Electric vehicle charging strategy construction method and system considering user charging behavior

Abstract

The invention discloses an electric vehicle charging strategy construction method and system considering user charging behaviors, and relates to the technical field of electric vehicle charging. The method comprises the following steps: firstly, acquiring original multi-dimensional data including charging load, travel time, weather and the like; repairing charging load data through structured tensor parallel factorization, and predicting travel time by using an xLSTM model after dimension reduction; and based on the complete load data and the travel time prediction result, outputting optimal charging power distribution and each time step charging decision by adopting a two-layer optimization strategy combined with a sparrow search algorithm. Power grid overload is avoided through multi-dimensional constraint, the user satisfaction is taken as an optimization target, the number of vehicles which do not reach the target SOC is remarkably reduced, the user charging experience is improved, and stable operation of the power grid is guaranteed.

Description

Technical Field

[0001] This application relates to the field of electric vehicle charging technology, specifically to a method and system for constructing an electric vehicle charging strategy that takes into account user charging behavior. Background Technology

[0002] In recent years, the global new energy vehicle industry has developed rapidly, and the number of electric vehicles on the road has increased significantly. Correspondingly, the demand for electricity for electric vehicle charging has also grown substantially. This growth trend has placed a heavy load on existing power generation and supply infrastructure.

[0003] Currently, electric vehicle (EV) users' charging behavior exhibits a significant concentration during peak hours, often coinciding with peak residential electricity consumption periods. This overlap easily leads to overload of the power distribution network, affecting not only the reliability of the power grid but also disrupting normal residential and commercial electricity use. Simultaneously, the increasing number of EVs is driving up the demand for power distribution network capacity. However, given the current fixed capacity of the existing power distribution network, the growing number of EVs will directly result in some EVs not being able to complete a full charge, significantly reducing the charging experience and efficiency. Particularly noteworthy is that the initial design of existing residential power distribution networks did not adequately consider the charging load demands of EVs, resulting in relatively low distribution capacity and making residential areas a high-risk area for incomplete charging.

[0004] Therefore, how to scientifically and rationally manage the charging behavior of electric vehicles and coordinate the contradiction between charging demand and grid carrying capacity has become an urgent technical problem to be solved. Summary of the Invention

[0005] In view of the technical problems mentioned in the background, the purpose of this invention is to provide a method and system for constructing electric vehicle charging strategies that take into account user charging behavior.

[0006] To achieve the objectives of this invention, the technical solution provided by this invention is as follows: First aspect This application provides a method for constructing an electric vehicle charging strategy that takes into account user charging behavior, including the following steps: Step 1: Collect raw multidimensional data; The original multidimensional data includes electric vehicle charging load data, historical observation data of electric vehicle arrival time and historical observation data of electric vehicle departure time, meteorological data, battery state of charge data, and target SOC; Step 2: Using the electric vehicle load data repair model based on structured tensor parallel factor decomposition, repair the electric vehicle charging load data in the original multidimensional data to obtain complete electric vehicle charging load data. Step 3: Perform dimensionality reduction processing on the original multidimensional data to obtain dimensionality-reduced data; based on the dimensionality-reduced data, use the electric vehicle travel time prediction model based on xLSTM to make predictions and obtain electric vehicle travel time prediction data. Step 4: Based on the complete electric vehicle charging load data and the electric vehicle travel time prediction data, a two-layer optimization strategy combining the sparrow search algorithm is used to calculate the optimal charging power allocation for each electric vehicle and the charging or not charging decision for each electric vehicle at each time step.

