is determined as the final prediction model.

[0047] The preset accuracy metric refers to a pre-set quantitative threshold used to determine whether the model's predictive performance meets the target; for example, requiring the model to have an average absolute percentage error of less than a specific percentage when predicting charging demand on the test set. The final prediction model refers to a big data prediction model instance that has reached the preset accuracy metric after training and validation and has been selected for actual prediction; for example, after multiple rounds of training and optimization, the model with the smallest error on the test set is selected as the final prediction model.

[0048] The historical charging data, user charging behavior characteristics, real-time grid load data, and weather forecast data within the historical period that corresponds continuously in time to the future scheduling cycle are input into the final prediction model as input features to obtain the regional charging demand prediction results and grid load prediction results that correspond to the future scheduling cycle in time and geospatial space.

[0049] The historical time period that corresponds continuously in time to the future scheduling cycle refers to a time interval that is immediately before the future scheduling cycle and has the same length as the future scheduling cycle; for example, if the future scheduling cycle is tomorrow, then the corresponding historical time period is today. The regional charging demand forecast result refers to the sequence of predicted values ​​of the total charging power of the target geographical area within the future scheduling cycle, output by the final forecast model; for example, the forecast model outputs the total charging demand power of the target area every fifteen minutes, forming a sequence containing multiple time points. The grid load forecast result refers to the sequence of predicted values ​​of the total power consumption of the target grid (excluding charging load) within the future scheduling cycle, output by the final forecast model; for example, the forecast model outputs the basic grid load power of the target area every fifteen minutes, forming another sequence.

[0050] In the above-mentioned optional approach, the fused dataset is further divided into a training set and a test set. A prediction model that meets the accuracy requirements is established through a training and verification mechanism, and historical data is used to predict future demand, thereby improving the accuracy of regional charging demand and grid load prediction.

[0051] In one alternative approach, S4 specifically includes:

[0052] A collaborative scheduling optimization model is established, with the start / stop status and output power of the multiple charging modules in each time period within the future scheduling cycle as decision variables.

[0053] In this context, "start / stop state" refers to a binary variable representing whether the charging module is in a powered-on or powered-off state during a certain time period; for example, a starting / stop state of 1 for a charging module during a specific time period indicates that it is in the starting / operating state. "Output power" refers to the electrical power value output by the charging module within a certain time period; for example, a charging module being instructed to output a specific number of kilowatts during a specific time period. "Cooperative scheduling optimization model" refers to a mathematical model that uses the starting / stop state and output power of the charging modules as decision variables, aims for optimal system energy efficiency, and considers various constraints; for example, establishing a mathematical programming model that considers cost and smooths out fluctuations for a certain number of charging modules.

[0054] The optimization objectives of the collaborative scheduling optimization model are set, including smoothing the total grid load fluctuation caused by the superposition of the grid load forecast results and the total charging power, minimizing the total operating cost including grid power purchase cost and charging module loss cost, and maximizing the renewable energy consumption rate actually consumed in the renewable energy output forecast data.

[0055] The optimization objective refers to the mathematical indicators that the collaborative scheduling optimization model aims to achieve. For example, the optimization objective of this model is to minimize the total operating cost, smooth out the fluctuations in the total grid load, and maximize the use of renewable energy. The fluctuations in the total grid load refer to the degree of instability in the total power consumption sequence formed by the superposition of the base grid load and the total charging power. For example, the magnitude of the fluctuation is quantified by calculating the variance of the total load sequence, and one of the optimization objectives is to minimize this variance. The grid purchase cost refers to the fee paid for purchasing electricity from the upper-level grid, which is usually related to the amount of electricity purchased and the time-of-use pricing. For example, the total operating cost includes the cost of purchasing electricity from the grid at different times and prices. The charging module loss cost refers to the depreciation and maintenance costs of equipment caused by frequent start-ups and shutdowns of charging modules or their operation in inefficient ranges. For example, in the model, a fixed loss cost coefficient is assigned to each start-up and shutdown action of each charging module. The total operating cost refers to the sum of the grid purchase cost and the loss cost of all charging modules within the scheduling cycle. For example, one sub-objective of the model is to minimize the sum of these two costs within future scheduling cycles. Renewable energy output forecast data refers to a series of predicted values ​​for local renewable energy power generation during future dispatch cycles; for example, predicting the average power generation of photovoltaic systems supporting charging stations in a target area every fifteen minutes based on weather forecasts. Renewable energy utilization rate refers to the proportion of energy actually consumed that originates from renewable energy sources; for example, the goal is to utilize as much of the predicted renewable energy power generation as possible in the total charging volume.

