A method and system for predicting electric vehicle charging load based on spatiotemporal characteristics
Through spatiotemporal and spatial characteristics analysis and comprehensive prediction modules, the problem of relying on experience and historical data in the charging load prediction method of electric vehicle is solved, and more accurate charging station site selection and resource allocation are achieved, improving the efficiency and users of electric vehicle charging infrastructure Experience.
Patent Information
- Application Number
- CN202411107819.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-13
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-08-13
AI Technical Summary
The existing charging load prediction methods for electric vehicles rely on experience and historical data, resulting in unbalanced load of charging stations, unreasonable resource allocation, and inability to adapt to changes in charging demand in time, affecting the stability of the power grid and user experience.
Using the charging load prediction method of electric vehicles based on spatiotemporal characteristics, the data acquisition module, spatiotemporal characteristic analysis module, spatiotemporal joint prediction module, site selection evaluation module and site selection optimization module are adopted to collect and preprocess data in real time to build a timing and spatial analysis model. , generate comprehensive charging demand forecast values, and optimize charging station site selection and resource configuration.
It improves the accuracy and scientificity of the site selection of charging stations, optimizes resource allocation, reduces user waiting time, improves the utilization rate and economic benefits of charging facilities, and dynamically responds to changes in charging demand.
Smart Images

Figure CN118627698B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of electric vehicles, and in particular to a method and system for predicting charging load of electric vehicles based on spatiotemporal characteristics. Background Art
[0002] With the popularity and development of electric vehicles (EVs), smart grids and new energy management have become important research areas in modern urban planning and energy management. In smart grids, the charging load management of electric vehicles is a key issue because it directly affects the stability and efficiency of the power grid. In this specific area, the reasonable site selection and layout of charging stations are crucial. The electric vehicle charging load forecasting system aims to provide accurate charging demand forecasts by analyzing the charging demand data of electric vehicles, helping decision makers optimize the site selection of charging stations, and improving the charging service level and economic benefits.
[0003] In the Chinese invention application with application publication number CN113569372A, a charging pile site selection method based on machine learning is disclosed, which is characterized by including an online model recognition unit, an offline model recognition unit and a feedback optimization unit; in order to realize intelligent decision-making on the site selection of charging piles, the present invention will analyze the data of built charging piles, obtain their attribute characteristics such as the flow of people, vehicle flow and regional attributes in each time period, and associate this data with the frequency of use and the period of use of the charging piles to establish a machine learning algorithm classification model; at the same time, considering that the data of built charging piles is relatively small, the present invention will also use the addresses and attribute characteristics of unbuilt charging piles, combine them with the data of built charging piles, and use the tri-training algorithm to fully learn and utilize the unlabeled data, thereby improving the accuracy of the model and achieving performance for practical application.
[0004] In combination with the prior art, the above application still has the following deficiencies:
[0005] At present, the traditional site selection for electric vehicle load forecasting, inaccurate recharging demand forecasting, and over-reliance on experience and historical data will lead to the following abnormal effects: uneven load at charging stations, charging queues in some areas, affecting user experience; while in other areas, charging facilities have low utilization rates, resulting in waste of resources. This will not only increase operating costs, but also affect the stability of the power grid, ultimately hindering the popularization and development of electric vehicles. By introducing a charging load forecasting system based on spatiotemporal characteristics, the accuracy and rationality of charging station site selection can be significantly improved, resource allocation can be optimized, and overall operational efficiency can be improved;
[0006] Decisions usually rely on historical data and empirical judgments, lacking accurate predictions of real-time data and future demand. This approach is not only prone to site selection deviations, but also fails to adapt to changing charging needs and user behavior patterns in a timely manner. In addition, existing site selection methods usually ignore the combined impact of spatial and temporal factors, and are not easy to effectively identify hot spots and time periods for charging demand, resulting in irrational resource allocation, low utilization of charging facilities, and even overloading of some charging stations and idleness of some charging stations. Summary of the invention
[0007] In view of the deficiencies in the prior art, the present invention provides a method and system for predicting electric vehicle charging load based on spatiotemporal characteristics, which solves the problems mentioned in the background technology.
[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: including a data acquisition module, a spatiotemporal characteristic analysis module, a spatiotemporal joint prediction module, a site selection evaluation module and a site selection optimization module;
[0009] The data acquisition module is used to set up an integrated API application program interface set, collect electric vehicle charging data sets in real time, build a database for storage, and then pre-process the stored data;
[0010] The spatiotemporal characteristic analysis module is used to construct a charging load prediction model, set a time series analysis algorithm model and a spatial analysis algorithm model, and calculate and obtain a charging demand prediction value P(t) and a charging location demand prediction value P(x, y) based on the collected electric vehicle charging data set;
[0011] The spatiotemporal joint prediction module is used to correlate the acquired charging demand prediction value P(t) and the charging location demand prediction value P(x, y) to obtain a comprehensive charging demand prediction value P(x, y, t);
[0012] The site selection evaluation module comprehensively analyzes the charging load demand, performs a preliminary comparative evaluation between the preset site selection evaluation threshold F1 and the obtained comprehensive charging demand prediction value P (x, y, t), and determines the rationality of the site selection scheme;
[0013] The site selection optimization module is used to construct a charging station site selection optimization model when the site selection is unreasonable, set the objective function U, and use the optimization algorithm to obtain the variant individual , and substitute it into the objective function U to calculate and obtain the individual fitness value U ( ), and at the same time set the operation threshold F2 to compare and evaluate with the solved objective function U to determine the location plan of the charging station.