[0007] Second aspect This application provides an electric vehicle charging strategy construction system that takes into account user charging behavior, used to implement the electric vehicle charging strategy construction method that takes into account user charging behavior, including the following units: a data acquisition unit, a data repair unit, a prediction data calculation unit, and a two-layer optimization strategy calculation unit. The data acquisition unit is used to acquire raw multidimensional data; The original multidimensional data includes electric vehicle charging load data, historical observation data of electric vehicle arrival time and historical observation data of electric vehicle departure time, meteorological data, battery state of charge data, and target SOC; The data repair unit is used to repair the electric vehicle charging load data in the original multidimensional data using an electric vehicle load data repair model based on structured tensor parallel factor decomposition, so as to obtain complete electric vehicle charging load data. The prediction data calculation unit is used to perform dimensionality reduction processing on the original multidimensional data to obtain dimensionality-reduced data; based on the dimensionality-reduced data, an electric vehicle travel time prediction model based on xLSTM is used to make predictions to obtain electric vehicle travel time prediction data. The two-layer optimization strategy calculation unit is used to calculate the optimal charging power allocation for each electric vehicle and the charging or not charging decision for each electric vehicle at each time step based on the complete electric vehicle charging load data and the electric vehicle travel time prediction data, using a two-layer optimization strategy combined with the sparrow search algorithm.

[0008] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention employs a two-layer optimization strategy that incorporates load constraints from the residential power distribution network. By constraining charging volume and charging power across multiple dimensions, it strictly controls the total charging load to ensure it does not exceed the grid's capacity limit. This effectively addresses the grid overload problem caused by the overlap of peak electric vehicle charging and residential electricity consumption, guaranteeing stable grid operation and normal electricity use for residents. Furthermore, with user satisfaction as the core optimization objective, it achieves precise decision-making through a two-layer progressive optimization: the first layer determines the optimal charging power allocation, ensuring that vehicles participating in the scheduling maximize their charging needs under power constraints; the second layer clarifies the charging / non-charging decision at each time step, adapting to the predicted vehicle departure time.

[0009] Experimental results show that, compared to the traditional grid selection method GPM, this scheme reduces the number of electric vehicles that cannot reach the target SOC by 50%, increases overall user satisfaction by 2.77%, and significantly improves the user charging experience. Attached Figure Description

[0010] Figure 1 A schematic flowchart of an electric vehicle charging strategy construction method that takes into account user charging behavior provided in an embodiment of the present invention; Figure 2 The figure shows the experimental results comparing the method of this invention with those of the grid selection method (GPM). Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0012] like Figure 1 As shown, this embodiment provides a method for constructing an electric vehicle charging strategy that takes into account user charging behavior, including the following steps: Step 1: Collect raw multidimensional data; The original multidimensional data includes electric vehicle charging load data, historical observation data of electric vehicle arrival time and historical observation data of electric vehicle departure time, meteorological data, battery state of charge data, and target SOC; Step 2: Using the electric vehicle load data repair model based on structured tensor parallel factor decomposition, repair the electric vehicle charging load data in the original multidimensional data to obtain complete electric vehicle charging load data. It should be noted that electric vehicle charging load data is correlated with historical observation data of electric vehicle arrival time and departure time, meteorological data (temperature, humidity, air pressure), battery state of charge data throughout the charging period, target SOC and other multi-dimensional data. Therefore, an electric vehicle charging load data repair method based on multi-dimensional tensor parallel factor decomposition (dataset dimension is 11520×1) is adopted to obtain high-quality complete electric vehicle charging load data.

[0013] The objective function of the electric vehicle load data repair model based on structured tensor parallel factor decomposition is as follows: ; In the formula, It is the repaired data tensor used to represent the predicted charging load data; g r Indicates the adaptive calculation factor. It is the first factor in the factor matrix. It is the second factor in the factor matrix. It is the Nth factor in the factor matrix; A tensor representing the original electric vehicle load data; This represents the Frobenius norm, used to calculate the error of a matrix or tensor. This represents the observation portion of the data; These are used to control parameters during tensor decomposition. It is a factor matrix in tensor decomposition, used as a low-dimensional representation of charging load data; This is the error term.

[0014] The objective function of the electric vehicle load data repair model based on structured tensor parallel factor decomposition is solved using hierarchical least squares method to obtain complete electric vehicle charging load data. It should be noted that using hierarchical least squares method is a prior art solution, and its principle will not be detailed here.

[0015] Step 3: Perform dimensionality reduction processing on the original multidimensional data to obtain dimensionality-reduced data; based on the dimensionality-reduced data, use the electric vehicle travel time prediction model based on xLSTM to make predictions and obtain electric vehicle travel time prediction data. It should be noted that considering the long training time of multidimensional data input into neural network models, we first use the existing dimensionality reduction method based on elastic network regression to reduce the amount of input data, and then establish an electric vehicle travel time prediction model based on xLSTM.