[0056] The regional charging demand forecast results are used as the equality constraint that the sum of the output power of all charging modules in each time period must satisfy in the collaborative scheduling optimization model. The grid load forecast results and the renewable energy output forecast data are used as the input parameters of the collaborative scheduling optimization model.

[0057] Among the above-mentioned optional methods, a collaborative scheduling optimization model with start-stop status and output power as decision variables is further constructed. Three major objectives are set: smoothing fluctuations, minimizing costs, and maximizing absorption rate. Charging demand is used as an equation constraint to achieve multi-module collaborative optimization.

[0058] In one alternative approach, the multi-objective function expression of the optimization objective is:

[0059]

[0060] Where T represents the total number of time periods included in the future scheduling period, N represents the total number of charging modules, t represents the time period index, i represents the charging module index, and L represents the charging module index. t P represents the power grid load forecast result for time period t. i,t The output power of charging module i during time period t is a decision variable. The renewable energy output forecast data represents the time period t. S represents the grid purchase price of electricity during time period t. i,t The start / stop state of charging module i during time period t is represented by k, which is a binary decision variable; i This indicates the corresponding loss coefficient. These are preset positive weighting coefficients used to balance the three sub-objectives: the total grid load fluctuation, the total operating cost, and the renewable energy absorption rate.

[0061] The decision variable P i,t Satisfy constraints ;D t This represents the predicted charging demand for the region during time period t.

[0062] It should be noted that the above multi-objective function expression is constructed by using a linear weighted summation method to integrate the three sub-objectives of smoothing the total grid load fluctuation, minimizing the total operating cost, and maximizing the renewable energy absorption rate into a single mathematical objective that can be optimized holistically. Specifically, the first term quantifies and penalizes the fluctuation of the total grid load by calculating the sum of the squares of the differences between the grid base load, the total charging power, and the predicted renewable energy output. Its function is to directly serve the optimization objective of peak shaving and valley filling, and smoothing fluctuations. The second term consists of the sum of the grid electricity purchase cost and the charging module loss cost. Its function is to minimize the total economic operating cost of the system while meeting charging demand. The third term represents the total amount of renewable energy actually absorbed by summing the total charging power and the minimum predicted renewable energy output. Its function is to incentivize the charging load to align with the peak renewable energy generation time, thereby maximizing the green electricity utilization rate. The positive weighting coefficients introduced in the formula are used to flexibly adjust the relative importance of the three sub-objectives in the comprehensive optimization.

[0063] Among the above-mentioned optional methods, the three major optimization objectives are further quantified through multi-objective function expressions, and weight coefficients are introduced to achieve flexible adjustment between objectives, providing an accurate and calculable mathematical model and providing a clear basis for scheduling decisions.

[0064] In one alternative approach, S5 specifically includes:

[0065] The cooperative scheduling optimization model, which includes the multi-objective function and the constraints, is transformed into a single-objective mixed-integer linear programming model through linear weighting and piecewise linearization methods.

[0066] Among them, the linear weighting and piecewise linearization method refers to mathematical processing techniques that combine multi-objective optimization problems into a single objective by assigning weights and approximating nonlinear functions as piecewise linear functions for easier solution; for example, the nonlinear terms in the original objective function are transformed by introducing auxiliary variables and linear inequality constraints. A single-objective mixed-integer linear programming model refers to a mathematical model obtained after transformation where the objective function and all constraints are linear, and it includes both continuous and integer variables; for example, the transformed charging scheduling problem becomes a standard mixed-integer linear programming model that can be processed by general-purpose commercial solvers.

[0067] The transformed single-objective mixed-integer linear programming model is solved using the branch and bound algorithm.

[0068] Among them, the branch and bound algorithm refers to a class of exact algorithms used to solve mixed integer linear programming models, which find the optimal solution through systematic enumeration and pruning; for example, calling the built-in branch and bound algorithm of the standard optimization solver to solve the model.

[0069] Extract each time period within the future scheduling cycle from the solution results. The start / stop status of all corresponding charging modules With the output power The optimized scheduling instruction sequence is formed by arranging the instructions in chronological order.

[0070] The solution result refers to the specific values ​​of the decision variables that satisfy all constraints and optimize the objective function, output by the algorithm after it has run. For example, the algorithm outputs the optimal start / stop state and output power value for each charging module in each time interval within a future scheduling cycle. A charging module refers to the core power electronic unit inside a DC charging pile that performs AC / DC power conversion and whose start / stop and output power can be controlled by signals. For example, each DC charging pile has one charging module with a rated power of a specific kilowatt. The optimized scheduling instruction sequence refers to a complete set of control instructions organized chronologically, containing the start / stop state and output power of each charging module in future time intervals. For example, based on the solution result, a control instruction table covering the future scheduling cycle and divided into fixed time intervals is generated for all charging modules.