[0014] Preferably, the data acquisition module includes a data acquisition unit, an operation cost analysis unit and a data storage and preprocessing unit;
[0015] The data collection unit is used to set an integrated API application program interface set, which is respectively connected to the charging pile management system, the vehicle networking platform, the census system, the traffic management system, the meteorological system and the GIS geographic information system, and is set to collect charging data, vehicle data, population data, traffic data and weather data every 12 hours, and summarize and generate an electric vehicle charging data set;
[0016] The charging data includes time, charging location, charging duration and charging amount;
[0017] The vehicle data includes parking location, GPS location and driving route;
[0018] The population data include residential population, working population and total population density;
[0019] The traffic data includes road traffic volume and the number of parking lots in public parking lots;
[0020] The weather data includes vehicle travel volume and charging demand under different weather conditions;
[0021] The operating cost analysis unit is used to obtain and analyze the geographic information of the site k through the GIS geographic information system, and at the same time collect land purchase or lease costs, construction costs, equipment purchase prices, electricity costs, maintenance costs, labor costs, operating costs and construction costs in combination with big data, and use linear regression, decision tree and random forest models to build a cost evaluation model, and then use AnyLogic simulation software to simulate the cost situation, and then use the ARIMA model to predict long-term costs and benefits to obtain operating cost data Ck;
[0022] The data storage and preprocessing unit is used to clean and process the collected electric vehicle charging data set, including processing missing values, outliers and standardization of data formats, and then extract features of the electric vehicle charging data set, and extract the charging feature vector X and spatial influence vector J of the electric vehicle based on the charging data, vehicle data, population data, traffic data and weather data in the electric vehicle charging data set;
[0023] The charging feature vector X={charging amount X1, charging time X2, charging times X3, charging cost X4, charging start time X5, charging end time X6}
[0024] The spatial influence vector J={total population density J1, road traffic volume J2, vehicle distribution density J3, number of parking lots J3, parking lot capacity J4, average vehicle travel rate under the influence of weather variables J5, average charging demand rate under the influence of weather variables J6}; then a time series database is constructed, and the charging feature vector X and the spatial influence vector J after feature extraction are marked with timestamps, and sent to the time series database through wireless transmission. According to the marked timestamps, the charging feature vector X and the spatial influence vector J are sorted and stored according to the order of the timestamps, and the data stored for 24 hours is transferred to historical data and backed up.
[0025] Preferably, the spatiotemporal characteristic analysis module includes a time series analysis unit and a space analysis unit;
[0026] The timing analysis unit is used to construct a timing analysis algorithm model, then extract the charging feature vector X from the timing database, and then use the historical charging data to train the timing analysis algorithm model to analyze the demand regression coefficient. , based on the charging characteristic vector X and the demand regression coefficient , construct the time series analysis algorithm formula, calculate and obtain the charging demand prediction value P(t), and predict the future charging demand;
[0027] The charging demand prediction value P(t) is obtained by the following timing analysis algorithm formula:
[0028] ;
[0029] In the formula, represents the first constant term, which represents the baseline charging demand, that is, the charging demand when there are no other influencing factors. It is used to measure the influence of historical charging feature vectors on the predicted value. T represents the total number of lag periods. T is the number of historical data points included in the time series analysis algorithm model. This value determines how many past time points the time series analysis algorithm model considers to predict the value at the current time point. For example, if n=3, the model will consider the data of the last three time points to predict the charging demand at the current moment. o represents the lag period. In the time series analysis algorithm model, o represents the lag data of the first o time units in the time series. Xi represents the demand for the i-th charging feature vector X. represents the i-th charging characteristic vector X demand at time to, which is used to capture the impact of historical data on current charging demand. Represents the error term, which represents the unexplained part of the prediction and is used to deal with random errors or noise that the model fails to capture.
[0030] Preferably, the spatial analysis unit is used to construct a spatial analysis algorithm model, then extract the spatial influence vector J from the time series database, train the spatial algorithm model, and obtain the spatial regression coefficient , and combined with the spatial influence vector J and the spatial regression coefficient , construct a spatial analysis algorithm formula, and calculate and obtain the charging location demand prediction value P (x, y);
[0031] The charging location demand prediction value P (x, y) is calculated and obtained by the following spatial analysis algorithm formula:
[0032] ;
[0033] In the formula, represents the second constant term, which represents the baseline charging demand, that is, the charging demand when there are no other influencing factors. (x, y) represents longitude and latitude. represents the spatial regression coefficient, which measures the influence of each spatial influence vector J on the charging demand. Indicates location The related spatial influence vector J, Jj represents the j-th spatial influence vector, represents the first error term, which represents the unexplained part of the prediction. m represents the total number of spatial influencing factors that affect the charging demand at location (x, y). For example, if you choose 5 different spatial influencing vectors J for prediction, then m is 5. Each influencing factor has a corresponding spatial regression coefficient. .
[0034] Preferably, the spatiotemporal joint prediction module is used to construct a spatiotemporal joint prediction algorithm model, and performs combined calculation to obtain a comprehensive charging demand prediction value P (x, y, t) based on the obtained charging demand prediction value P (t) and charging location demand prediction value P (x, y);
[0035] The comprehensive charging demand prediction value P (x, y, t) is calculated by the following algorithm formula:
[0036] ;
[0037] In the formula, represents the third constant term, which represents the baseline charging demand. P(x, y, t) represents the comprehensive charging demand forecast value at time t and location (x, y), which comprehensively considers the charging demand of time and space factors. Represents the second error term, which represents the random error or noise that the model fails to capture when predicting the charging demand value. It can include the influence of factors in the data that are not considered by the model, or random fluctuations in the data itself.
[0038] Preferably, the site selection evaluation module obtains the site selection evaluation threshold F1 by performing a comprehensive average calculation on the demand coverage rate, utilization rate, service radius and cost-effectiveness ratio of the charging piles, and then performs a preliminary comparative evaluation with the obtained comprehensive charging demand forecast value P (x, y, t) to analyze the rationality of the site selection plan;
[0039] The specific evaluation plan is as follows:
[0040] When the comprehensive charging demand prediction value P (x, y, t) > the site selection evaluation threshold F1, it means that the existing charging station load abnormality site selection is reasonable. At this time, the current site selection is planned to plan the electric vehicle power load supply;
[0041] This area may be a hot spot for charging demand and suitable for adding charging stations. Existing charging stations may face high loads and need to be sited to ensure sufficient charging facilities and power supply;
[0042] When the comprehensive charging demand prediction value P (x, y, t) ≤ the site selection evaluation threshold F1, it means that the existing charging station load normal site selection is unreasonable. At this time, the current site selection is not selected and the site selection optimization is performed again;
[0043] This area may not be a hot spot for charging demand and is not suitable for prioritizing the establishment of charging stations. Existing charging facilities may be idle or under-used during this time period, and there is no need to increase the power load supply for electric vehicles.
