Medium voltage line load value prediction method, system and storage medium

Abstract

The present invention relates to the field of power management technology, and in particular to a medium-voltage line load value prediction method, system and storage medium. The method comprises the following steps: performing charging event behavior analysis based on historical charging data of charging stations to construct charging mode portrait data; performing regional charging demand analysis on the charging mode portrait data to generate spatiotemporal charging demand matrix data; performing load baseline prediction based on the spatiotemporal charging demand matrix data to generate load baseline prediction curve data; constructing a line load prediction model; performing real-time medium-voltage line load prediction using the line load prediction model to generate corrected load prediction curve data; performing load curve weighted fusion of the corrected load prediction curve data and the load baseline prediction curve data to generate comprehensive load prediction curve data. The present invention realizes accurate prediction of the medium-voltage line load of a charging station through charging demand analysis, and effectively solves the problem of load fluctuation caused by charging stations on the power grid.

Description

Technical Field

[0001] The present invention relates to the technical field of power management, and in particular to a medium voltage line load value prediction method, system and storage medium. Background Art

[0002] With the popularity of electric vehicles, the number and scale of charging stations have increased rapidly, and the impact on the power grid, especially the medium-voltage power grid, has become increasingly significant. The access of charging stations not only increases the load on the lines, but also poses new challenges to the operational stability and reliability of the power grid. Since the power load of charging stations is highly random and volatile, medium-voltage lines play a vital role in connecting charging piles to the power grid. The load changes directly affect the safety and stability of the power grid. It is particularly important to accurately predict the load changes of medium-voltage lines. Therefore, the effectiveness of charging station load management is directly related to the smooth operation of the entire power grid. Through reasonable prediction and management, we can better cope with the load fluctuations of the power grid caused by the access of charging stations, and provide guarantees for the safe and reliable operation of the power grid. However, traditional medium-voltage line load value prediction methods mainly rely on traditional statistical models or empirical analysis. These methods can provide a reference for load trends to a certain extent, but due to the volatility and uncertainty of charging demand, these changes cannot be reflected in a timely manner, resulting in limited accuracy in load value prediction and difficulty in meeting actual needs. Summary of the invention

[0003] Based on this, the present invention provides a medium voltage line load value prediction method, system and storage medium to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for predicting a medium voltage line load value comprises the following steps:

[0005] Step S1: Acquire historical charging data of the charging station; perform charging event behavior analysis based on the historical charging data of the charging station to generate historical charging behavior feature data; construct a charging mode portrait based on the historical charging behavior feature data to generate charging mode portrait data;

[0006] Step S2: Acquire actual configuration data of the charging station; perform regional charging demand analysis on the charging mode portrait data through the actual configuration data of the charging station to generate spatiotemporal charging demand matrix data; perform dynamic load value calculation based on the spatiotemporal charging demand matrix data to generate dynamic load distribution matrix data;

[0007] Step S3: Perform medium voltage line topology analysis according to the actual configuration data of the charging station to generate medium voltage line power topology data; perform line load constraint processing according to the medium voltage line power topology data to generate line load constraint data; perform load baseline prediction on the dynamic load distribution matrix data through the medium voltage line power topology data to generate load baseline prediction curve data;

[0008] Step S4: constructing a line load prediction model based on line load constraint data and dynamic load distribution matrix data; using the line load prediction model to perform real-time medium voltage line load prediction and generate dynamic load prediction curve data; performing nonlinear correction processing on the dynamic load prediction curve data and generating corrected load prediction curve data;

[0009] Step S5: Perform load curve weighted fusion on the modified load forecast curve data and the load baseline forecast curve data to generate comprehensive load forecast curve data; perform overload risk assessment based on the comprehensive load forecast curve data to obtain medium voltage line overload warning data.

[0010] By acquiring the historical charging data of the charging station, the present invention can fully understand the usage of the charging station and the charging behavior pattern of the user. By analyzing the charging event behavior of these historical data, the user's charging behavior characteristics can be extracted, such as charging time period preference, charging frequency and charging amount. Based on these characteristic data, a charging mode portrait is constructed, which helps to identify the main characteristics and trends of charging demand. By obtaining the actual configuration data of the charging station (such as the number, type, power, etc. of charging piles), and combining the charging mode portrait data, the charging demand of different regions at different time periods can be accurately analyzed, and the distribution of charging demand at different times and places can be reflected in detail, which helps to identify the peak charging time period and area, calculate the dynamic load value to generate dynamic load distribution matrix data, and can reflect the dynamic change characteristics of the charging station load. By analyzing the topological structure of the medium-voltage line, detailed medium-voltage line power topology data can be generated, and the structure of the power grid and the connection relationship of each node can be fully understood, ensuring that the capacity limit of the line is fully considered in the load forecast, and avoiding the risk of power grid overload caused by inaccurate load forecasting. By using these data to predict the load baseline of the dynamic load distribution matrix, a load change trend curve based on existing conditions can be generated, providing a reference baseline for dynamic load prediction. By considering the physical topology and line restrictions, the reality and feasibility of load prediction can be ensured. Based on the previously generated line load constraint data and dynamic load distribution matrix data, a line load prediction model is constructed to achieve real-time prediction of medium-voltage line loads and capture the dynamic changes of line loads. By performing nonlinear correction processing on the prediction results and generating corrected load prediction curve data, the prediction accuracy can be further improved, and the prediction error caused by nonlinear factors can be reduced. Through real-time prediction and correction processing, the high accuracy and timeliness of load prediction results can be ensured. Taking into account the respective advantages of prediction and baseline, the future load change trend can be more accurately reflected. Overload risk assessment based on comprehensive load prediction curve data can identify potential medium-voltage line overload risks, improve the comprehensiveness and reliability of predictions, and effectively prevent medium-voltage line overloads and ensure the safe and stable operation of the power system. Therefore, a medium-voltage line load value prediction method of the present invention constructs a charging mode portrait based on the user's charging behavior characteristics by deeply analyzing the historical data and actual configuration of the charging station, considers the topological structure and physical characteristics of the medium-voltage line, and constructs a prediction model that comprehensively considers the line constraints, thereby realizing real-time prediction of dynamic loads. The overload risk assessment and early warning based on the comprehensive load prediction curve significantly improves the accuracy and timeliness of medium-voltage line load prediction, effectively meets the dynamic needs in the actual operation of charging stations, and provides a timely and effective risk monitoring and response mechanism for the safe operation of the power grid.

[0011] Preferably, step S1 comprises the following steps:

[0012] Step S11: Acquire historical charging data of the charging station; perform data preprocessing on the historical charging data of the charging station to obtain standard historical charging data;

[0013] Step S12: extracting timing characteristics according to standard historical charging data to generate charging timing characteristic data;

[0014] Step S13: Analyze historical charging event behaviors according to the charging time series characteristic data to generate historical charging behavior characteristic data;

[0015] Step S14: performing charging group cluster analysis on the charging piles based on the historical charging behavior feature data, and annotating the behavior patterns using preset behavior pattern label data to obtain charging pile behavior pattern data;

[0016] Step S15: Perform group behavior feature analysis on the charging pile behavior pattern data through historical charging behavior feature data, and construct a charging pattern portrait of the charging pile to generate charging pattern portrait data.

[0017] The present invention ensures the accuracy and consistency of the data used by performing data preprocessing on the original charging station historical charging data, such as cleaning outliers, filling missing values, and unifying formats. By extracting time series features based on standard historical charging data, the time pattern and regularity of charging behavior can be identified, and the changing trend of charging demand can be captured. By analyzing the behavior of historical charging events based on charging time series feature data, the key features of charging behavior can be revealed, and information such as the peak charging hours and preferred charging locations of users can be identified. Based on the historical charging behavior feature data, charging piles are clustered for charging groups, and charging piles with similar charging behaviors can be classified into one category to identify the charging habits and needs of different user groups. Each charging pile is given a specific behavior pattern label to form charging pile behavior pattern data. By analyzing the group behavior characteristics of charging pile behavior pattern data based on historical charging behavior feature data, the usage characteristics and user preferences of different charging piles can be deeply explored, the commonalities and differences of charging behaviors can be identified, and the usage patterns and user preferences of different charging piles can be intuitively displayed.

[0018] Preferably, step S2 comprises the following steps:

[0019] Step S21: Acquire actual configuration data of the charging station; wherein the actual configuration data of the charging station includes configuration data of the charging pile and configuration data of the medium voltage line of the charging station;

[0020] Step S22: Perform regional charging demand analysis on the charging mode portrait data through the charging pile configuration data to generate spatiotemporal charging demand matrix data;

[0021] Step S23: Calculate the charging demand change frequency according to the spatiotemporal charging demand matrix data to generate charging demand change frequency data;

[0022] Step S24: performing time step division according to the charging demand change frequency data to obtain time step division data;

[0023] Step S25: Based on the time step division data, the time-series operation characteristics of the charging piles are analyzed through the time-space charging demand matrix data, and local load fitting is performed to generate time-series operation load data;

[0024] Step S26: Performing physical mapping processing on the spatiotemporal charging demand matrix data and the time series operation load data based on the charging pile configuration data to obtain charging demand-operation load mapping data;

[0025] Step S27: Calculate the dynamic load value according to the charging demand-operation load mapping data to generate dynamic load distribution matrix data.