[0071] In the above-mentioned optional approach, the multi-objective model is further transformed into a single-objective mixed-integer linear programming model, and the branch and bound algorithm is used to solve it efficiently. The start-stop state and power value are extracted from the result to form an executable instruction sequence.

[0072] In an alternative approach, the process of transforming the cooperative scheduling optimization model into a single-objective mixed-integer linear programming model is represented by the following formula:

[0073]

[0074] The constraints are satisfied:

[0075]

[0076] in, The auxiliary variable introduced represents the linearized upper bound of the square of the total power grid load fluctuation over time period t; The auxiliary variable introduced represents the time period. The actual amount of renewable energy power consumed; The auxiliary variable introduced represents the net power purchased from the grid during time period t.

[0077] Among the above optional methods, auxiliary variables are further introduced to linearize the nonlinear terms, transforming the squared fluctuation terms into a solvable form, ensuring the mathematical rigor and computational feasibility of the model, and facilitating the application of mature solution tools.

[0078] In one alternative approach, S6 specifically includes:

[0079] The optimized scheduling instruction sequence corresponding to each charging module is analyzed to obtain the start / stop status of each time period within the future scheduling cycle. With the output power .

[0080] According to the start / stop status Generate the corresponding relay control signal, and based on the output power Generate the corresponding pulse width modulation control parameters.

[0081] Among them, relay control signals refer to electrical signals used to drive the opening and closing of the contactor in the main circuit inside the charging module to realize the start and stop of the module; for example, the central controller generates a voltage signal of a specific level and sends it to the designated module according to the start and stop status values ​​in the instruction sequence. Pulse width modulation control parameters refer to the values ​​used to adjust the duty cycle of the switching devices of the power converter in the charging module, thereby precisely controlling its output power; for example, the controller calculates and sends a specific pulse width modulation duty cycle value to the controller of the charging module according to the output power value in the instruction sequence.

[0082] The relay control signal and the pulse width modulation control parameters are sent to the corresponding charging module through a preset communication interface to drive the corresponding charging module to perform the start-stop control and the power distribution.

[0083] The preset communication interface refers to the physical channel and protocol specifications determined during system design for transmitting control commands and data between the central controller and the charging modules; for example, Ethernet communication based on the TCP / IP protocol is used, and the controller is connected to the communication module of each charging module through a network switch.

[0084] In the above-mentioned optional methods, the optimized scheduling command is further parsed into relay control signals and pulse width modulation parameters, and sent to the charging module through a standard communication interface to complete the closed-loop control from decision-making to execution, thereby achieving precise power allocation.

[0085] In one alternative approach, the data cleaning and format standardization process includes: performing spatial downscaling on the weather forecast data using Kriging interpolation, and performing anomaly detection and reconstruction on the historical charging data using a variational autoencoder-based method.

[0086] Spatial downscaling refers to the process of transforming low-spatial-resolution meteorological forecast data into high-resolution data suitable for a smaller geographical area through interpolation and other methods. For example, using Kriging interpolation to generate temperature data at a grid scale suitable for the target area from city-scale temperature forecast data. Variational autoencoders (VAEs) are neural network models capable of learning the latent distribution of data and using it for generation or reconstruction. For example, VAEs can be used to model historical charging power curves to learn their normal patterns. Anomaly detection and reconstruction methods refer to the technical process of identifying outliers in data based on models such as VAEs and replacing them with reasonable values ​​generated by the model. For example, a trained VAE is first used to calculate the reconstruction error of historical data, marking points with excessive errors as anomalies, and then replacing these outliers with reconstructed values ​​generated by the model.

[0087] Among the above-mentioned optional methods, Kriging interpolation is further used to spatially downscale the meteorological data, and variational autoencoders are used to detect and reconstruct anomalies in historical charging data, thereby improving the resolution and accuracy of the input data.

[0088] The following is a complete example based on the method of this embodiment, used to illustrate the specific implementation process of the charging module energy efficiency optimization scheduling method based on big data prediction in this embodiment:

[0089] S10. Obtain historical charging data and user charging behavior characteristics related to charging load from the charging pile data platform. Obtain real-time grid load data related to grid status from the grid dispatch system. Obtain meteorological forecast data for the target geographical area and future dispatch cycles from the meteorological data interface. Historical charging data, user charging behavior characteristics, real-time grid load data, and meteorological forecast data together constitute multi-source heterogeneous data.