[0044] Preferably, the site selection optimization module includes an objective function analysis unit, an optimization calculation unit and an evaluation iteration unit;
[0045] The objective function analysis unit is used to construct an objective function U to sum up the utility of the site k based on the obtained comprehensive charging demand forecast value P (x, y, t) and the operating cost data Ck when it is determined that the current site selection is unreasonable;
[0046] The objective function U is calculated by the following algorithm formula:
[0047] ;
[0048] Where wk represents the weight value, which indicates the importance of charging station k. The weight can be set based on various factors, such as user demand and policy support. Indicates the location of site k and the predicted value of charging demand at time t, Maximize represents the objective of maximizing utility, and K represents the total number of candidate charging station locations.
[0049] Preferably, the optimization calculation unit is used to perform reproduction according to the objective function U to generate two individuals and , exchange some of the site selection point k information, set the intersection point c to generate a new individual and , for the newly generated individuals and Some site selection points k are randomly changed to obtain mutant individuals , and re-substitute it into the objective function U to calculate and analyze the fitness value U of each individual ( );
[0050] The individual fitness value U( ) is obtained through the following algorithm formula:
[0051] ;
[0052] In the formula, Indicates the location of the kth charging station in the ath candidate solution The predicted value of charging demand at time t, Represents the result after the mutation operation of the a-th individual.
[0053] Preferably, the evaluation iteration unit presets an operation threshold F2 based on the operation cost and then compares it with the obtained individual fitness value U ( ) to conduct comparative evaluation, analyze the operation status of the electric vehicle charging load at the optimized site k, and generate an iterative mechanism;
[0054] When the individual fitness value U( ) ≥ operation threshold F2, it means that the total utility of the current site k is normal. Specifically, it means that the comprehensive charging demand forecast value is high at the selected candidate location. , and the construction and operation costs of these locations are relatively low, that is, the operating cost data Ck is low. Overall, such a solution is more economical, can provide a higher level of charging service, and is more cost-effective;
[0055] When the individual fitness value U( ) < operation threshold F2, indicating that the total utility of the current site k is abnormal. At this time, the iteration mechanism is executed to regenerate two individuals for mutation and calculate and analyze the fitness value U of each new individual ( ), indicating that the total utility of this site selection is low, which may be due to the high demand for the comprehensive charging demand forecast value of the selected location. The lower the cost of construction and operation, or the higher the cost of operation, the higher the cost data Ck. Such a scheme is inferior to the scheme with higher utility value in terms of economic benefit and service level.
[0056] A method for predicting electric vehicle charging load based on spatiotemporal characteristics comprises the following steps:
[0057] S1. First, an integrated API application program interface set is set up to collect electric vehicle charging data sets in real time, and a database is built to store the data, and then the stored data is pre-processed;
[0058] S2. By constructing a charging load prediction model, setting a time series analysis algorithm model and a spatial analysis algorithm model, and calculating and obtaining a charging demand prediction value P(t) and a charging location demand prediction value P(x, y) based on the collected electric vehicle charging data set;
[0059] S3, then correlate the obtained charging demand prediction value P(t) and the charging location demand prediction value P(x, y) to obtain a comprehensive charging demand prediction value P(x, y, t);
[0060] S4. By comprehensively analyzing the charging load demand, a preliminary comparative evaluation is performed between the preset site selection evaluation threshold F1 and the obtained comprehensive charging demand prediction value P (x, y, t) to determine the rationality of the site selection plan;
[0061] S5. Finally, in the case of unreasonable site selection, a charging station site selection optimization model is constructed, the objective function U is set, and the optimization algorithm is used to solve and obtain the individual fitness value U ( ), and set the operation threshold F2 to compare and evaluate with the solved objective function U to determine the charging station site selection plan.
[0062] The present invention provides a method and system for predicting electric vehicle charging load based on spatiotemporal characteristics. It has the following beneficial effects:
[0063] (1) Through the comprehensive data collection and preprocessing of the data acquisition module, the system can acquire and store charging data, vehicle data, population data, traffic data and weather data in real time to form a comprehensive electric vehicle charging data set. The spatiotemporal characteristic analysis module combines the time series analysis algorithm and the spatial analysis algorithm to calculate the charging demand prediction value P(t) and the charging location demand prediction value P(x, y), respectively, while the spatiotemporal joint prediction module further integrates these two types of prediction values to generate a comprehensive charging demand prediction value P(x, y, t). This multi-level and multi-dimensional prediction method greatly improves the accuracy and scientificity of the charging station site selection, so that the charging station layout can better match the actual charging demand and optimize resource allocation.
[0064] (2) The system uses the data acquisition module, the operating cost analysis unit, the GIS geographic information system and big data analysis to evaluate the various costs of the site selection point. The cost evaluation model is constructed through linear regression, decision tree and random forest models, and the cost situation is simulated using AnyLogic simulation software. The long-term cost and benefit forecast is then carried out through the ARIMA model to obtain the operating cost data Ck. The site selection optimization module further constructs the objective function U based on the operating cost data Ck and the comprehensive charging demand forecast value P (x, y, t) to optimize the site selection plan of the charging station. Through this comprehensive cost-benefit analysis and optimization, the construction and operation costs of the charging station can be effectively optimized and the overall economic benefits can be improved.