[0016] It should be noted that the electric vehicle travel time prediction model based on xLSTM includes: an input feature construction module, an mLSTM sub-network, an sLSTM sub-network, a feature fusion module, and a regression output module. The input feature construction module is used to construct an input sequence of length T from the dimensionality-reduced data in chronological order. The mLSTM subnetwork is used to perform temporal modeling on the input sequence and output the first hidden feature sequence; The sLSTM subnetwork is used to perform temporal modeling on the input sequence and output a second hidden feature sequence; The feature fusion module is used to concatenate or weightedly fuse the first hidden feature sequence and the second hidden feature sequence to obtain fused features; The regression output module is used to output electric vehicle travel time prediction data based on the fusion features.

[0017] Step 4: Based on the complete electric vehicle charging load data and the electric vehicle travel time prediction data, a two-layer optimization strategy combining the sparrow search algorithm is used to calculate the optimal charging power allocation for each electric vehicle and the charging or not charging decision for each electric vehicle at each time step.

[0018] It should be noted that in existing charging solutions, meeting users' charging needs is the primary prerequisite for all charging strategies. However, with the increasing number of electric vehicles, meeting users' charging needs is no longer the primary prerequisite, especially in older residential areas. The two-layer optimization strategy proposed in this invention uses user satisfaction as the global optimization objective in the calculation.

[0019] In the first optimization strategy, the power allocation for each electric vehicle is treated as a variable. To meet the constraints of safe operation of the power distribution network, assuming no more vehicles will be added, the optimal charging power allocation for each electric vehicle is obtained with the total user satisfaction of existing vehicles as the objective function.

[0020] In the second-level optimization strategy, the results of the first optimization strategy are combined to participate in the current charging cycle. The average user satisfaction rate of the number of vehicles is used as the objective function to obtain the charging or not charging decision for each electric vehicle at each time step.

[0021] Specifically, the two-layer optimization strategy combining the sparrow search algorithm includes a first-layer optimization strategy and a second-layer optimization strategy; The first-level optimization strategy includes the following: (1.1) First decision variable Q t First decision variable Qt It is time t Time to give N t The power allocated to the vehicle is calculated using the following formula: ; In the formula, Q t j It is time t Vehicle j The power allocation value, j The value is 1- N t , N t It is time t The number of vehicles willing to participate in optimization at that time , It is time t The total number of vehicles at that time It is time t The number of vehicles not participating in optimization; (1.2) First objective function The first objective function is time. t The formula for overall user satisfaction is as follows: ; In the formula, S j It is a vehicle j The state of satisfaction. ,when F j When it exceeds 80%, a single user S j Set to 1, F j yes and proportion, ; C j It is a vehicle j Battery capacity, It is a vehicle j The charging needs, It is a vehicle j Target SOC It is a vehicle j The initial SOC; (1.3) The first model constraint includes the following: (1.3.1) The charging amount constraint is: ; in, For time tAfter deducting the base load and the load of electric vehicles not participating in dispatch, the remaining allocable charging power of the residential power distribution network that can be used for dispatched vehicles is given by the following formula: ; In the formula, This is the upper limit of the residential power distribution network; This is the predicted value of the base load for this charging cycle. It is the total charging load of electric vehicles that do not participate in the scheduling. The total charging load of electric vehicles that do not participate in the scheduling is the sum of the load corresponding to the set of vehicles that do not participate in the scheduling, which is obtained from the complete electric vehicle charging load data according to the user's wishes. It is derived data formed by filtering the complete charging load data according to "vehicles that do not participate in the scheduling" and summarizing it hourly in the time dimension. yes The average number of vehicles included in the user's predicted departure time is calculated using the following formula: ; In the formula, It is a vehicle j The predicted departure time is obtained from electric vehicle travel time prediction data; in, The calculation formula is as follows: ; In the formula, For time i EV loads that do not participate in scheduling This is the start time of this charging cycle. This is the end time of this charging cycle; ; In the formula, A not It is a sequence of EV numbers that does not participate in scheduling. It is a vehicle j The charging power; (1.3.2) The charging power constraint is: ; (1.3.3) The charging demand constraint is: ; In the formula, For the target SOC of this charging cycle, This represents the initial SOC for this charging cycle. (1.4) Solving the problem using the sparrow search algorithm. Q t As decision variables, with Given an optimization problem with an objective function that satisfies the first model constraint, we obtain: ,according to To allocate the optimal charging power to each electric vehicle; The second-level optimization strategy includes the following: (2.1) Second decision variable The details are as follows: ; in, w t It is a matrix of decision variables. Indicates vehicle j In time i Charge, Indicates vehicle j In time i Not charged ; (2.2) Second objective function The second objective function is To participate in the current charging cycle The average user satisfaction rate based on the number of vehicles is calculated using the following formula: ; ; ; In the formula, Indicates vehicle The amount of electricity / charge that can be obtained within the charging time window, where the charging time window refers to the period from time... To its predicted departure time Obtained from electric vehicle travel time prediction data; Indicates vehicle The target electricity demand; For vehicles The charging completion rate, among which, This refers to the amount of charge per unit time step or the increase in charge corresponding to the charging power. For single-bike satisfaction, a threshold rule is used to determine: when Time to take Otherwise take ; (2.3) The second model constraints include the following: (2.3.1) Obtaining energy constraints: ; (2.3.2) Total power allocation constraint: ; In the formula, It is timei Basic load value at time, It is time i The total charging load of electric vehicles that do not participate in the scheduling process; (2.4) Solving the problem using the Sparrow Search algorithm As decision variables, with Given an optimization problem with an objective function that satisfies the second model constraint, we obtain: ,according to The system obtains the charging or non-charging decision for each electric vehicle at each time step.