[0090] S20. Data cleaning and format standardization are performed on historical charging data, user charging behavior characteristics, real-time grid load data, and weather forecast data. Specifically, Kriging interpolation is used for spatial downscaling of weather forecast data, and an anomaly detection and reconstruction method based on variational autoencoders is used for historical charging data. The processed data are then time-aligned based on a unified time series benchmark. Finally, the time-aligned data are spatially correlated and feature-stitched according to a pre-defined geospatial grid to generate a spatiotemporally correlated fusion dataset.

[0091] S30. Divide the fused dataset into a training set and a test set in chronological order. Use the training set to train the big data prediction model, and use the test set to verify and evaluate the trained big data prediction model. The big data prediction model that achieves the preset accuracy index is determined as the final prediction model. Input the final prediction model with historical charging data, user charging behavior characteristics, real-time grid load data, and weather forecast data from historical periods that correspond continuously in time to the future scheduling cycle. This yields the regional charging demand prediction results and grid load prediction results corresponding to the future scheduling cycle in time and geographic space.

[0092] S40. Establish a collaborative scheduling optimization model with the start / stop status and output power of multiple charging modules in each time period of the future scheduling cycle as decision variables. Set the optimization objectives of the collaborative scheduling optimization model, which include smoothing the total grid load fluctuation caused by the superposition of grid load forecast results and total charging power, minimizing the total operating cost including grid power purchase cost and charging module loss cost, and maximizing the renewable energy absorption rate actually consumed in the renewable energy output forecast data. Use the regional charging demand forecast results as the equality constraint that the sum of the output power of all charging modules in each time period must satisfy in the collaborative scheduling optimization model, and use the grid load forecast results and renewable energy output forecast data as the input parameters of the collaborative scheduling optimization model.

[0093] S50. The collaborative scheduling optimization model, which includes multiple objective functions and constraints, is transformed into a single-objective mixed-integer linear programming model through linear weighting and piecewise linearization. The branch-and-bound algorithm is then used to solve the transformed single-objective mixed-integer linear programming model. From the solution results, the start / stop status and output power of all charging modules corresponding to each time period in the future scheduling cycle are extracted and arranged in chronological order to form an optimized scheduling instruction sequence.

[0094] S60. Analyze the optimized scheduling instruction sequence corresponding to each charging module to obtain the start / stop status and output power for each time period within the future scheduling cycle. Generate corresponding relay control signals based on the start / stop status and corresponding pulse width modulation control parameters based on the output power. Send the relay control signals and pulse width modulation control parameters to the corresponding charging modules through a preset communication interface to drive the corresponding charging modules to perform start / stop control and power distribution.

[0095] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0096] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

[0097] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for optimizing and scheduling the energy efficiency of charging modules based on big data prediction, characterized in that, The method includes: Collect multi-source heterogeneous data related to charging load and power grid status, including historical charging data, real-time power grid load data, weather forecast data, and user charging behavior characteristics; The multi-source heterogeneous data is preprocessed and feature fused to generate a spatiotemporally correlated fused dataset; Based on the fused dataset, a big data prediction model is trained and applied to output the regional charging demand prediction results and the power grid load prediction results within the future scheduling cycle. The regional charging demand forecast, the power grid load forecast, and the renewable energy output forecast are input into a collaborative scheduling optimization model established for multiple charging modules. The collaborative scheduling optimization model aims to smooth out fluctuations in the total power grid load, minimize total operating costs, and maximize the renewable energy absorption rate. Solving the cooperative scheduling optimization model yields an optimized scheduling instruction sequence for the start / stop status and output power of each charging module in the future scheduling cycle; Based on the optimized scheduling instruction sequence for each charging module, start / stop control and power allocation are performed for each charging module; The step of inputting the regional charging demand forecast, the power grid load forecast, and the renewable energy output forecast into the collaborative scheduling optimization model established for multiple charging modules further includes: A collaborative scheduling optimization model is established, with the start / stop status and output power of the multiple charging modules in each time period within the future scheduling cycle as decision variables. The optimization objectives of the collaborative scheduling optimization model are set, including smoothing the total grid load fluctuation caused by the superposition of the grid load forecast results and the total charging power, minimizing the total operating cost including grid power purchase cost and charging module loss cost, and maximizing the renewable energy consumption rate actually consumed in the renewable energy output forecast data. The regional charging demand forecast results are used as the equality constraint that the sum of the output power of all charging modules in each time period must satisfy in the collaborative scheduling optimization model. The grid load forecast results and the renewable energy output forecast data are used as the input parameters of the collaborative scheduling optimization model. The multi-objective function expression for the optimization objective is: Where T represents the total number of time periods included in the future scheduling period, N represents the total number of charging modules, t represents the time period index, i represents the charging module index, and L represents the charging module index. t P represents the power grid load forecast result for time period t. i,t The output power of charging module i during time period t is a decision variable. The renewable energy output forecast data represents the time period t. S represents the grid purchase price of electricity during time period t. i,t The start / stop state of charging module i during time period t is represented by k, which is a binary decision variable; i This indicates the corresponding loss coefficient. These are preset positive weighting coefficients used to balance the three sub-objectives: the total grid load fluctuation, the total operating cost, and the renewable energy absorption rate. The decision variable P i,t Satisfy constraints ;D t The predicted charging demand for the region during time period t is represented. The step of solving the cooperative scheduling optimization model to obtain the optimized scheduling instruction sequence of the start / stop state and output power of each charging module in the future scheduling cycle further includes: The cooperative scheduling optimization model, which includes the multi-objective function and the constraints, is transformed into a single-objective mixed-integer linear programming model through linear weighting and piecewise linearization methods. The transformed single-objective mixed-integer linear programming model is solved using the branch and bound algorithm. Extract the start / stop status of all charging modules for each time period t within the future scheduling cycle from the solution results. With the output power The optimized scheduling instruction sequence is formed by arranging the instructions in chronological order. The process of transforming the cooperative scheduling optimization model into a single-objective mixed-integer linear programming model is represented by the following formula: The constraints are satisfied: in, The auxiliary variable introduced represents the linearized upper bound of the square of the total power grid load fluctuation over time period t; The auxiliary variable introduced represents the time period. The actual amount of renewable energy power consumed; The auxiliary variable introduced represents the net power purchased from the grid during time period t.