[0026] The present invention obtains the actual configuration data of the charging station, including the configuration data of the charging pile and the configuration data of the medium-voltage line of the charging station. The configuration data of the charging pile can reveal the number, type and distribution of the charging equipment, and ensure the accuracy of the charging demand analysis. The regional charging demand analysis of the charging mode portrait is performed through the charging pile configuration data, and the expected charging demand of each region or a single charging pile can be clarified. The charging demand change frequency is calculated according to the spatiotemporal charging demand matrix data, which can reveal the fluctuation of the charging demand and identify the peak, trough and stable period. According to the charging demand change frequency data, the time step is divided, and the time period can be divided into different levels, making the analysis of the charging demand more refined. Based on the time step division data, the time sequence operation characteristics of the charging pile are analyzed through the spatiotemporal charging demand matrix data, which can identify the working characteristics and load performance of the charging pile in different time periods, and support the efficient use of the charging pile. The actual workload of the charging pile is simulated, the realistic compliance of the load forecast is enhanced, and it is helpful to more accurately evaluate the instantaneous pressure of the medium-voltage line. Based on the charging pile configuration data, the spatiotemporal charging demand matrix data and the time-series operating load data are physically mapped, which can effectively connect the charging demand with the actual operating load and realize the spatial and temporal integration of data. The dynamic load value is calculated based on the charging demand-operating load mapping data, which can reflect the historical load changes of the charging piles.

[0027] Preferably, step S22 comprises the following steps:

[0028] Step S221: establishing a mapping relationship between the charging pile number and the actual geographical location according to the charging pile configuration data, and generating charging pile location mapping data;

[0029] Step S222: spatially associating the charging mode portrait data with the charging pile based on the charging pile position mapping data to generate charging portrait spatial distribution data;

[0030] Step S223: performing charging feature spatial interpolation on the charging portrait spatial distribution data to generate interpolated spatial charging feature data;

[0031] Step S224: performing time slicing processing on the interpolated spatial charging characteristic data using preset time slicing data to generate segmented spatial charging characteristic data;

[0032] Step S225: performing charging frequency statistics according to the charging characteristic data of the segmented space to obtain charging demand activity curve data;

[0033] Step S226: allocating time slice weights to preset time slice data using charging demand activity curve data to generate time slice weight data;

[0034] Step S227: performing time dimension weighted superposition on the segmented space charging feature data based on the time slice weight data, and performing charging demand intensity calculation to generate charging pile time series charging demand data;

[0035] Step S228: The time-space charging demand matrix is ​​converted to the charging pile sequential charging demand data through the charging pile position mapping data, and normalized to generate the time-space charging demand matrix data.

[0036] The present invention ensures the geographical accuracy of charging demand analysis by establishing a mapping relationship between the charging pile number and the actual geographical location. Based on the charging pile location mapping data, the charging mode portrait data is spatially associated, so that the charging behavior characteristics are combined with the specific geographical location, which is helpful to deeply understand the charging preferences and demand patterns in different regions. Through the charging feature space interpolation, the charging portrait spatial distribution data is supplemented and refined, and the spatial resolution is improved, so that a relatively continuous and detailed charging demand estimate can be obtained even in areas with sparse data, which improves the comprehensiveness and accuracy of the analysis. The interpolated spatial charging feature data is time-sliced ​​using preset time slice data, and the charging demand analysis is converted from the continuous time domain to discrete time points, which is convenient for the refined analysis of charging demand in a specific time period, and is conducive to identifying the peak and trough periods of demand. By statistically analyzing the activity of charging demand, the intensity of charging activities over time is revealed. By allocating time slice weights to the preset time slice data, the focus on high-demand periods is strengthened, which can better reflect the time preference of actual charging behavior. Based on the time slice weight data, the time dimension weighted superposition of the segmented space charging feature data can comprehensively consider the charging demand in different time periods, calculate the charging demand intensity, integrate the information of both time and space, and form a dynamic charging pile demand sequence. Through the charging pile location mapping data, the charging pile time series charging demand data is converted into a time-space charging demand matrix, which can integrate the charging demand in the time-space dimension, reflecting the distribution and changes of the charging demand in the time-space dimension.

[0037] Preferably, step S3 comprises the following steps:

[0038] Step S31: performing medium voltage line topology analysis according to the medium voltage line configuration data of the charging station to generate medium voltage line power topology data;

[0039] Step S32: mapping the dynamic load distribution matrix data to corresponding power nodes through the medium voltage line power topology data to generate node load distribution matrix data;

[0040] Step S33: using the medium voltage line power topology data to perform line load constraint processing on the node load distribution matrix data to generate line load constraint data;

[0041] Step S34: performing medium voltage line load analysis according to the node load distribution matrix data to generate medium voltage line load data;

[0042] Step S35: performing load time series decomposition on the medium voltage line load data, and using a preset autoregressive moving average model to perform load forecasting to generate medium voltage line load forecasting data;

[0043] Step S36: extracting the basic load value according to the medium voltage line load prediction data, and generating load baseline prediction curve data.

[0044] The present invention performs topological analysis through medium-voltage line configuration data, which helps to clarify the connection relationship and power flow direction between each line. By mapping the dynamic load distribution matrix data to the corresponding power nodes through the medium-voltage line power topology data, the charging demand can be matched with the specific nodes of the power system. The line load constraint processing of the node load distribution matrix data using the medium-voltage line power topology data can take into account the environmental conditions and safe operation conditions of the power line. The medium-voltage line load analysis based on the node load distribution matrix data can reflect the load conditions of the medium-voltage line in different time periods. The medium-voltage line load data is predicted by load time series decomposition and autoregressive moving average model (ARIMA), and the generated medium-voltage line load prediction data can capture the trend and seasonal fluctuations of the load over time, thereby enhancing the adaptability and accuracy of the prediction model. The basic load value is extracted from the medium-voltage line load prediction data, and the constant or slowly changing part in the load prediction is separated, providing a comparison benchmark for identifying and analyzing abnormal fluctuations in the load.

[0045] Preferably, step S33 includes the following steps:

[0046] Step S331: extracting line electrical physical parameters from medium voltage line power topology data to generate line physical attribute parameter data; collecting historical ambient temperature values ​​according to node load distribution matrix data to generate historical ambient temperature data;

[0047] Step S332: Calculate the power flow of the node load distribution matrix data using the line physical attribute parameter data to generate node power flow distribution data;

[0048] Step S333: deriving power transfer characteristics according to the node power flow distribution data to obtain power transfer characteristic data;

[0049] Step S334: constructing a line thermal physical model for the medium voltage line power topology data using the line physical property parameter data to obtain a line thermal physical model;

[0050] Step S335: Calculate the dissipated power of the power transfer characteristic data using the line thermal physics model to generate node power dissipation data;

[0051] Step S336: Calculate the node temperature rise value through the line thermal physics model based on the node power dissipation data and the historical ambient temperature data to generate the node temperature rise value data;

[0052] Step S337: obtaining a critical temperature threshold of the line material according to the line physical property parameter data; performing a temperature critical ratio analysis on the node temperature rise value data according to the critical temperature threshold of the line material to generate temperature critical ratio matrix data;

[0053] Step S338: Calculate the thermal stability coefficient according to the temperature critical ratio matrix data to generate a thermal stability coefficient matrix;

[0054] Step S339: performing line load constraint condition processing on the thermal stability coefficient matrix and the power transfer characteristic data, thereby obtaining line load constraint data.

[0055] The present invention extracts physical property parameters of the line, such as resistance, reactance, etc., through in-depth analysis of medium-voltage line power topology data. At the same time, historical ambient temperature values ​​are collected and historical ambient temperature data is formed. Taking into account the impact of environmental factors on the operation of the power system, the environmental adaptability and accuracy of load prediction are increased. The power flow of the node load is calculated using the line physical property parameter data, revealing the flow path and distribution of electric energy in the power grid. The power transfer characteristics are derived according to the node power flow distribution data, which helps to understand the mutual influence of load changes between different nodes. The line thermal physical model is constructed for the medium-voltage line power topology data through the line physical property parameter data, and the thermal effect of the current in the conductor can be analyzed. The power transfer characteristic data is calculated by the line thermal physical model to calculate the dissipated power, which can reflect the energy loss of the line under different load conditions. The node temperature rise value is calculated by the line thermal physical model based on the node power dissipation data and the historical ambient temperature data, and the temperature change of the power line during operation can be accurately evaluated. The critical temperature threshold of the line material is obtained according to the line physical property parameter data to ensure that the power line operates within a safe range. The temperature critical ratio analysis of the node temperature rise value data through the critical temperature threshold of the line material can reveal the temperature stability of each node. The thermal stability coefficient is calculated based on the temperature critical ratio matrix data, which can evaluate the thermal stability of the line under different load conditions. The line load constraint condition processing of the thermal stability coefficient matrix and the power transfer characteristic data can ensure that the safety restrictions of the line are fully considered in the load forecast. The risk of grid overload caused by inaccurate load forecasting is avoided to ensure the stable operation of the power system.