[0065] (3) The system can comprehensively analyze the charging load demand and conduct reasonable site selection evaluation and optimization through the site selection evaluation module and the site selection optimization module. In the preliminary evaluation, if the comprehensive charging demand prediction value P (x, y, t) exceeds the preset site selection evaluation threshold F1, it means that the current site selection is reasonable and the load is high, and it needs to be optimized to increase charging facilities; if it does not exceed, no adjustment is required. Through the optimization calculation of the objective function U and the individual fitness value U ( ) to ensure that charging facilities are added in high-demand areas and that resource waste is avoided in low-demand areas. This dynamic adjustment mechanism not only improves the service level of charging stations, enabling charging stations to better meet user needs and reduce user waiting time, but also improves user satisfaction and experience. BRIEF DESCRIPTION OF THE DRAWINGS
[0066] Figure 1 This is a schematic diagram of the process of an electric vehicle charging load prediction system based on time and space characteristics of the present invention;
[0067] Figure 2 The figure is a schematic diagram of the steps of a method for predicting charging load of an electric vehicle based on spatiotemporal characteristics according to the present invention. DETAILED DESCRIPTION
[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0069] Example 1
[0070] See also Figure 1The present invention provides an electric vehicle charging load prediction system based on spatiotemporal characteristics. To achieve the above purpose, the present invention is implemented through the following technical solutions: including a data acquisition module, a spatiotemporal characteristic analysis module, a spatiotemporal joint prediction module, a site selection evaluation module and a site selection optimization module;
[0071] The data acquisition module is used to set up an integrated API application program interface set, collect electric vehicle charging data sets in real time, build a database for storage, and then pre-process the stored data;
[0072] The spatiotemporal characteristic analysis module is used to construct a charging load prediction model, set up a time series analysis algorithm model and a spatial analysis algorithm model, and calculate the charging demand prediction value P(t) and the charging location demand prediction value P(x, y) based on the collected electric vehicle charging data set;
[0073] The spatiotemporal joint prediction module is used to correlate the obtained charging demand prediction value P(t) and the charging location demand prediction value P(x, y) to obtain a comprehensive charging demand prediction value P(x, y, t);
[0074] The site selection evaluation module comprehensively analyzes the charging load demand, performs a preliminary comparative evaluation between the preset site selection evaluation threshold F1 and the obtained comprehensive charging demand prediction value P (x, y, t), and determines the rationality of the site selection plan;
[0075] The site selection optimization module is used to build a charging station site selection optimization model when the site selection is unreasonable, set the objective function U, and use the optimization algorithm to obtain the variant individual , and substitute it into the objective function U to calculate and obtain the individual fitness value U ( ), and at the same time set the operation threshold F2 to compare and evaluate with the solved objective function U to determine the location plan of the charging station.
[0076] In this embodiment, the system collects and preprocesses various relevant data in real time through the data acquisition module, ensuring the accuracy and timeliness of the system input data. The spatiotemporal characteristic analysis module accurately predicts the acquired charging demand prediction value P(t) and charging location demand prediction value P(x, y) through time series analysis and spatial analysis algorithms. The spatiotemporal joint prediction module further integrates these two types of prediction values to generate a comprehensive charging demand prediction value P(x, y, t). These steps ensure a comprehensive understanding of future charging demand and provide a solid data foundation for subsequent site selection evaluation and optimization. The site selection evaluation module comprehensively analyzes the charging load demand, sets the site selection evaluation threshold F1, and makes a preliminary comparison with the comprehensive charging demand prediction value P(x, y, t) to evaluate the rationality of the existing charging sites and determine whether optimization is needed. In the case of unreasonable site selection, the site selection optimization module constructs a charging station site selection optimization model, sets the objective function U, and uses the optimization algorithm to solve the individual fitness value U( ), and then compared with the operating threshold F2, and finally determined the best charging station site selection plan. This process ensures the scientific and economical site selection of charging stations, reduces resource waste, and improves the utilization efficiency of charging facilities. The system improves the accuracy and reliability of charging demand forecasting through real-time data collection and multi-dimensional prediction models. Secondly, by dynamically evaluating and optimizing the site selection plan, the rationality and economic benefits of the charging station layout are ensured, and the construction and operation costs are reduced. Finally, the system can respond to changes in charging demand in a timely manner, improving the service level and user satisfaction of charging stations. Overall, these improvements not only improve the overall efficiency of electric vehicle charging infrastructure, but also provide strong support for the popularization and promotion of electric vehicles.
[0077] Example 2
[0078] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically: the data acquisition module includes a data acquisition unit, an ,operation cost analysis unit and a data storage and pre-processing unit;
[0079] The data collection unit is used to set up an integrated API application program interface set, which is connected to the charging pile management system, vehicle networking platform, census system, traffic management system, meteorological system and GIS geographic information system respectively. It is set to collect charging data, vehicle data, population data, traffic data and weather data every 12 hours, and summarize them to generate electric vehicle charging data sets;
[0080] Charging data includes time, charging location, charging duration and charging amount;
[0081] Vehicle data includes parking location, GPS location and driving route;
[0082] Population data include residential population, working population and total population density;
[0083] Traffic data include road traffic volume and the number of people parked in public parking lots;
[0084] Weather data includes vehicle travel and charging demand under different weather conditions;
[0085] The operating cost analysis unit is used to obtain and analyze the geographic information of the site k through the GIS geographic information system. At the same time, it combines big data to collect land acquisition or leasing costs, construction costs, equipment purchase prices, electricity costs, maintenance costs, labor costs, operating costs and construction costs. It uses linear regression, decision tree and random forest models to build a cost evaluation model, and then uses AnyLogic simulation software to simulate the cost situation. Then, it uses the ARIMA model to predict long-term costs and benefits to obtain operating cost data Ck;
[0086] The data storage and preprocessing unit is used to clean and process the collected electric vehicle charging data set, including processing missing values, outliers and standardization of data formats, and then extract features of the electric vehicle charging data set. Based on the charging data, vehicle data, population data, traffic data and weather data in the electric vehicle charging data set, the charging feature vector X and spatial influence vector J of the electric vehicle are extracted;
[0087] Charging feature vector X = {charging amount X1, charging time X2, charging times X3, charging cost X4, charging start time X5, charging end time X6}
[0088] Spatial influence vector J={total population density J1, road traffic volume J2, vehicle distribution density J3, number of parking lots J3, parking lot capacity J4, average vehicle travel rate under the influence of weather variables J5, average charging demand rate under the influence of weather variables J6}; then construct a time series database, mark the timestamps of the charging feature vector X and the spatial influence vector J after feature extraction, and send them to the time series database through wireless transmission. According to the marked timestamps, the charging feature vector X and the spatial influence vector J are sorted and stored according to the order of the timestamps, and the data stored for 24 hours is transferred to historical data and backed up.