2. The charging module energy efficiency optimization scheduling method based on big data prediction according to claim 1, characterized in that, The step of collecting multi-source heterogeneous data related to charging load and grid status further includes: Obtain the historical charging data related to charging load and the user charging behavior characteristics from the charging pile data platform; Obtain the real-time grid load data related to the grid status from the grid dispatching system; Obtain the meteorological forecast data for the corresponding geographical area and future scheduling cycle from the meteorological data interface.

3. The charging module energy efficiency optimization scheduling method based on big data prediction according to claim 2, characterized in that, The step of preprocessing and feature fusion of the multi-source heterogeneous data to generate a spatiotemporally correlated fused dataset further includes: The historical charging data, user charging behavior characteristics, real-time power grid load data, and weather forecast data are cleaned and format standardized. The historical charging data, user charging behavior characteristics, real-time power grid load data, and weather forecast data, after the data cleaning and format standardization processes, are time-aligned based on a unified time series benchmark. The various types of data that have completed time alignment are spatially correlated and feature-stitched according to a preset geospatial grid to generate the spatiotemporally correlated fused dataset.

4. The charging module energy efficiency optimization scheduling method based on big data prediction according to claim 3, characterized in that, The step of training and applying a big data prediction model based on the fused dataset to output the regional charging demand prediction results and the power grid load prediction results for the future scheduling cycle further includes: The fused dataset is divided into a training set and a test set in chronological order. The big data prediction model is trained using the training set, and the trained big data prediction model is validated and evaluated using the test set. The big data prediction model that achieves the preset accuracy index is determined as the final prediction model; The historical charging data, user charging behavior characteristics, real-time grid load data, and weather forecast data within the historical period that corresponds continuously in time to the future scheduling cycle are input into the final prediction model as input features to obtain the regional charging demand prediction results and grid load prediction results that correspond to the future scheduling cycle in time and geospatial space.

5. The charging module energy efficiency optimization scheduling method based on big data prediction according to claim 1, characterized in that, The step of performing start-stop control and power allocation for each charging module according to the optimized scheduling instruction sequence for each charging module further includes: The optimized scheduling instruction sequence corresponding to each charging module is analyzed to obtain the start / stop status of each time period within the future scheduling cycle. With the output power ; According to the start / stop status Generate the corresponding relay control signal, and based on the output power Generate the corresponding pulse width modulation control parameters; The relay control signal and the pulse width modulation control parameters are sent to the corresponding charging module through a preset communication interface to drive the corresponding charging module to perform the start-stop control and the power distribution.

6. The charging module energy efficiency optimization scheduling method based on big data prediction according to claim 3, characterized in that, The data cleaning and format standardization process includes: using Kriging interpolation to perform spatial downscaling on the meteorological forecast data, and using an anomaly detection and reconstruction method based on variational autoencoder on the historical charging data.

Images