[0056] Preferably, step S4 comprises the following steps:

[0057] Step S41: taking a real-time charging data snapshot according to the charging station management system to generate real-time charging data; performing charging vehicle analysis according to the real-time charging data to obtain real-time charging vehicle data; and obtaining charging station temperature forecast data;

[0058] Step S42: performing charging mode portrait matching on real-time charging vehicle data through the charging mode portrait data to generate real-time charging portrait matching data;

[0059] Step S43: performing real-time charging demand processing according to the real-time charging portrait matching data to generate real-time charging demand data;

[0060] Step S44: obtaining a long short-term memory network model and model initialization parameters; performing initialization architecture processing on the long short-term memory network model using the model initialization parameters and line load constraint data to generate an initial line load prediction model;

[0061] Step S45: performing random data partitioning on the dynamic load distribution matrix data to obtain training set data and verification set data respectively;

[0062] Step S46: Perform model back propagation training on the initial line load prediction model using the training set data, and perform hyperparameter tuning and model verification using the validation set data, thereby obtaining a line load prediction model;

[0063] Step S47: transmitting the real-time charging demand data and the charging station temperature forecast data to the line load prediction model to perform real-time load value prediction and generate dynamic load prediction curve data;

[0064] Step S48: Calculate the charging mode weighting coefficient according to the real-time charging portrait matching data to obtain real-time charging weight data; use the real-time charging weight data to perform nonlinear correction processing on the dynamic load prediction curve data to generate corrected load prediction curve data.

[0065] The present invention combines the temperature forecast data of the charging station to ensure that the prediction model can operate based on the latest charging activity status and environmental conditions, thereby improving the timeliness and accuracy of the prediction. The acquisition of real-time charging vehicle data directly reflects the current usage of the charging station, while the temperature forecast data of the charging station takes into account the impact of environmental factors on power demand and transmission efficiency. The historical behavior pattern is linked to the current actual charging activity, so that the prediction model can better understand and predict the user's charging habits and behaviors, thereby improving the personalization and adaptability of the prediction model. The real-time charging demand processing is based on the matched charging portrait data, which directly reflects the charging demand at the current moment and in the short term in the future. The model is initialized and trained based on the model initialization parameters and the line load constraint data. The final line load prediction model can effectively learn the time series pattern in the historical data, and make predictions in combination with the actual carrying capacity of the line, thereby improving the prediction performance and practical value of the model. The tuning of hyperparameters during model training and verification ensures the generalization ability and stability of the model. The real-time charging demand data and the charging station temperature forecast data are transmitted to the line load prediction model for real-time load value prediction, which can timely reflect the impact of charging demand changes on line load. By calculating the real-time charging weight data and using these weights to perform nonlinear correction on the dynamic load forecast curve data, the accuracy and reliability of the prediction are further improved. This nonlinear correction step can better handle complex and nonlinear load change relationships, making the final corrected load forecast curve data more consistent with actual load changes, which is conducive to more accurate overload risk assessment and prevention.

[0066] Preferably, step S5 comprises the following steps:

[0067] Step S51: performing load curve time alignment on the modified load forecast curve data and the load baseline forecast curve data, and performing time step load level calculation to generate forecast load level data;

[0068] Step S52: performing load curve weighted fusion on the modified load forecast curve data and the load baseline forecast curve data by predicting the load level data to generate comprehensive load forecast curve data;

[0069] Step S53: performing overload risk assessment on the medium voltage line by integrating the load forecast curve data to generate line overload risk data;

[0070] Step S54: dynamically grading risk events according to the line overload risk data, and performing risk warning processing, thereby obtaining medium voltage line overload warning data.

[0071] The present invention aligns the load curve time of the modified load forecast curve data with the load baseline forecast curve data, which can ensure that the load data in different time periods are consistent. The time step load level calculation provides the load change situation in each time period by analyzing the load change in a unified time frame, ensuring a more accurate understanding of the load. By applying the predicted load level data to the weighted fusion of the modified load forecast curve and the load baseline forecast curve, the short-term dynamic changes and the long-term basic trends are integrated, and the comprehensiveness and reliability of the prediction are improved. This fusion strategy can more accurately reflect the actual load behavior and reduce the limitations of single model prediction. By analyzing the comprehensive load forecast curve data, the overload risk of the medium voltage line is evaluated, and the generated line overload risk data directly points out the existing safety hazards and vulnerable links. The dynamic rating of risk events based on the line overload risk data is a further refinement and classification of the overload risk, so that power grid managers can quickly identify the severity of the risk. On this basis, the risk warning processing triggers the corresponding warning signal, providing operators with an immediate action guide to take preventive measures, such as adjusting load distribution, starting backup power supplies, or temporarily restricting the use of certain charging stations, thereby effectively controlling risks and ensuring the safe and stable operation of the power grid.

[0072] Preferably, the present invention further provides a medium voltage line load value prediction system, which executes the medium voltage line load value prediction method as described above, and the medium voltage line load value prediction system comprises:

[0073] A charging profile building module is used to obtain historical charging data of a charging station; analyze charging event behavior based on the historical charging data of the charging station to generate historical charging behavior feature data; and build a charging mode profile based on the historical charging behavior feature data to generate charging mode profile data;

[0074] The charging demand analysis module is used to obtain the actual configuration data of the charging station; perform regional charging demand analysis on the charging mode portrait data through the actual configuration data of the charging station to generate spatiotemporal charging demand matrix data; perform dynamic load value calculation based on the spatiotemporal charging demand matrix data to generate dynamic load distribution matrix data;

[0075] The load baseline prediction module is used to perform medium-voltage line topology analysis based on the actual configuration data of the charging station to generate medium-voltage line power topology data; perform line load constraint processing based on the medium-voltage line power topology data to generate line load constraint data; perform load baseline prediction on the dynamic load distribution matrix data through the medium-voltage line power topology data to generate load baseline prediction curve data;

[0076] The real-time line load prediction module is used to build a line load prediction model based on line load constraint data and dynamic load distribution matrix data; use the line load prediction model to perform real-time medium voltage line load prediction and generate dynamic load prediction curve data; perform nonlinear correction processing on the dynamic load prediction curve data and generate corrected load prediction curve data;

[0077] The comprehensive load assessment module is used to perform load curve weighted fusion on the modified load forecast curve data and the load baseline forecast curve data to generate comprehensive load forecast curve data; perform overload risk assessment based on the comprehensive load forecast curve data to obtain medium voltage line overload warning data.

[0078] Preferably, a medium voltage line load value prediction storage medium stores a computer program, and when the computer program is executed, the medium voltage line load value prediction method as described in any one of the above is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 It is a schematic diagram of the steps of the method for predicting the load value of a medium voltage line of the present invention;

[0080] Figure 2 for Figure 1 Detailed implementation steps of step S1 in FIG.

[0081] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0082] Figure 4 for Figure 1 Detailed implementation steps of step S5;

[0083] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0084] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0085] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0086] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0087] To achieve this, please refer to Figures 1 to 4 The present invention provides a method for predicting a medium voltage line load value, comprising the following steps:

[0088] Step S1: Acquire historical charging data of the charging station; perform charging event behavior analysis based on the historical charging data of the charging station to generate historical charging behavior feature data; construct a charging mode portrait based on the historical charging behavior feature data to generate charging mode portrait data;

[0089] Step S2: Acquire actual configuration data of the charging station; perform regional charging demand analysis on the charging mode portrait data through the actual configuration data of the charging station to generate spatiotemporal charging demand matrix data; perform dynamic load value calculation based on the spatiotemporal charging demand matrix data to generate dynamic load distribution matrix data;

[0090] Step S3: Perform medium voltage line topology analysis according to the actual configuration data of the charging station to generate medium voltage line power topology data; perform line load constraint processing according to the medium voltage line power topology data to generate line load constraint data; perform load baseline prediction on the dynamic load distribution matrix data through the medium voltage line power topology data to generate load baseline prediction curve data;

[0091] Step S4: constructing a line load prediction model based on line load constraint data and dynamic load distribution matrix data; using the line load prediction model to perform real-time medium voltage line load prediction and generate dynamic load prediction curve data; performing nonlinear correction processing on the dynamic load prediction curve data and generating corrected load prediction curve data;

[0092] Step S5: Perform load curve weighted fusion on the modified load forecast curve data and the load baseline forecast curve data to generate comprehensive load forecast curve data; perform overload risk assessment based on the comprehensive load forecast curve data to obtain medium voltage line overload warning data.

[0093] In the embodiment of the present invention, reference Figure 1 The above is a schematic diagram of the steps of the medium voltage line load value prediction method of the present invention. In this embodiment, the medium voltage line load value prediction method includes the following steps:

[0094] Step S1: Acquire historical charging data of the charging station; perform charging event behavior analysis based on the historical charging data of the charging station to generate historical charging behavior feature data; construct a charging mode portrait based on the historical charging behavior feature data to generate charging mode portrait data;

[0095] In an embodiment of the present invention, the historical charging data of the charging station is obtained through the charging station management system or the intelligent metering device, including the charging pile ID, charging start time, charging end time, charging power, user type and other information. The historical charging data is cleaned, screened and statistically analyzed using data analysis tools (such as the Python Pandas library) to extract the key behavioral characteristics of the charging event, such as: the charging time preferences of different user types (for example, online car-hailing drivers prefer to charge late at night, and private car owners prefer to charge at night), the utilization rate differences of different charging piles, the average charging time, etc. According to the analysis results of the charging event behavior, the charging behavior is classified using a clustering algorithm (such as the K-means algorithm), and charging events with similar charging behavior characteristics are classified into one category to construct different charging mode portraits. For example, the charging behavior can be divided into different modes such as "fast charging type", "slow charging type", "peak time concentrated type", and the characteristic parameters such as the proportion of each mode, the average charging power, and the charging time distribution are calculated to generate the charging mode portrait data.