[0089] In this embodiment, the system sets an integrated API application program interface set through the data acquisition unit, and regularly collects charging, vehicle, population, traffic and weather data from multiple systems such as charging pile management system, vehicle networking platform, census system, traffic management system, meteorological system and GIS geographic information system to generate a comprehensive electric vehicle charging data set. The operating cost analysis unit combines GIS geographic information system and big data technology, uses linear regression, decision tree and random forest models to build a cost evaluation model, and uses AnyLogic simulation and ARIMA model to perform long-term cost and benefit forecasts to provide detailed operating cost data Ck. The data storage and preprocessing unit cleans, processes and extracts features from the data, generates charging feature vectors X and spatial influence vectors J, and builds a time series database through wireless transmission to ensure the timeliness and accuracy of the data. These improvements enable the system to complete highly automated and standardized tasks in data collection and processing, effectively reduce data missing and anomalies, and improve data integrity and reliability. Compared with traditional technical means, the system not only provides more accurate charging load predictions through regular, multi-source data collection and complex cost analysis models, but also accurately evaluates and optimizes the construction and operation costs of charging stations. This series of improvements has brought about significant improvements in results, including more accurate charging demand forecasting, more reasonable charging station site selection, lower operating costs and higher service levels, thus providing strong support for the optimized layout and efficient operation of electric vehicle charging infrastructure.
[0090] Example 3
[0091] This embodiment is explained in Example 2. Please refer to Figure 1 ,Specifically: the spatiotemporal characteristic analysis module includes a time series analysis unit and a ,spatial analysis unit;
[0092] The timing analysis unit is used to build a timing analysis algorithm model, extract the charging feature vector X from the timing database, and then use the historical charging data to train the timing analysis algorithm model to analyze the demand regression coefficient. , based on the charging characteristic vector X and the demand regression coefficient , construct the time series analysis algorithm formula, calculate and obtain the charging demand prediction value P(t), and predict the future charging demand;
[0093] The charging demand prediction value P(t) is obtained by the following timing analysis algorithm formula:
[0094] ;
[0095] In the formula, represents the first constant term, which represents the baseline charging demand, that is, the charging demand when there are no other influencing factors, It is used to measure the influence of historical charging feature vectors on the predicted value. T represents the total number of lag periods. T is the number of historical data points included in the time series analysis algorithm model. This value determines how many past time points the time series analysis algorithm model considers to predict the value at the current time point. For example, if n=3, the model will consider the data of the last three time points to predict the charging demand at the current moment. o represents the lag period. In the time series analysis algorithm model, o represents the lag data of the first o time units in the time series. Xi represents the demand for the i-th charging feature vector X. represents the i-th charging characteristic vector X demand at time to, which is used to capture the impact of historical data on current charging demand. Represents the error term, which represents the unexplained part of the prediction and is used to deal with random errors or noise that the model fails to capture.
[0096] The spatial analysis unit is used to build a spatial analysis algorithm model, extract the spatial influence vector J from the time series database, train the spatial algorithm model, and obtain the spatial regression coefficient , and combined with the spatial influence vector J and the spatial regression coefficient , construct a spatial analysis algorithm formula, and calculate and obtain the charging location demand prediction value P (x, y);
[0097] The predicted value of charging location demand P (x, y) is calculated using the following spatial analysis algorithm formula:
[0098] ;
[0099] In the formula, represents the second constant term, which represents the baseline charging demand, that is, the charging demand when there are no other influencing factors. (x, y) represents longitude and latitude. represents the spatial regression coefficient, which measures the influence of each spatial influence vector J on the charging demand. Indicates location The related spatial influence vector J, Jj represents the j-th spatial influence vector, represents the first error term, which represents the unexplained part of the prediction. m represents the total number of spatial influencing factors that affect the charging demand at location (x, y). For example, if you choose 5 different spatial influencing vectors J for prediction, then m is 5. Each influencing factor has a corresponding spatial regression coefficient. .
[0100] In this embodiment, the system significantly improves the accuracy and reliability of electric vehicle charging demand prediction through the spatiotemporal characteristic analysis module. The time series analysis unit accurately predicts the future charging demand value P(t) by constructing and training the time series analysis algorithm model and using the historical charging data and the charging feature vector X. The spatial analysis unit calculates the charging location demand prediction value P(x, y) by constructing the spatial analysis algorithm model, combining the spatial influence vector J and the spatial regression coefficient, and realizes the spatial distribution prediction of the charging demand. Compared with traditional technical means, this module not only improves the complexity and accuracy of the prediction model through the comprehensive application of time series and spatial analysis, but also can refine the demand prediction to the specific time and place. This improvement brings many beneficial effects. First, the accurate charging demand prediction P(t) and the charging location demand prediction value P(x, y) provide reliable data support for the subsequent site selection and operation of charging stations. Secondly, through the combination of time series and spatial data, the system can capture the spatiotemporal dynamic changes of charging demand, and improve the flexibility and practicality of prediction. Finally, compared with the traditional single-dimensional prediction method, this system can more comprehensively reflect the actual charging demand after comprehensively considering time and space factors, reduce prediction errors, and improve the efficiency of charging station layout and resource allocation.