[0022] This invention was experimentally verified using a real dataset of electric vehicles from a northern city and employed a local time-of-use electricity pricing mechanism. A comparative experiment was conducted using the grid selection method (GPM) for comparison. Figure 2 The figure shown is a comparison of experimental results between the method of this invention and the grid selection method (GPM). Figure 2 In the case of using the method of the present invention, 4 electric vehicles failed to reach the target SOC, while using the GPM method, 8 electric vehicles failed to reach the target SOC, and 2 electric vehicles deviated from the target SOC.

[0023] In addition, calculations show that the method of this invention improves overall user satisfaction (GPM) by 2.77% compared to the comparative method.

[0024] In addition, this embodiment also provides an electric vehicle charging strategy construction system that takes into account user charging behavior, used to implement the electric vehicle charging strategy construction method that takes into account user charging behavior, including the following units: data acquisition unit, data repair unit, prediction data calculation unit, and two-layer optimization strategy calculation unit; The data acquisition unit is used to acquire raw multidimensional data; The original multidimensional data includes electric vehicle charging load data, historical observation data of electric vehicle arrival time and historical observation data of electric vehicle departure time, meteorological data, battery state of charge data, and target SOC; The data repair unit is used to repair the electric vehicle charging load data in the original multidimensional data using an electric vehicle load data repair model based on structured tensor parallel factor decomposition, so as to obtain complete electric vehicle charging load data. The prediction data calculation unit is used to perform dimensionality reduction processing on the original multidimensional data to obtain dimensionality-reduced data; based on the dimensionality-reduced data, an electric vehicle travel time prediction model based on xLSTM is used to make predictions to obtain electric vehicle travel time prediction data. The two-layer optimization strategy calculation unit is used to calculate the optimal charging power allocation for each electric vehicle and the charging or not charging decision for each electric vehicle at each time step based on the complete electric vehicle charging load data and the electric vehicle travel time prediction data, using a two-layer optimization strategy combined with the sparrow search algorithm.

[0025] The objective function of the electric vehicle load data repair model based on structured tensor parallel factor decomposition is as follows: ; In the formula, It is the repaired data tensor used to represent the predicted charging load data; g r Indicates the adaptive calculation factor. It is the first factor in the factor matrix. It is the second factor in the factor matrix. It is the Nth factor in the factor matrix; A tensor representing the original electric vehicle load data; This represents the Frobenius norm, used to calculate the error of a matrix or tensor. This represents the observation portion of the data; These are used to control parameters during tensor decomposition. It is a factor matrix in tensor decomposition, used as a low-dimensional representation of charging load data; This is the error term.