[0096] Step S2: Acquire actual configuration data of the charging station; perform regional charging demand analysis on the charging mode portrait data through the actual configuration data of the charging station to generate spatiotemporal charging demand matrix data; perform dynamic load value calculation based on the spatiotemporal charging demand matrix data to generate dynamic load distribution matrix data;

[0097] In an embodiment of the present invention, the actual configuration information of the target charging station is collected, such as the number of charging piles, the type of charging pile (AC / DC), the rated power, the capacity of the transformer connected to the charging station, the medium voltage line topology, etc. The charging demand of different regions (for example, charging pile groups in different regions) is analyzed in combination with the charging mode portrait data and the actual configuration data of the charging station. For example, the charging load demand of the region in different time periods is estimated based on the proportion of different charging modes in the region and the average charging power of each mode. The charging station service area is divided into several small geographic grids, and based on the analysis results of step 2, a three-dimensional matrix is ​​constructed to represent the charging load demand of different geographic grids in different time periods. The rows of the matrix represent the geographic grids, the columns represent the time periods (for example, 15 minutes is a time period), and the values ​​of the matrix elements represent the charging load demand (kW) of the geographic grid in the time period. The charging load demands of all geographic grids in each time period are superimposed to obtain the total charging load value of each time period, and generate dynamic load distribution matrix data.

[0098] Step S3: Perform medium voltage line topology analysis according to the actual configuration data of the charging station to generate medium voltage line power topology data; perform line load constraint processing according to the medium voltage line power topology data to generate line load constraint data; perform load baseline prediction on the dynamic load distribution matrix data through the medium voltage line power topology data to generate load baseline prediction curve data;

[0099] In an embodiment of the present invention, the topological structure data of the medium-voltage line to which the target charging station is connected is obtained, including line length, conductor model, substation location, line connection relationship, etc. A medium-voltage line model is established according to the actual configuration data of the charging station, and power flow calculation and topological analysis are performed to generate medium-voltage line power topology data containing information such as line impedance, transformer capacity, and switch status. Based on the topological data, combined with the power system operation specifications, the maximum allowable load of the line is calculated, and the safety margin is considered to obtain the line load constraint data. Utilize historical load data (for example, the load data of the medium-voltage line in the past year) and time series analysis methods (for example, ARIMA model) to predict the load baseline of the medium-voltage line for a period of time in the future (for example, the next week), and generate load baseline prediction curve data.

[0100] Step S4: constructing a line load prediction model based on line load constraint data and dynamic load distribution matrix data; using the line load prediction model to perform real-time medium voltage line load prediction and generate dynamic load prediction curve data; performing nonlinear correction processing on the dynamic load prediction curve data and generating corrected load prediction curve data;

[0101] In an embodiment of the present invention, a line load prediction model is constructed using a machine learning algorithm (e.g., support vector machine, neural network, etc.) based on line load constraint data, dynamic load distribution matrix data, and other influencing factors (e.g., temperature, holidays, etc.). The real-time collected charging station operation data (e.g., charging pile status, charging power, etc.) is input into the line load prediction model to predict the load value of the medium-voltage line for a period of time in the future (e.g., the next 24 hours), and generate dynamic load prediction curve data. Due to the volatility and randomness of the charging load, the dynamic load prediction curve data usually has deviations and needs to be processed nonlinearly. For example, the Kalman filter algorithm can be used to smooth the prediction results, or the prediction results can be corrected according to historical prediction errors to generate corrected load prediction curve data.

[0102] Step S5: Perform load curve weighted fusion on the modified load forecast curve data and the load baseline forecast curve data to generate comprehensive load forecast curve data; perform overload risk assessment based on the comprehensive load forecast curve data to obtain medium voltage line overload warning data.

[0103] In an embodiment of the present invention, according to the confidence of the corrected load forecast curve data and the load baseline forecast curve data, the two curves are weighted and fused to generate comprehensive load forecast curve data. For example, the load level calculation of the time step can be performed according to the predicted load value, and the load forecast curve is given a higher weight to improve the accuracy of short-term prediction. For example, the exponential smoothing method can be used to calculate the recent accuracy of each prediction method, and the accuracy is used as a weight factor. The comprehensive load forecast curve data is compared with the preset medium voltage line overload threshold to determine whether there is a line overload risk in the future. For example, if the predicted load exceeds 90% of the maximum allowable load of the line, an early warning message is issued. According to the overload risk assessment result, medium voltage line overload early warning data is generated, including information such as early warning time period, early warning level, and early warning area. For example, the early warning level can be divided into three levels of "mild", "moderate", and "severe", and the early warning information is sent to relevant operation and maintenance personnel to solve the problem of volatility caused by charging station access to the safe and stable operation of the distribution network.

[0104] As an example of the present invention, refer to Figure 2 As shown, Figure 1 Detailed implementation steps of step S1 in the flowchart, in this example, step S1 includes:

[0105] Step S11: Acquire historical charging data of the charging station; perform data preprocessing on the historical charging data of the charging station to obtain standard historical charging data;

[0106] In an embodiment of the present invention, historical charging data is exported from the charging station management system database, and the data includes information such as: charging pile ID, charging start time, charging end time, charging power, user type, payment method, etc. For missing charging power data, the average value filling method can be used to fill it; for missing charging start / end time, it can be inferred based on the previous and subsequent records or the record can be deleted. For records whose charging duration or charging power obviously exceeds the normal range, they can be regarded as abnormal data and eliminated. Convert date and time data to a unified format, for example, to a timestamp or date type. Standardize numerical data such as charging duration and charging power, such as Min-Max standardization or Z-score standardization.

[0107] Step S12: extracting timing characteristics according to standard historical charging data to generate charging timing characteristic data;

[0108] In an embodiment of the present invention, a time series analysis tool (such as the Python Statsmodels library) is used to extract time series features from standard historical charging data. For example, features such as the hour, day of the week, and whether it is a holiday when charging starts are extracted, the duration of each charge is calculated, the average charging power is calculated based on the charging power and the charging duration, the time interval between two charges of the same charging pile is calculated, and a time window (such as 1 hour) is set. The number of charges, average charging duration, total charging power, etc. in each time window are counted as charging time series feature data.

[0109] Step S13: Analyze historical charging event behaviors according to the charging time series characteristic data to generate historical charging behavior characteristic data;

[0110] In an embodiment of the present invention, the changing trends of charging frequency, charging time and charging power on weekdays, weekends, holidays and different hours are analyzed. The differences in charging behaviors of different user types (for example, online car-hailing drivers, private car owners) are analyzed. The correlation between characteristics such as charging time, charging power, and charging interval is analyzed. The charging behavior characteristic indicators obtained from the analysis are quantified as historical charging behavior characteristic data. For example, the average charging power during the evening peak (18:00-20:00) on weekdays can be used as a behavior characteristic indicator.

[0111] Step S14: performing charging group cluster analysis on the charging piles based on the historical charging behavior feature data, and annotating the behavior patterns using preset behavior pattern label data to obtain charging pile behavior pattern data;

[0112] In an embodiment of the present invention, a clustering algorithm (such as a K-means algorithm) is used to analyze the historical charging behavior feature data, and charging piles with similar charging behaviors are clustered into different groups. For example, the charging piles can be divided into different groups such as "fast charging type", "slow charging type", "peak time concentrated type" and so on according to the characteristics such as the charging time period, charging duration, and charging power. According to the preset behavior pattern label data (for example, the charging behavior pattern defined according to expert experience or industry standards), the behavior pattern of each charging pile group is labeled. For example, the charging pile group with charging time concentrated in the middle of the night and higher charging power can be labeled as "online car-hailing charging mode".

[0113] Step S15: Perform group behavior feature analysis on the charging pile behavior pattern data through historical charging behavior feature data, and construct a charging pattern portrait of the charging pile to generate charging pattern portrait data.

[0114] In an embodiment of the present invention, a statistical analysis is performed on the charging pile behavior pattern data to calculate the characteristic indicators of each behavior pattern group in different dimensions, such as: the charging frequency, charging duration and charging power distribution of different behavior pattern groups in different time periods (e.g., hours, days of the week, holidays). Statistical characteristics of different behavior pattern groups in terms of charging duration, charging power, charging interval, etc. (e.g., mean, standard deviation, maximum value, minimum value, etc.). The proportion of different user types in different behavior pattern groups. Based on the results of the group behavior feature analysis, a portrait description is performed for each behavior pattern to generate charging pattern portrait data. For example, the "online car-hailing charging mode" can be described as: the charging time is concentrated in the late night and early morning, the charging frequency is high, the charging time is short, and the charging power is high.