[0101] Example 4
[0102] This embodiment is explained in Example 1, please refer to Figure 1 Specifically: the spatiotemporal joint prediction module is used to construct a spatiotemporal joint prediction algorithm model, and obtains a comprehensive charging demand prediction value P(x, y, t) by combining and calculating the obtained charging demand prediction value P(t) and charging location demand prediction value P(x, y);
[0103] The comprehensive charging demand prediction value P (x, y, t) is calculated by the following algorithm formula:
[0104] ;
[0105] In the formula, represents the third constant term, which represents the baseline charging demand. P(x, y, t) represents the comprehensive charging demand forecast value at time t and location (x, y), which comprehensively considers the charging demand of time and space factors. Represents the second error term, which represents the random error or noise that the model fails to capture when predicting the charging demand value. It can include the influence of factors in the data that are not considered by the model, or random fluctuations in the data itself.
[0106] The site selection evaluation module obtains the site selection evaluation threshold F1 by performing a comprehensive average calculation of the demand coverage rate, utilization rate, service radius and cost-effectiveness ratio of the charging piles, and then makes a preliminary comparative evaluation with the obtained comprehensive charging demand forecast value P (x, y, t) to analyze the rationality of the site selection plan;
[0107] The specific evaluation plan is as follows:
[0108] When the comprehensive charging demand prediction value P (x, y, t) > the site selection evaluation threshold F1, it means that the existing charging station load abnormality site selection is reasonable. At this time, the current site selection is planned to plan the electric vehicle power load supply;
[0109] This area may be a hot spot for charging demand and suitable for adding charging stations. Existing charging stations may face high loads and need to be sited to ensure sufficient charging facilities and power supply;
[0110] When the comprehensive charging demand prediction value P (x, y, t) ≤ the site selection evaluation threshold F1, it means that the existing charging station load normal site selection is unreasonable. At this time, the current site selection is not selected and the site selection optimization is performed again;
[0111] This area may not be a hot spot for charging demand and is not suitable for prioritizing the establishment of charging stations. Existing charging facilities may be idle or under-used during this time period, and there is no need to increase the power load supply for electric vehicles.
[0112] In this embodiment, the system constructs a spatiotemporal joint prediction algorithm model through a spatiotemporal joint prediction module, comprehensively considers time and space factors, and calculates the comprehensive charging demand prediction value P (x, y, t), which provides a scientific basis for the site selection and capacity planning of charging stations. The site selection evaluation module sets the site selection evaluation threshold F1 by comprehensively setting indicators such as demand coverage, utilization, service radius and cost-effectiveness ratio, and compares and analyzes it with the comprehensive charging demand prediction value to judge the rationality of the site selection plan. This innovative design brings significant beneficial effects. First, through the accurate comprehensive charging demand prediction value P (x, y, t), the system can identify the hot spots of charging demand in real time to ensure the reasonable layout and efficient utilization of charging facilities. Secondly, the site selection evaluation module avoids blind expansion and waste of resources through quantitative evaluation methods, and improves the scientific nature of charging station site selection and the accuracy of decision-making. Compared with traditional technical means, the spatiotemporal joint prediction and comprehensive evaluation method of the system not only improves the accuracy and comprehensiveness of the prediction, but also dynamically responds to actual demand changes and optimizes the layout and operation strategy of charging stations.
[0113] Example 5
[0114] This embodiment is explained in Example 1, please refer to Figure 1,Specifically: the site selection optimization module includes the objective function ,analysis unit, the optimization calculation unit and the evaluation iteration unit;
[0115] The objective function analysis unit is used to construct an objective function U to sum up the utility of the site k based on the obtained comprehensive charging demand forecast value P (x, y, t) and the operating cost data Ck when it is judged that the current site selection is unreasonable;
[0116] The objective function U is calculated using the following algorithm formula:
[0117] ;
[0118] Where wk represents the weight value, which indicates the importance of charging station k. The weight can be set based on various factors, such as user demand and policy support. Indicates the location of site k and the predicted value of charging demand at time t, Maximize represents the objective of maximizing utility, and K represents the total number of candidate charging station locations.
[0119] The optimization calculation unit is used to reproduce according to the objective function U and generate two individuals and , exchange some of the site selection point k information, set the intersection point c to generate a new individual and , for the newly generated individuals and Some site selection points k are randomly changed to obtain mutant individuals , and re-substitute it into the objective function U to calculate and analyze the fitness value U of each individual ( );
[0120] Individual fitness value U( ) is obtained through the following algorithm formula:
[0121] ;
[0122] In the formula, Indicates the location of the kth charging station in the ath candidate solution The predicted value of charging demand at time t, Represents the result after the mutation operation of the a-th individual.
[0123] The evaluation iteration unit presets the operation threshold F2 based on the operation cost and then compares it with the obtained individual fitness value U ( ) to conduct comparative evaluation, analyze the operation status of the electric vehicle charging load at the optimized site k, and generate an iterative mechanism;
[0124] When the individual fitness value U( ) ≥ operation threshold F2, it means that the total utility of the current site k is normal. Specifically, it means that the comprehensive charging demand forecast value is high at the selected candidate location. , and the construction and operation costs of these locations are relatively low, that is, the operating cost data Ck is low. Overall, such a solution is more economical, can provide a higher level of charging service, and is more cost-effective;
[0125] When the individual fitness value U( ) < operation threshold F2, indicating that the total utility of the current site k is abnormal. At this time, the iteration mechanism is executed to regenerate two individuals for mutation and calculate and analyze the fitness value U of each new individual ( ), indicating that the total utility of this site selection is low, which may be due to the high demand for the comprehensive charging demand forecast value of the selected location. The lower the cost of construction and operation, or the higher the cost of operation, the higher the cost data Ck. Such a scheme is inferior to the scheme with higher utility value in terms of economic benefit and service level.