[0026] Specifically, the objective function of the electric vehicle load data repair model based on structured tensor parallel factor decomposition is solved using the hierarchical least squares method to obtain complete electric vehicle charging load data.

[0027] The electric vehicle travel time prediction model based on xLSTM includes: an input feature construction module, an mLSTM sub-network, an sLSTM sub-network, a feature fusion module, and a regression output module. The input feature construction module is used to construct an input sequence of length T from the dimensionality-reduced data in chronological order. The mLSTM subnetwork is used to perform temporal modeling on the input sequence and output the first hidden feature sequence; The sLSTM subnetwork is used to perform temporal modeling on the input sequence and output a second hidden feature sequence; The feature fusion module is used to concatenate or weightedly fuse the first hidden feature sequence and the second hidden feature sequence to obtain fused features; The regression output module is used to output electric vehicle travel time prediction data based on the fusion features.

[0028] The two-layer optimization strategy combining the sparrow search algorithm includes a first-layer optimization strategy and a second-layer optimization strategy. The first-level optimization strategy includes the following: (1.1) First decision variable Q t First decision variable Q t It is time t Time to give N t The power allocated to the vehicle is calculated using the following formula: ; In the formula, Q t j It is time t Vehicle j The power allocation value, j The value is 1- N t , N t It is time t The number of vehicles willing to participate in optimization at that time , It is time t The total number of vehicles at that time It is time t The number of vehicles not participating in optimization; (1.2) First objective function The first objective function is time. t The formula for overall user satisfaction is as follows: ; In the formula, S j It is a vehicle j The state of satisfaction. ,when F j When it exceeds 80%, a single user S j Set to 1, F j yes and proportion, ; C j It is a vehicle j Battery capacity, It is a vehicle j The charging needs, It is a vehicle j Target SOC It is a vehicle jThe initial SOC; (1.3) The first model constraint includes the following: (1.3.1) The charging amount constraint is: ; in, For time t After deducting the base load and the load of electric vehicles not participating in dispatch, the remaining allocable charging power of the residential power distribution network that can be used for dispatched vehicles is given by the following formula: ; In the formula, This is the upper limit of the residential power distribution network; This is the predicted value of the base load for this charging cycle. It is the total charging load of electric vehicles that do not participate in the scheduling. The total charging load of electric vehicles that do not participate in the scheduling is the sum of the load corresponding to the set of vehicles that do not participate in the scheduling, which is obtained from the complete electric vehicle charging load data according to the user's wishes. yes The average number of vehicles included in the user's predicted departure time is calculated using the following formula: ; In the formula, It is a vehicle j The predicted departure time is obtained from electric vehicle travel time prediction data; in, The calculation formula is as follows: ; In the formula, For time i EV loads that do not participate in scheduling This is the start time of this charging cycle. This is the end time of this charging cycle; ; In the formula, A not It is a sequence of EV numbers that does not participate in scheduling. It is a vehicle j The charging power; (1.3.2) The charging power constraint is: ; (1.3.3) The charging demand constraint is: ; In the formula, For the target SOC of this charging cycle, This represents the initial SOC for this charging cycle. (1.4) Solving the problem using the sparrow search algorithm. Q t As decision variables, with Given an optimization problem with an objective function that satisfies the first model constraint, we obtain: ,according to To allocate the optimal charging power to each electric vehicle; The second-level optimization strategy includes the following: (2.1) Second decision variable The details are as follows: ; in, w t It is a matrix of decision variables. Indicates vehicle j In time i Charge, Indicates vehicle j In time i Not charged ; (2.2) Second objective function The second objective function is To participate in the current charging cycle The average user satisfaction rate based on the number of vehicles is calculated using the following formula: ; ; ; In the formula, Indicates vehicle The amount of electricity / charge that can be obtained within the charging time window, where the charging time window refers to the period from time... To its predicted departure time Obtained from electric vehicle travel time prediction data; Indicates vehicle The target electricity demand; For vehicles The charging completion rate, among which, This refers to the amount of charge per unit time step or the increase in charge corresponding to the charging power. For single-bike satisfaction, a threshold rule is used to determine: when Time to take Otherwise take ; (2.3) The second model constraints include the following: (2.3.1) Obtaining energy constraints: ; (2.3.2) Total power allocation constraint: ; In the formula, It is time i Basic load value at time, It is time i The total charging load of electric vehicles that do not participate in the scheduling process; (2.4) Solving the problem using the Sparrow Search algorithm As decision variables, with Given an optimization problem with an objective function that satisfies the second model constraint, we obtain: ,according to The system obtains the charging or non-charging decision for each electric vehicle at each time step.