[0115] Preferably, step S2 comprises the following steps:

[0116] Step S21: Acquire actual configuration data of the charging station; wherein the actual configuration data of the charging station includes configuration data of the charging pile and configuration data of the medium voltage line of the charging station;

[0117] Step S22: Perform regional charging demand analysis on the charging mode portrait data through the charging pile configuration data to generate spatiotemporal charging demand matrix data;

[0118] Step S23: Calculate the charging demand change frequency according to the spatiotemporal charging demand matrix data to generate charging demand change frequency data;

[0119] Step S24: performing time step division according to the charging demand change frequency data to obtain time step division data;

[0120] Step S25: Based on the time step division data, the time-series operation characteristics of the charging piles are analyzed through the time-space charging demand matrix data, and local load fitting is performed to generate time-series operation load data;

[0121] Step S26: Performing physical mapping processing on the spatiotemporal charging demand matrix data and the time series operation load data based on the charging pile configuration data to obtain charging demand-operation load mapping data;

[0122] Step S27: Calculate the dynamic load value according to the charging demand-operation load mapping data to generate dynamic load distribution matrix data.

[0123] In an embodiment of the present invention, the charging pile configuration data is obtained by reading the charging station management system database or calling the API interface, including the charging pile ID, charging pile type (AC / DC, power size), region, installation location and other information. The charging station power system design drawings and related technical documents are consulted to collect information such as the topology of the medium-voltage line, transformer capacity, and line parameters. For example, using professional power system analysis software, the power system drawings of the area where the target charging station is located are imported to extract the topology data of the medium-voltage line, including line ID, start node, end node, rated voltage, rated current, etc. For example, the charging event data is grouped according to the charging pile ID, and the charging demand of each area in different time periods (for example, 24 time periods are divided at intervals of 1 hour) is counted, including the number of charging vehicles, charging power demand, etc. According to the regional charging demand analysis results, a spatiotemporal charging demand matrix is ​​constructed. The rows of the matrix represent different time periods, the columns represent different charging areas, and the matrix elements represent the charging demand of the time period and the area. Using the spatiotemporal charging demand matrix data, the rate of change of charging demand in each area in adjacent time periods is calculated. For example, using Python's NumPy library, calculate the change in charging power demand in each area within adjacent hours, and divide it by the length of the time period (1 hour) to obtain the charging demand change rate (kW / h) of the area within the time period. Count the frequency of the charging demand change rate in each area within a day (24 hours) to generate charging demand change frequency data. For example, using Python's Collections library, count the number of different charging demand change rates in each area within 24 hours, and store the results in a dictionary format. Divide the data based on the time step, and analyze the operating characteristics of the charging pile in each time step, such as the number of charging vehicles, average charging time, average charging power, etc. For example, group the charging event data according to the time step, count the number of charging times, charging time, charging power, etc. of each charging pile in each time step, and calculate the average value. For example, using Python's SciPy library, select a suitable curve fitting model (for example, polynomial fitting, spline interpolation, etc.), fit the load data of each charging pile in each time step, and obtain the value describing the load change of the charging pile in the time step. The data can be represented as a three-dimensional array, where the first dimension represents the charging pile ID, the second dimension represents the time step, and the third dimension represents the load value of the charging pile within the time step. According to the physical connection relationship, the time series operation load data of each charging pile in each time step is mapped to the corresponding medium-voltage line to generate charging demand-operation load mapping data. For example, the load values ​​of all charging piles connected to the same medium-voltage line in the same time step are added to obtain the total load value of the line in the time step. According to the charging demand-operation load mapping data, the dynamic load value of each medium-voltage line in each time step is calculated.For example, the total load value of each line in each time step is divided by the rated capacity of the line to obtain the load rate of the line in the time step. The dynamic load values ​​of all medium-voltage lines in all time steps are constructed into a dynamic load distribution matrix. The rows of the matrix represent different time steps, the columns represent different medium-voltage lines, and the matrix elements represent the dynamic load values ​​(e.g., load rate) of the line in the time step.

[0124] Preferably, step S22 comprises the following steps:

[0125] Step S221: establishing a mapping relationship between the charging pile number and the actual geographical location according to the charging pile configuration data, and generating charging pile location mapping data;

[0126] Step S222: spatially associating the charging mode portrait data with the charging pile based on the charging pile position mapping data to generate charging portrait spatial distribution data;

[0127] Step S223: performing charging feature spatial interpolation on the charging portrait spatial distribution data to generate interpolated spatial charging feature data;

[0128] Step S224: performing time slicing processing on the interpolated spatial charging characteristic data using preset time slicing data to generate segmented spatial charging characteristic data;

[0129] Step S225: performing charging frequency statistics according to the charging characteristic data of the segmented space to obtain charging demand activity curve data;

[0130] Step S226: allocating time slice weights to preset time slice data using charging demand activity curve data to generate time slice weight data;

[0131] Step S227: performing time dimension weighted superposition on the segmented space charging feature data based on the time slice weight data, and performing charging demand intensity calculation to generate charging pile time series charging demand data;

[0132] Step S228: The time-space charging demand matrix is ​​converted to the charging pile sequential charging demand data through the charging pile position mapping data, and normalized to generate the time-space charging demand matrix data.

[0133] In an embodiment of the present invention, the charging pile ID and geographic location information (latitude and longitude coordinates) are extracted from the charging pile configuration data. Use spatial data processing tools (such as GeoPandas, ArcGIS, etc.) to associate the charging pile ID with its corresponding longitude and latitude coordinates to generate charging pile location mapping data. According to the user's charging behavior data or other relevant information, the charging mode portrait data is spatially associated with the charging pile. For example, according to the user's home address or workplace, the charging mode to which the user belongs is associated with the nearest charging pile. The charging mode information after spatial association is merged with the charging pile location mapping data to generate charging portrait spatial distribution data, which contains the proportion or quantity information of different charging modes on each charging pile. Using spatial interpolation methods (such as inverse distance weighted interpolation, Kriging interpolation, etc.), the charging mode information on the charging pile is interpolated to the entire charging station area to generate interpolated spatial charging feature data. The interpolated spatial charging feature data can be represented as a raster data, each raster represents a small spatial area, and the value of the raster represents the proportion or quantity information of different charging modes in the area. According to the time variation law of charging demand, 24 hours a day is divided into multiple time periods, each of which is called a time slice. For example, a day can be divided into 24 1-hour time slices, or 48 0.5-hour time slices. Using the preset time slice data, the interpolated spatial charging feature data is segmented according to the time dimension to generate segmented spatial charging feature data. Each time slice corresponds to an independent interpolated spatial charging feature data, which indicates the distribution of charging modes at different spatial locations in the time slice. According to historical charging data or other relevant information, the charging frequency of each spatial location in each time slice is counted to generate charging demand activity curve data. For example, the number of charging times or the number of charging vehicles in each grid in each time slice can be counted to obtain the spatial distribution and activity of charging demand in the time slice. According to the charging demand activity curve data, the preset time slice data is assigned time slice weights to generate time slice weight data. For example, a time slice with a higher charging demand activity is given a larger weight; otherwise, a smaller weight is given. Based on the time slice weight data, the time dimension weighted superposition of the segmented space charging feature data is performed, that is, the interpolated space charging feature data of each time slice is multiplied by the corresponding weight, and then the results of all time slices are added. According to the results of weighted superposition, the charging demand intensity of each spatial location is calculated to generate the charging pile time series charging demand data. For example, the proportion of different charging modes in each grid can be multiplied by the average charging power of the mode, and then the results of all modes are added to obtain the charging demand intensity of the grid.According to the charging pile location mapping data, the charging pile time series charging demand data is converted into a spatiotemporal charging demand matrix, where the rows of the matrix represent different time slices, the columns represent different charging piles, and the matrix elements represent the charging demand intensity of the charging pile within the time slice.

[0134] As an example of the present invention, refer to Figure 3 As shown, Figure 1 Detailed implementation steps of step S3 in the flowchart, in this example, step S3 includes:

[0135] Step S31: performing medium voltage line topology analysis according to the medium voltage line configuration data of the charging station to generate medium voltage line power topology data;

[0136] In the embodiment of the present invention, the collected medium voltage line topology data is processed and analyzed using a network topology analysis tool. For example, the medium voltage line topology data is converted into a graph data structure using the NetworkX library of the Python language, and the distance, path and other information between the lines are calculated using the algorithm provided by the library. The medium voltage line power topology data is generated, which can be a graphical network topology diagram or a data structure such as an adjacency matrix that describes the line connection relationship.

[0137] Step S32: mapping the dynamic load distribution matrix data to corresponding power nodes through the medium voltage line power topology data to generate node load distribution matrix data;

[0138] In an embodiment of the present invention, each charging area in the dynamic load distribution matrix data is mapped to a corresponding power node in the medium voltage line power topology data according to the actual wiring conditions of the charging station. For example, a mapping table can be constructed to record which power node each charging area is connected to. According to the mapping relationship, the dynamic load distribution matrix data is converted into node load distribution matrix data, the rows of which represent different time steps, the columns represent different power nodes, and the matrix elements represent the dynamic load value of the power node within the time step.