[0126] In this embodiment, when the system determines that the current site selection is unreasonable through the site selection optimization module, the objective function analysis unit constructs the objective function U according to the comprehensive charging demand prediction value P (x, y, t) and the operating cost data Ck, and performs optimization calculations, generates new individuals through reproduction, crossover and mutation, and calculates the fitness value U ( ). The evaluation iteration unit evaluates the individual fitness value by comparing it with the preset operation threshold F2 to ensure that the total utility of the site selection scheme is optimal. This method brings significant beneficial effects. First, through the precise objective function U and individual fitness value U ( ) calculation and optimization, the system can find the optimal charging station site selection plan, improving the layout rationality and utilization efficiency of charging facilities. Secondly, through the evaluation iteration mechanism, the system can dynamically adjust and optimize the site selection plan to ensure that the construction and operation costs of the charging station are minimized while meeting efficient service needs. Compared with traditional site selection methods, the introduction of this module not only improves the scientificity and accuracy of site selection decisions, but also significantly reduces operating costs and improves overall benefits and service levels.
[0127] Example 7
[0128] See also Figure 1 and Figure 2 , a method for predicting electric vehicle charging load based on spatiotemporal characteristics, comprising the following steps:
[0129] S1. First, an integrated API application program interface set is set up to collect electric vehicle charging data sets in real time, and a database is built to store the data, and then the stored data is pre-processed;
[0130] S2. By constructing a charging load prediction model, setting a time series analysis algorithm model and a spatial analysis algorithm model, and calculating and obtaining a charging demand prediction value P(t) and a charging location demand prediction value P(x, y) based on the collected electric vehicle charging data set;
[0131] S3, then correlate the obtained charging demand prediction value P(t) and the charging location demand prediction value P(x, y) to obtain a comprehensive charging demand prediction value P(x, y, t);
[0132] S4. By comprehensively analyzing the charging load demand, a preliminary comparative evaluation is performed between the preset site selection evaluation threshold F1 and the obtained comprehensive charging demand prediction value P (x, y, t) to determine the rationality of the site selection plan;
[0133] S5. Finally, in the case of unreasonable site selection, a charging station site selection optimization model is constructed, the objective function U is set, and the optimization algorithm is used to solve and obtain the individual fitness value U ( ), and set the operation threshold F2 to compare and evaluate with the solved objective function U to determine the charging station site selection plan.
[0134] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.
Claims
1. An electric vehicle charging load prediction system based on spatiotemporal characteristics, characterized in that: It includes data acquisition module, spatiotemporal characteristic analysis module, spatiotemporal joint prediction module, site selection evaluation module and site selection optimization module; The data acquisition module is used to set up an integrated API application program interface set, collect electric vehicle charging data sets in real time, build a database for storage, and then pre-process the stored data; The spatiotemporal characteristic analysis module is used to construct a charging load prediction model, set a time series analysis algorithm model and a spatial analysis algorithm model, and calculate and obtain a charging demand prediction value P(t) and a charging location demand prediction value P(x, y) based on the collected electric vehicle charging data set; The spatiotemporal characteristic analysis module includes a time series analysis unit and a space analysis unit; The timing analysis unit is used to construct a timing analysis algorithm model, then extract the charging feature vector X from the timing database, and then use the historical charging data to train the timing analysis algorithm model to analyze the demand regression coefficient. , based on the charging characteristic vector X and the demand regression coefficient , construct the time series analysis algorithm formula, calculate and obtain the charging demand prediction value P(t), and predict the future charging demand; The charging demand prediction value P(t) is obtained by the following timing analysis algorithm formula: ; In the formula, represents the first constant term, which represents the baseline charging demand, that is, the charging demand when there are no other influencing factors, It is used to measure the influence of historical charging feature vectors on the predicted value. T represents the total number of lag periods, o represents the number of lag periods, Xi represents the demand for the i-th charging feature vector X, represents the i-th charging characteristic vector X demand at time to, represents the error term; The spatial analysis unit is used to construct a spatial analysis algorithm model, then extract the spatial influence vector J from the time series database, train the spatial algorithm model, and obtain the spatial regression coefficient , and combined with the spatial influence vector J and the spatial regression coefficient , construct a spatial analysis algorithm formula, and calculate and obtain the charging location demand prediction value P (x, y); The charging location demand prediction value P (x, y) is calculated and obtained by the following spatial analysis algorithm formula: ; In the formula, represents the second constant term, (x, y) represents longitude and latitude, represents the spatial regression coefficient, which measures the influence of each spatial influence vector J on the charging demand. Indicates location The related spatial influence vector J, Jj represents the j-th spatial influence vector, represents the first error term, which indicates the unexplained part of the prediction, and m represents the total number of spatial influencing factors that affect the charging demand at location (x, y). Each influencing factor has a corresponding spatial regression coefficient ; The spatiotemporal joint prediction module is used to correlate the acquired charging demand prediction value P(t) and the charging location demand prediction value P(x, y) to obtain a comprehensive charging demand prediction value P(x, y, t); The spatiotemporal joint prediction module is used to construct a spatiotemporal joint prediction algorithm model, and performs combined calculation to obtain a comprehensive charging demand prediction value P (x, y, t) based on the obtained charging demand prediction value P (t) and the charging location demand prediction value P (x, y); The comprehensive charging demand prediction value P (x, y, t) is calculated by the following algorithm formula: ; In the formula, represents the third constant term, represents the baseline charging demand, P(x, y, t) represents the comprehensive charging demand prediction value at time t and location (x, y), represents the second error term; The site selection evaluation module comprehensively analyzes the charging load demand, performs a preliminary comparative evaluation between the preset site selection evaluation threshold F1 and the obtained comprehensive charging demand prediction value P (x, y, t), and determines the rationality of the site selection scheme; The site selection optimization module is used to construct a charging station site selection optimization model when the site selection is unreasonable, set the objective function U, and use the optimization algorithm to obtain the variant individual , and substitute it into the objective function U to calculate and obtain the individual fitness value U ( ), and at the same time set the operation threshold F2 to compare and evaluate with the solved objective function U to determine the charging station site selection plan; The site selection optimization module includes an objective function analysis unit, an optimization calculation unit and an evaluation iteration unit; The objective function analysis unit is used to construct an objective function U to sum up the utility of the site k based on the obtained comprehensive charging demand forecast value P (x, y, t) and the operating cost data Ck when it is determined that the current site selection is unreasonable; The objective function U is calculated by the following algorithm formula: ; In the formula, wk represents the weight value, which indicates the importance of the charging station location k. Indicates the location of site k and the predicted value of charging demand at time t, K represents the total number of candidate charging station locations; The optimization calculation unit is used to reproduce according to the objective function U to generate two individuals and , exchange some of the site selection point k information, set the intersection point c to generate a new individual and , for the newly generated individuals and Some site selection points k are randomly changed to obtain mutant individuals , and re-substitute it into the objective function U to calculate and analyze the fitness value U of each individual ( ); The individual fitness value U ( ) is obtained through the following algorithm formula: ; In the formula, Indicates the location of the kth charging station in the ath candidate solution The predicted value of charging demand at time t, Represents the result after the ath individual undergoes mutation operation.