[0029] Finally, it should be noted that the above embodiments are merely illustrative and explanatory of the present invention, and are not intended to limit the present invention to the scope of the described embodiments. Furthermore, those skilled in the art will understand that the present invention is not limited to the above embodiments, and many more variations and modifications can be made based on the teachings of the present invention, all of which fall within the scope of protection claimed by the present invention.

Claims

1. A method for constructing an electric vehicle charging strategy considering user charging behavior, characterized in that, The method comprises the following steps: Step 1: collecting original multidimensional data; The original multidimensional data comprises electric vehicle charging load data, electric vehicle arrival time historical observation data and electric vehicle departure time historical observation data, meteorological data, battery state of charge data and target SOC; Step 2: repairing the electric vehicle charging load data in the original multidimensional data by using an electric vehicle load data repair model based on structured tensor parallel factorization to obtain complete electric vehicle charging load data; Step 3: performing dimension reduction processing on the original multidimensional data to obtain reduced data; based on the reduced data, an electric vehicle travel time prediction model based on xLSTM is used for prediction to obtain electric vehicle travel time prediction data; Step 4: based on the complete electric vehicle charging load data and the electric vehicle travel time prediction data, a two-layer optimization strategy combined with a sparrow search algorithm is used to calculate the optimal charging power distribution of each electric vehicle and the charging or non-charging decision of each electric vehicle at each time step. 2.The method of claim 1, wherein, In step 2, the objective function of the electric vehicle load data repair model based on structured tensor parallel factorization is as follows: ； wherein, is the repaired data tensor, which is used to represent the predicted charging load data; g r denotes an adaptive computation factor, is the 1st factor of the factor matrix, is the 2nd factor of the factor matrix, is the Nth factor of the factor matrix; denotes the tensor of original electric vehicle load data; denotes the Frobenius norm, which is used to calculate the error of a matrix or tensor; denotes the observed part of the data; is a parameter used to control the tensor decomposition process; is the factor matrix in the tensor decomposition, which is used as a low-dimensional representation of the charging load data; is the error term. 3.The method of claim 2, wherein, The objective function of the electric vehicle load data repair model based on structured tensor parallel factorization is solved by using hierarchical least squares to obtain complete electric vehicle charging load data.

4. The method of claim 3, wherein the charging strategy is constructed by considering the user's charging behavior. In step 3, the electric vehicle travel time prediction model based on xLSTM comprises an input feature construction module, an mLSTM subnetwork, an sLSTM subnetwork, a feature fusion module and a regression output module; The input feature construction module is used to construct an input sequence with a length of T in chronological order based on the reduced data; The mLSTM subnetwork is used to perform time series modeling on the input sequence and output a first hidden feature sequence; The sLSTM subnetwork is used to perform time series modeling on the input sequence and output a second hidden feature sequence; The feature fusion module is used to splice or weightedly fuse the first hidden feature sequence and the second hidden feature sequence to obtain fused features; The regression output module is used to output electric vehicle travel time prediction data based on the fused features.