[0139] Step S33: using the medium voltage line power topology data to perform line load constraint processing on the node load distribution matrix data to generate line load constraint data;

[0140] In the embodiment of the present invention, the load constraint value of each line is determined according to factors such as the rated capacity of the medium voltage line, the safety operation margin, etc. For example, according to the power industry standard, the maximum allowable load current of each line is determined taking into account the thermal stability limit and voltage stability limit of the line.

[0141] Step S34: performing medium voltage line load analysis according to the node load distribution matrix data to generate medium voltage line load data;

[0142] In an embodiment of the present invention, according to the power system flow calculation principle, the node load distribution matrix data is analyzed using the medium-voltage line power topology data, and the load current, voltage drop and other indicators of each line in each time step are calculated. The load information of each node in the total node load distribution matrix is ​​calculated, the total load of the entire medium-voltage line, and the load conditions of each sub-line are calculated to generate medium-voltage line load data.

[0143] Step S35: performing load time series decomposition on the medium voltage line load data, and using a preset autoregressive moving average model to perform load forecasting to generate medium voltage line load forecasting data;

[0144] In an embodiment of the present invention, the medium voltage line load data is time-series decomposed to extract components such as trend items, periodic items, and random items. For example, the seasonal_decompose function in the Statsmodels library of Python can be used for decomposition. Using a preset autoregressive moving average (ARMA) model or other time series prediction model, an autoregressive moving average model is constructed based on the decomposed load data. Analyze the historical load data, select appropriate AR and MA orders (for example, use the autocorrelation function and partial autocorrelation function graph to determine the optimal parameters), and then fit it through the ARMA model. After fitting, the model is used to predict future loads and generate medium voltage line load forecast data.

[0145] Step S36: extracting the basic load value according to the medium voltage line load prediction data, and generating load baseline prediction curve data.

[0146] In an embodiment of the present invention, a basic load value is extracted based on the medium voltage line load forecast data to generate load baseline forecast curve data. First, the basic load level is identified from the medium voltage line load forecast data. The basic load is usually defined as the minimum load demand of the power system without the influence of peak load. The lowest value in the load forecast data can be selected as the basic load, or a reasonable baseline level can be determined by analyzing the historical operation of the load data. This process can use the Pandas library in Python for data processing, and determine a reasonable baseline load value by calculating the mean, minimum value and standard deviation of the forecast data. The extracted basic load value is visualized to generate load baseline forecast curve data.

[0147] Preferably, step S33 includes the following steps:

[0148] Step S331: extracting line electrical physical parameters from medium voltage line power topology data to generate line physical attribute parameter data; collecting historical ambient temperature values ​​according to node load distribution matrix data to generate historical ambient temperature data;

[0149] Step S332: Calculate the power flow of the node load distribution matrix data using the line physical attribute parameter data to generate node power flow distribution data;

[0150] Step S333: deriving power transfer characteristics according to the node power flow distribution data to obtain power transfer characteristic data;

[0151] Step S334: constructing a line thermal physical model for the medium voltage line power topology data using the line physical property parameter data to obtain a line thermal physical model;

[0152] Step S335: Calculate the dissipated power of the power transfer characteristic data using the line thermal physics model to generate node power dissipation data;

[0153] Step S336: Calculate the node temperature rise value through the line thermal physics model based on the node power dissipation data and the historical ambient temperature data to generate the node temperature rise value data;

[0154] Step S337: obtaining a critical temperature threshold of the line material according to the line physical property parameter data; performing a temperature critical ratio analysis on the node temperature rise value data according to the critical temperature threshold of the line material to generate temperature critical ratio matrix data;

[0155] Step S338: Calculate the thermal stability coefficient according to the temperature critical ratio matrix data to generate a thermal stability coefficient matrix;

[0156] Step S339: performing line load constraint condition processing on the thermal stability coefficient matrix and the power transfer characteristic data, thereby obtaining line load constraint data.

[0157] In an embodiment of the present invention, electrical physical parameters such as resistance, reactance, length, and cross-sectional area of ​​the line are extracted from the medium-voltage line power topology data to generate line physical property parameter data. According to the geographical location and historical meteorological data of the area where the charging station is located, the ambient temperature data consistent with the time range of the node load distribution matrix data is mined to generate historical ambient temperature data. Based on the line physical property parameter data and the node load distribution matrix data, the power flow calculation is performed to obtain the voltage, current, power and other information of each node in each time step, and the node power flow distribution data is generated. According to the electrical parameters of the line and the node power flow distribution data, the power transfer characteristics of the line are derived by using circuit principles and power system analysis methods, that is, the relationship between the power transmitted on the line and variables such as voltage drop and current. For example, using power flow data, the power transfer relationship between each node can be analyzed, and a linear regression model or other statistical methods can be used to derive the power transfer characteristics. Use the SciPy library or Statsmodels library in Python for analysis to identify the power transfer efficiency and loss characteristics between nodes. Use the known heat conduction formula and the current-temperature relationship, combined with electrical parameters, to construct a thermal physical model of the line. Select a suitable line thermal physics model, such as the IEEE standard heat balance equation model, the IEC standard dynamic thermal rating model, etc. Take into account factors such as the thermal conductivity, specific heat capacity and ambient temperature of the line material. Through this model, the heat generated when the current passes through the line and its propagation in the material can be simulated. By analyzing the node power flow distribution data and the power transfer characteristic data, the dissipated power of each node is calculated to reflect the energy loss in the line under different load conditions. The node temperature rise value is calculated through the line thermal physics model to generate the node temperature rise value data. First, the temperature rise is calculated using the previously generated node power dissipation data combined with the historical ambient temperature data. The temperature rise is usually calculated using the following formula: Where ΔT is the temperature rise value, P dissis the power dissipation of the node, t is the time (usually taken as hours), m is the mass of the line, and C is the specific heat capacity of the line material. The critical temperature threshold of the line material is obtained according to the line physical property parameter data, and the temperature critical ratio analysis of the node temperature rise value data is performed through the critical temperature threshold, so as to generate the temperature critical ratio matrix data. The node temperature rise value data is compared with the critical temperature threshold of the line material, and the temperature critical ratio of each node in each time step is calculated to generate the temperature critical ratio matrix data. According to the temperature critical ratio matrix data, the thermal stability coefficient of each line in each time step is calculated using a preset formula or algorithm to generate a thermal stability coefficient matrix. The thermal stability coefficient is used to characterize the thermal stability of the line under the current load and ambient temperature conditions. The lower the coefficient, the closer the line is to the thermal limit and the more likely it is to have an overload risk. For example, a thermal stability coefficient threshold is set, such as 0.8. For each time step, all lines are traversed. If the thermal stability coefficient of the line is lower than the threshold, the maximum transmission power allowed for the line is calculated based on the power transfer characteristic data while meeting the thermal stability coefficient requirements. The calculated maximum transmission power is used as the load constraint value of the line in the time step, and the line load constraint data is updated.

[0158] Preferably, step S4 comprises the following steps:

[0159] Step S41: taking a real-time charging data snapshot according to the charging station management system to generate real-time charging data; performing charging vehicle analysis according to the real-time charging data to obtain real-time charging vehicle data; and obtaining charging station temperature forecast data;

[0160] Step S42: performing charging mode portrait matching on real-time charging vehicle data through the charging mode portrait data to generate real-time charging portrait matching data;

[0161] Step S43: performing real-time charging demand processing according to the real-time charging portrait matching data to generate real-time charging demand data;

[0162] Step S44: obtaining a long short-term memory network model and model initialization parameters; performing initialization architecture processing on the long short-term memory network model using the model initialization parameters and line load constraint data to generate an initial line load prediction model;

[0163] Step S45: performing random data partitioning on the dynamic load distribution matrix data to obtain training set data and verification set data respectively;

[0164] Step S46: Perform model back propagation training on the initial line load prediction model using the training set data, and perform hyperparameter tuning and model verification using the validation set data, thereby obtaining a line load prediction model;

[0165] Step S47: transmitting the real-time charging demand data and the charging station temperature forecast data to the line load prediction model to perform real-time load value prediction and generate dynamic load prediction curve data;

[0166] Step S48: Calculate the charging mode weighting coefficient according to the real-time charging portrait matching data to obtain real-time charging weight data; use the real-time charging weight data to perform nonlinear correction processing on the dynamic load prediction curve data to generate corrected load prediction curve data.