2. The electric vehicle charging load prediction system based on spatiotemporal characteristics according to claim 1 is characterized in that: The data acquisition module includes a data acquisition unit, an operation cost analysis unit and a data storage and preprocessing unit; The data collection unit is used to set an integrated API application program interface set, which is respectively connected to the charging pile management system, the vehicle networking platform, the census system, the traffic management system, the meteorological system and the GIS geographic information system, and is set to collect charging data, vehicle data, population data, traffic data and weather data every 12 hours, and summarize and generate an electric vehicle charging data set; The charging data includes time, charging location, charging duration and charging amount; The vehicle data includes parking location, GPS location and driving route; The population data include residential population, working population and total population density; The traffic data includes road traffic volume and the number of parking lots in public parking lots; The weather data includes vehicle travel volume and charging demand under different weather conditions; The operating cost analysis unit is used to obtain and analyze the geographic information of the site k through the GIS geographic information system, and at the same time collect land purchase or lease costs, construction costs, equipment purchase prices, electricity costs, maintenance costs, labor costs, operating costs and construction costs in combination with big data, and use linear regression, decision tree and random forest models to build a cost evaluation model, and then use AnyLogic simulation software to simulate the cost situation, and then use the ARIMA model to predict long-term costs and benefits to obtain operating cost data Ck; The data storage and preprocessing unit is used to clean and process the collected electric vehicle charging data set, including processing missing values, outliers and standardization of data formats, and then extract features of the electric vehicle charging data set, and extract the charging feature vector X and spatial influence vector J of the electric vehicle based on the charging data, vehicle data, population data, traffic data and weather data in the electric vehicle charging data set; The charging feature vector X={charging amount X1, charging time X2, charging times X3, charging cost X4, charging start time X5, charging end time X6} The spatial influence vector J={total population density J1, road traffic volume J2, vehicle distribution density J3, number of parking lots J3, parking lot capacity J4, average vehicle travel rate under the influence of weather variables J5, average charging demand rate under the influence of weather variables J6}; then a time series database is constructed, and the charging feature vector X and the spatial influence vector J after feature extraction are marked with timestamps, and sent to the time series database through wireless transmission. According to the marked timestamps, the charging feature vector X and the spatial influence vector J are sorted and stored according to the order of the timestamps, and the data stored for 24 hours is transferred to historical data and backed up.
3. The electric vehicle charging load prediction system based on spatiotemporal characteristics according to claim 1 is characterized in that: The site selection evaluation module obtains the site selection evaluation threshold F1 by performing a comprehensive average calculation on the demand coverage rate, utilization rate, service radius and cost-effectiveness ratio of the charging piles, and then performs a preliminary comparative evaluation with the obtained comprehensive charging demand prediction value P (x, y, t) to analyze the rationality of the site selection plan; The specific evaluation plan is as follows: When the comprehensive charging demand prediction value P (x, y, t) > the site selection evaluation threshold F1, it means that the existing charging station load abnormality site selection is reasonable. At this time, the current site selection is planned to plan the electric vehicle power load supply; When the comprehensive charging demand prediction value P (x, y, t) ≤ the site selection evaluation threshold F1, it means that the normal site selection of the existing charging station load is unreasonable. At this time, the current site selection is not selected and the site selection optimization is performed again.
4. The electric vehicle charging load prediction system based on spatiotemporal characteristics according to claim 1 is characterized in that: The evaluation iteration unit presets the operation threshold F2 based on the operation cost and then compares it with the obtained individual fitness value U ( ) to conduct comparative evaluation, analyze the operation status of the electric vehicle charging load at the optimized site k, and generate an iterative mechanism; When the individual fitness value U( ) ≥ operation threshold F2, it means that the total utility of the current site k is normal; When the individual fitness value U( ) < operation threshold F2, indicating that the total utility of the current site k is abnormal. At this time, the iteration mechanism is executed to regenerate two individuals for mutation and calculate and analyze the fitness value U of each new individual ( ).
5. A method for predicting electric vehicle charging load based on spatiotemporal characteristics, applied to an electric vehicle charging load prediction system based on spatiotemporal characteristics as claimed in any one of claims 1 to 4, characterized in that: The following steps are involved: S1. First, an integrated API application program interface set is set up to collect electric vehicle charging data sets in real time, and a database is built to store the data, and then the stored data is pre-processed; S2. By constructing a charging load prediction model, setting a time series analysis algorithm model and a spatial analysis algorithm model, and calculating and obtaining a charging demand prediction value P(t) and a charging location demand prediction value P(x, y) based on the collected electric vehicle charging data set; S3, then correlate the obtained charging demand prediction value P(t) and the charging location demand prediction value P(x, y) to obtain a comprehensive charging demand prediction value P(x, y, t); S4. By comprehensively analyzing the charging load demand, a preliminary comparative evaluation is performed between the preset site selection evaluation threshold F1 and the obtained comprehensive charging demand prediction value P (x, y, t) to determine the rationality of the site selection plan; S5. Finally, in the case of unreasonable site selection, a charging station site selection optimization model is constructed, the objective function U is set, and the optimization algorithm is used to solve and obtain the individual fitness value U ( ), and set the operation threshold F2 to compare and evaluate with the solved objective function U to determine the charging station site selection plan.