5. The method of claim 4, wherein the charging strategy is constructed by considering the user's charging behavior. In step 4, the two-layer optimization strategy combined with the sparrow search algorithm comprises a first-layer optimization strategy and a second-layer optimization strategy; The first-layer optimization strategy comprises the following: (1.1) first decision variable Q t : first decision variable Q t is time t to give N t power allocated to the vehicle, as follows: ； wherein Q t j is the time t is the power distribution value of the vehicle j at time j is 1- N t , N t is the time t is the number of vehicles willing to participate in the optimization at time , is the total number of vehicles at time t is the number of vehicles not participating in the optimization at time is the time t ; (1.2) the first objective function : the first objective function is time t user total satisfaction, which is given by the following formula: ； wherein S j is a satisfaction state of the vehicle j , when F j is greater than 80%, a single user S j is set to 1, F j is a ratio of , ; C j is a battery capacity of the vehicle j , is a charging demand of the vehicle j , is a target SOC of the vehicle j , is an initial SOC of the vehicle j ; (1.3) first model constraints, comprising the following: (1.3.1) charging quantity constraint is: ； wherein, is the time t The remaining allocable charging power of the residential power distribution network for participating in dispatching vehicles after deducting the basic load and the load of the electric vehicles not participating in dispatching is as follows: ； wherein is the upper limit of the residential power distribution network; is the predicted value of the basic load of the current charging period, is the total charging load of the electric vehicles not participating in the dispatch, which is the load summary amount corresponding to the set of vehicles not participating in the dispatch, obtained by dividing the complete electric vehicle charging load data according to the user's intention; is is the average value of the number of vehicles participating in the user's predicted departure time, and the calculation formula is as follows: ； In the formula, is a predicted departure time of the vehicle j from the electric vehicle trip time prediction data; wherein The calculation formula of the following: The calculation formula of the following: ； wherein is time i EV load not participating in dispatch, is the start time of the current charging period, is the end time of the present charging cycle; ； wherein A not is a sequence of EV numbers that do not participate in scheduling, is a vehicle j charging power; (1.3.2) charging power constraint is: ； (1.3.3) charging demand constraint is: ； wherein is the target SOC for the present charging cycle, is the initial SOC for the present charging cycle; (1.4) solving the optimization problem with sparrow search algorithm Q t as decision variables, with as the objective function, and satisfying the first model constraints, obtaining , and distributing the optimal charging power for each electric vehicle according to ; The second-layer optimization strategy comprises the following: (2.1) Second decision variable , as follows: ； wherein, w t is a matrix of decision variables, denotes a vehicle j at time i charged, denotes a vehicle j at time i not charged ; (2.2) Second objective function The second objective function is To participate in the current charging cycle The average user satisfaction rate based on the number of vehicles is calculated using the following formula: ； ； ； In the formula, represents the vehicle The amount of electricity / charge that can be obtained within the charging time window, which refers to the time to its predicted departure time , from the electric vehicle travel time prediction data; represents the target demand of the vehicle ; is the charging completion degree of the vehicle , wherein is the charging power or the amount of electricity corresponding to the charging power per unit time step; is the single-vehicle satisfaction, determined by the threshold rule: when , take , otherwise take ; (2.3) second model constraints, comprising the following: (2.3.1) energy acquisition constraint: ； (2.3.2) total amount of distributed power constraint: ； In the formula, is the time i the basic load value at the time, is the time i the total charging load of the electric vehicles not participating in the dispatch at the time; (2.4) Solving the problem using the Sparrow Search algorithm As decision variables, with Given an optimization problem with an objective function that satisfies the second model constraint, we obtain: ,according to The system obtains the charging or non-charging decision for each electric vehicle at each time step. 6.A system for constructing an electric vehicle charging strategy considering user charging behavior, for implementing the method for constructing an electric vehicle charging strategy considering user charging behavior according to any one of claims 1-5, characterized in that, It comprises the following units: data acquisition unit, data repair unit, prediction data calculation unit and two-layer optimization strategy calculation unit; The data acquisition unit is used to acquire original multidimensional data; The original multi-dimensional data includes electric vehicle charging load data, electric vehicle arrival time historical observation data and electric vehicle departure time historical observation data, meteorological data, battery state of charge data, and target SOC; The data repairing unit is configured to repair the electric vehicle charging load data in the original multi-dimensional data by using an electric vehicle load data repairing model based on structured tensor parallel factorization to obtain complete electric vehicle charging load data. The prediction data calculation unit is configured to perform dimension reduction processing on the original multi-dimensional data to obtain reduced dimension data, and perform prediction based on the reduced dimension data by using an electric vehicle travel time prediction model based on xLSTM to obtain electric vehicle travel time prediction data. The two-layer optimization strategy calculation unit is configured to calculate the optimal charging power distribution of each electric vehicle and the charging or non-charging decision of each electric vehicle at each time step based on the complete electric vehicle charging load data and the electric vehicle travel time prediction data by using a two-layer optimization strategy combined with a sparrow search algorithm.