[0167] In the embodiment of the present invention, by calling the charging station management system API interface, the information of the vehicle being charged at the current moment is obtained, including the charging pile ID, charging start time, real-time charging power, etc., to generate real-time charging data. The information of the vehicle being charged, such as the vehicle ID, charging pile ID, charging start time, real-time charging power, etc., is extracted from the real-time charging data to generate real-time charging vehicle data. The vehicle information in the real-time charging vehicle data is matched with the charging mode portrait data to find the charging mode most similar to each charging vehicle, and generate real-time charging portrait matching data. A similarity calculation method can be used, for example, to calculate the similarity between the charging behavior characteristics (such as charging time, charging power, etc.) of each charging vehicle and each charging mode characteristic, and select the charging mode with the highest similarity as the matching result. According to the real-time charging portrait matching data, the future charging demand of each charging vehicle, such as charging time, charging power, etc., is predicted using the matched charging mode information. The predicted charging demand of all charging vehicles is summarized to generate real-time charging demand data. A long short-term memory network (LSTM) is selected as the line load prediction model, and a pre-trained LSTM model and model initialization parameters are obtained. Using the obtained model initialization parameters and line load constraint data, the LSTM model is initialized for architecture processing, such as setting the model's input and output layer dimensions, the number of hidden layer nodes, etc., to generate an initial line load prediction model. The dynamic load distribution matrix data is randomly divided into training set data and validation set data according to a certain ratio (e.g., 80% as a training set and 20% as a validation set). The initial line load prediction model is trained using the training set data, such as using the Adam optimization algorithm for back propagation training, and the model parameters are updated to minimize the error between the model's predicted output and the actual load value. According to the performance of the model on the validation set, the model's hyperparameters (e.g., learning rate, number of iterations, number of hidden layer nodes, etc.) are adjusted to select the optimal hyperparameter combination. The trained line load prediction model is used to predict real-time load values ​​and generate dynamic load prediction curve data. According to the real-time charging portrait matching data, the weight coefficient of each charging mode is calculated to obtain real-time charging weight data. For example, the weight of each charging mode can be determined based on the proportion of each charging mode in real-time charging vehicles. The dynamic load forecast curve data generated in step S47 is subjected to nonlinear correction processing using the real-time charging weight data to generate corrected load forecast curve data. For example, a weighted average method may be used to multiply the predicted load values ​​corresponding to different charging modes by the corresponding weights and then sum them to obtain a corrected load forecast value.

[0168] As an example of the present invention, refer to Figure 4 As shown, Figure 1 Detailed implementation steps of step S5 in the flowchart, in this example, step S5 includes:

[0169] Step S51: performing load curve time alignment on the modified load forecast curve data and the load baseline forecast curve data, and performing time step load level calculation to generate forecast load level data;

[0170] In an embodiment of the present invention, a smaller time interval is selected as a uniform step size, such as 15 minutes, to ensure that the details of the load changes can be captured. A common time range is determined, such as the next 24 hours. Data with a larger time step (such as baseline load data) are interpolated to a uniform time step using methods such as linear interpolation and spline interpolation. For each time step, the average load value of the modified load forecast curve and the load baseline forecast curve in the time period are calculated respectively. For example, if the time step is 15 minutes, the load values ​​of all data points within the 15 minutes are summed and divided by the number of data points to obtain the average load value of the time step.

[0171] Step S52: performing load curve weighted fusion on the modified load forecast curve data and the load baseline forecast curve data by predicting the load level data to generate comprehensive load forecast curve data;

[0172] In an embodiment of the present invention, the weighting coefficient can be dynamically adjusted according to the degree of deviation of the predicted load level data relative to the historical load data. For example, a threshold can be set. When the predicted load level is higher than the historical level of the same period by a certain proportion, the weight of the modified load forecast curve is increased and the weight of the load baseline forecast curve is reduced, and vice versa. For each time step, the modified load forecast value and the baseline load forecast value are weighted averaged according to the determined weighting coefficient to obtain the comprehensive load forecast value of the time step. For example, assuming that at a certain time step, the modified load forecast value is 100kW, the baseline load forecast value is 80kW, the weight coefficient of the modified load forecast curve is 0.7, and the weight coefficient of the baseline load forecast curve is 0.3, then the comprehensive load forecast value of the time step is: 0.7×100kW+0.3×80kW=94kW. For example, during the peak period of electricity consumption, a higher weight is given to the "dynamic load curve" to better reflect the real-time changing load demand; and during the trough period of electricity consumption, a higher weight is given to the "load baseline curve" to utilize its stable prediction ability. The two curves are weightedly added to generate a fused load forecast curve that takes into account both historical trends and real-time fluctuations.

[0173] Step S53: performing overload risk assessment on the medium voltage line by integrating the load forecast curve data to generate line overload risk data;

[0174] In an embodiment of the present invention, each time step of the comprehensive load forecast curve data is traversed. For each time step, the comprehensive load forecast value of each line is obtained, such as the predicted current value of the time step. One or more warning thresholds are set in combination with the design capacity, historical operation data and safety margin of the medium voltage line. For example, a standard threshold and an emergency threshold can be set, corresponding to different levels of load overload risks. These thresholds are organized into a matrix according to time steps and line nodes, and a load warning threshold matrix can be obtained. The predicted load value of each line is compared with the load warning threshold of the line to determine whether there is an overload risk. If the predicted value exceeds the set threshold, it is marked as a high risk or an emergency risk at the corresponding matrix position. Through this process, a risk assessment matrix is ​​generated to clearly identify the load risk information of each time step and node.

[0175] Step S54: dynamically grading risk events according to the line overload risk data, and performing risk warning processing, thereby obtaining medium voltage line overload warning data.

[0176] In the embodiment of the present invention, a risk rating standard is formulated according to factors such as the degree of overload, duration, and importance of the line. For example, the overload risk can be divided into the following levels: Mild risk: the predicted load exceeds the maximum allowable load by less than 10%, or the duration does not exceed 1 hour; Medium risk: the predicted load exceeds the maximum allowable load by 10% to 20%, or the duration is between 13 hours; Severe risk: the predicted load exceeds the maximum allowable load by more than 20%, or the duration exceeds 3 hours. The risk information of each time step of each line is traversed, and each overload risk event is rated according to the pre-set risk rating standard to determine its risk level. For different risk levels, corresponding early warning methods are defined, for example: Mild risk: send a text message to notify the relevant operation and maintenance personnel; Medium risk: send an email to notify the relevant person in charge, and start the load monitoring system to pay close attention to the changes in the line load; Severe risk: notify the relevant leaders by phone, and automatically or manually reduce the line load according to the pre-established load control strategy to avoid overload accidents. Based on the supply and demand balance analysis, a variety of intelligent scheduling strategies are designed, including power adjustment, load transfer, and temporary closure of some charging piles. Multi-objective optimization algorithms (such as genetic algorithms and particle swarm optimization) are used to generate the optimal scheduling strategy to ensure that line overload is avoided while meeting user needs. The scheduling strategy is converted into an executable instruction set, which includes adjusting the charging power of a certain charging pile, delaying the charging of certain low-priority vehicles, or temporarily shutting down certain non-critical charging piles. Each instruction contains detailed information such as the execution time, target charging pile, adjustment parameters, etc. to ensure accurate implementation. The load scheduling instruction set will be issued to each charging pile control system in real time to guide the dynamic load management of the charging station.

[0177] The beneficial effect of this application is that by analyzing the charging behavior characteristics of charging station users, a variety of typical charging mode portraits are constructed, and they are combined with the spatial location information of the charging pile to construct a refined spatiotemporal charging demand matrix. This method incorporates user behavior patterns into the prediction model, which can more accurately predict the charging load demand at different times and in different regions, and improve the adaptability to the volatility and uncertainty of the charging load. The influence of factors such as ambient temperature and line material characteristics on the line load capacity is considered. By constructing a line thermal physics model, calculating the thermal stability coefficient of the line under different load and ambient temperature conditions, and using it as a line load constraint, the actual carrying capacity of the line can be more accurately evaluated, avoiding the problem of overly conservative or overly aggressive prediction results caused by setting line load constraints based on experience or static parameters, and further improving the prediction accuracy. The LSTM model constructs a dynamic load prediction model by learning historical load data and real-time charging demand. Using this model for training and verification not only improves the accuracy of load prediction, but also enhances the model's adaptability to future charging demand fluctuations. By back-propagating the training set and using the validation set for hyperparameter tuning, the resulting load forecasting model can be updated in real time, ensuring that the model can quickly adjust its forecast results when charging demand changes. By applying the predicted load level data to the weighted fusion of the modified load forecast curve and the load baseline forecast curve, short-term dynamic changes and long-term basic trends are combined to improve the comprehensiveness and reliability of the forecast. This fusion strategy can more accurately reflect actual load behavior and reduce the limitations of single model predictions. Through comprehensive load forecasting and overload risk assessment, the safety and reliability of the power system under high load conditions are ensured, fundamentally improving the accuracy of load value predictions and meeting actual operational needs.

[0178] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0179] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be 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 present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for predicting medium voltage line load value, characterized in that: The following steps are involved: Step S1: Acquire historical charging data of the charging station; perform charging event behavior analysis based on the historical charging data of the charging station to generate historical charging behavior feature data; Construct a charging mode profile based on historical charging behavior feature data to generate charging mode profile data; Step S2: Acquire actual configuration data of the charging station; Perform regional charging demand analysis on the charging mode portrait data through the actual configuration data of the charging station to generate spatiotemporal charging demand matrix data; perform dynamic load value calculation based on the spatiotemporal charging demand matrix data to generate dynamic load distribution matrix data; Step S3: performing medium voltage line topology analysis according to actual configuration data of the charging station to generate medium voltage line power topology data; Perform line load constraint processing according to the medium-voltage line power topology data to generate line load constraint data; perform load baseline prediction on the dynamic load distribution matrix data through the medium-voltage line power topology data to generate load baseline prediction curve data; wherein step S3 is specifically as follows: Step S31: performing medium voltage line topology analysis according to the medium voltage line configuration data of the charging station to generate medium voltage line power topology data; Step S32: mapping the dynamic load distribution matrix data to corresponding power nodes through the medium voltage line power topology data to generate node load distribution matrix data; Step S33: using the medium voltage line power topology data to perform line load constraint processing on the node load distribution matrix data to generate line load constraint data; Step S34: performing medium voltage line load analysis according to the node load distribution matrix data to generate medium voltage line load data; Step S35: performing load time series decomposition on the medium voltage line load data, and using a preset autoregressive moving average model to perform load forecasting to generate medium voltage line load forecasting data; Step S36: extracting the basic load value according to the medium voltage line load prediction data to generate load baseline prediction curve data; Step S4: constructing a line load prediction model based on line load constraint data and dynamic load distribution matrix data; using the line load prediction model to perform real-time medium voltage line load prediction and generate dynamic load prediction curve data; performing nonlinear correction processing on the dynamic load prediction curve data and generating corrected load prediction curve data; Step S5: Perform load curve weighted fusion on the modified load forecast curve data and the load baseline forecast curve data to generate comprehensive load forecast curve data; perform overload risk assessment based on the comprehensive load forecast curve data to obtain medium voltage line overload warning data.

2. The method for predicting medium voltage line load value according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Acquire historical charging data of the charging station; perform data preprocessing on the historical charging data of the charging station to obtain standard historical charging data; Step S12: extracting timing characteristics according to standard historical charging data to generate charging timing characteristic data; Step S13: Analyze historical charging event behaviors according to the charging time series characteristic data to generate historical charging behavior characteristic data; Step S14: performing charging group cluster analysis on the charging piles based on the historical charging behavior feature data, and annotating the behavior patterns using preset behavior pattern label data to obtain charging pile behavior pattern data; Step S15: Perform group behavior feature analysis on the charging pile behavior pattern data through historical charging behavior feature data, and construct a charging pattern portrait of the charging pile to generate charging pattern portrait data.

3. The method for predicting medium voltage line load value according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: Acquire actual configuration data of the charging station; wherein the actual configuration data of the charging station includes configuration data of the charging pile and configuration data of the medium voltage line of the charging station; Step S22: Perform regional charging demand analysis on the charging mode portrait data through the charging pile configuration data to generate spatiotemporal charging demand matrix data; Step S23: Calculate the charging demand change frequency according to the spatiotemporal charging demand matrix data to generate charging demand change frequency data; Step S24: performing time step division according to the charging demand change frequency data to obtain time step division data; Step S25: Based on the time step division data, the time-series operation characteristics of the charging piles are analyzed through the time-space charging demand matrix data, and local load fitting is performed to generate time-series operation load data; Step S26: Perform physical mapping processing on the spatiotemporal charging demand matrix data and the time series operation load data based on the charging pile configuration data to obtain charging demand-operation load mapping data; Step S27: Calculate the dynamic load value according to the charging demand-operation load mapping data to generate dynamic load distribution matrix data.

4. The method for predicting medium voltage line load value according to claim 3, characterized in that: Step S22 includes the following steps: Step S221: establishing a mapping relationship between the charging pile number and the actual geographical location according to the charging pile configuration data, and generating charging pile location mapping data; Step S222: spatially associating the charging mode portrait data with the charging pile based on the charging pile position mapping data to generate charging portrait spatial distribution data; Step S223: performing charging feature spatial interpolation on the charging portrait spatial distribution data to generate interpolated spatial charging feature data; Step S224: performing time slicing processing on the interpolated spatial charging characteristic data using preset time slicing data to generate segmented spatial charging characteristic data; Step S225: performing charging frequency statistics according to the charging characteristic data of the segmented space to obtain charging demand activity curve data; Step S226: allocating time slice weights to preset time slice data using charging demand activity curve data to generate time slice weight data; Step S227: performing time dimension weighted superposition on the segmented space charging feature data based on the time slice weight data, and performing charging demand intensity calculation to generate charging pile time series charging demand data; Step S228: The time-space charging demand matrix is ​​converted to the charging pile sequential charging demand data through the charging pile position mapping data, and normalized to generate the time-space charging demand matrix data.

5. The method for predicting medium voltage line load value according to claim 1, characterized in that: Step S33 includes the following steps: Step S331: extracting line electrical physical parameters from medium voltage line power topology data to generate line physical attribute parameter data; collecting historical ambient temperature values ​​according to node load distribution matrix data to generate historical ambient temperature data; Step S332: Calculate the power flow of the node load distribution matrix data using the line physical attribute parameter data to generate node power flow distribution data; Step S333: deriving power transfer characteristics according to the node power flow distribution data to obtain power transfer characteristic data; Step S334: constructing a line thermal physical model for the medium voltage line power topology data using the line physical property parameter data to obtain a line thermal physical model; Step S335: Calculate the dissipated power of the power transfer characteristic data using the line thermal physics model to generate node power dissipation data; Step S336: Calculate the node temperature rise value through the line thermal physics model based on the node power dissipation data and the historical ambient temperature data to generate the node temperature rise value data; Step S337: obtaining a critical temperature threshold of the line material according to the line physical property parameter data; performing a temperature critical ratio analysis on the node temperature rise value data according to the critical temperature threshold of the line material to generate temperature critical ratio matrix data; Step S338: Calculate the thermal stability coefficient according to the temperature critical ratio matrix data to generate a thermal stability coefficient matrix; Step S339: performing line load constraint condition processing on the thermal stability coefficient matrix and the power transfer characteristic data, thereby obtaining line load constraint data.

6. The method for predicting medium voltage line load value according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: taking a real-time charging data snapshot according to the charging station management system to generate real-time charging data; performing charging vehicle analysis according to the real-time charging data to obtain real-time charging vehicle data; and obtaining charging station temperature forecast data; Step S42: performing charging mode portrait matching on real-time charging vehicle data through the charging mode portrait data to generate real-time charging portrait matching data; Step S43: performing real-time charging demand processing according to the real-time charging portrait matching data to generate real-time charging demand data; Step S44: obtaining a long short-term memory network model and model initialization parameters; performing initialization architecture processing on the long short-term memory network model using the model initialization parameters and line load constraint data to generate an initial line load prediction model; Step S45: performing random data partitioning on the dynamic load distribution matrix data to obtain training set data and verification set data respectively; Step S46: Perform model back propagation training on the initial line load prediction model using the training set data, and perform hyperparameter tuning and model verification using the validation set data, thereby obtaining a line load prediction model; Step S47: transmitting the real-time charging demand data and the charging station temperature forecast data to the line load prediction model to perform real-time load value prediction and generate dynamic load prediction curve data; Step S48: Calculate the charging mode weighting coefficient according to the real-time charging portrait matching data to obtain real-time charging weight data; use the real-time charging weight data to perform nonlinear correction processing on the dynamic load prediction curve data to generate corrected load prediction curve data.

7. The method for predicting medium voltage line load value according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: performing load curve time alignment on the modified load forecast curve data and the load baseline forecast curve data, and performing time step load level calculation to generate forecast load level data; Step S52: performing load curve weighted fusion on the modified load forecast curve data and the load baseline forecast curve data by predicting the load level data to generate comprehensive load forecast curve data; Step S53: performing overload risk assessment on the medium voltage line by integrating the load forecast curve data to generate line overload risk data; Step S54: dynamically grading risk events according to the line overload risk data, and performing risk warning processing, thereby obtaining medium voltage line overload warning data.

8. A medium voltage line load value prediction system, characterized in that: Used to execute the medium voltage line load value prediction method according to claim 1, the medium voltage line load value prediction system comprises: A charging profile building module is used to obtain historical charging data of a charging station; analyze charging event behavior based on the historical charging data of the charging station to generate historical charging behavior feature data; and build a charging mode profile based on the historical charging behavior feature data to generate charging mode profile data; The charging demand analysis module is used to obtain the actual configuration data of the charging station; perform regional charging demand analysis on the charging mode portrait data through the actual configuration data of the charging station to generate spatiotemporal charging demand matrix data; perform dynamic load value calculation based on the spatiotemporal charging demand matrix data to generate dynamic load distribution matrix data; The load baseline prediction module is used to perform medium-voltage line topology analysis based on the actual configuration data of the charging station to generate medium-voltage line power topology data; perform line load constraint processing based on the medium-voltage line power topology data to generate line load constraint data; perform load baseline prediction on the dynamic load distribution matrix data through the medium-voltage line power topology data to generate load baseline prediction curve data; The real-time line load prediction module is used to build a line load prediction model based on line load constraint data and dynamic load distribution matrix data; use the line load prediction model to perform real-time medium voltage line load prediction and generate dynamic load prediction curve data; perform nonlinear correction processing on the dynamic load prediction curve data and generate corrected load prediction curve data; The comprehensive load assessment module is used to perform load curve weighted fusion on the modified load forecast curve data and the load baseline forecast curve data to generate comprehensive load forecast curve data; perform overload risk assessment based on the comprehensive load forecast curve data to obtain medium voltage line overload warning data.

9. A medium voltage line load value prediction storage medium storing a computer program, characterized in that: When the computer program is executed, a medium voltage line load value prediction method as described in any one of claims 1 to 7 is implemented.

Images