7. The system for constructing electric vehicle charging strategy considering user charging behavior according to claim 6, wherein, The objective function of the electric vehicle load data repairing model based on structured tensor parallel factorization is as follows: ； wherein, is the repaired data tensor, which is used to represent the predicted charging load data; g r denotes an adaptive computation factor, is the 1st factor of the factor matrix, is the 2nd factor of the factor matrix, is the Nth factor of the factor matrix; denotes the tensor of the original electric vehicle load data; denotes the Frobenius norm, which is used to calculate the error of a matrix or a tensor; denotes the observed part of the data; is a parameter used to control the tensor decomposition process; is the factor matrix in the tensor decomposition, which is used as a low-dimensional representation of the charging load data; is the error term. 8.The system of claim 7, wherein, The objective function of the electric vehicle load data repairing model based on structured tensor parallel factorization is solved by using a hierarchical least squares method to obtain complete electric vehicle charging load data. 9.The system of claim 8, wherein, The electric vehicle travel time prediction model based on xLSTM includes an input feature construction module, an mLSTM subnetwork, an sLSTM subnetwork, a feature fusion module, and a regression output module. The input feature construction module is configured to construct the reduced dimension data into an input sequence with a length of T in chronological order. The mLSTM subnetwork is configured to perform time series modeling on the input sequence and output a first hidden feature sequence. The sLSTM subnetwork is configured to perform time series modeling on the input sequence and output a second hidden feature sequence. The feature fusion module is configured to splice or weightedly fuse the first hidden feature sequence and the second hidden feature sequence to obtain fused features. The regression output module is configured to output electric vehicle travel time prediction data based on the fused features. 10.The system of claim 9, wherein, The two-layer optimization strategy combined with the sparrow search algorithm includes a first-layer optimization strategy and a second-layer optimization strategy. The first-layer optimization strategy includes the following: (1.1) first decision variable Q t : first decision variable Q t is time t when the power N t allocated to the vehicle, as follows: ； wherein Q t j is the time t is the power distribution value of the vehicle j at time j is 1- N t , N t is the number of vehicles t willing to participate in the optimization at time , is the total number of vehicles t at time is the number of vehicles t not participating in the optimization at time (1.2) the first objective function : the first objective function is time t of total user satisfaction, which is given by the following formula: ； wherein S j is a satisfaction state of the vehicle j , when F j is greater than 80%, a single user S j is set to 1, F j is a ratio of ; C j is a battery capacity of the vehicle j , is a charging demand of the vehicle j , is a target SOC of the vehicle j , is an initial SOC of the vehicle j ;​ (1.3) First model constraints include the following: (1.3.1) The charging amount constraint is: ； wherein, t is time t The remaining allocable charging power of the residential power distribution network after deducting the basic load and the non-scheduled electric vehicle load for participating in the dispatch of vehicles is as follows: ； wherein is the upper limit of the residential power distribution network; is the forecast value of the basic load of the current charging period, is the total charging load of the electric vehicles not participating in the dispatch, which is the load summary amount corresponding to the set of vehicles not participating in the dispatch, obtained by dividing the complete electric vehicle charging load data according to the user's intention; is is the average value of the number of vehicles participating in the user's predicted departure time, and the calculation formula is as follows: ； In the formula, is a predicted departure time of the vehicle j from the electric vehicle trip time prediction data; wherein The calculation formula of the following: The calculation formula of the following: ； wherein is time i EV load not participating in dispatch, is the start time of the current charging period, is the end time of the present charging cycle; ； wherein A not is a sequence of EV numbers that do not participate in scheduling, is a charging power of the vehicle j ; (1.3.2) The charging power constraint is: ； (1.3.3) The charging demand constraint is: ； wherein is the target SOC for the present charging cycle, is the initial SOC for the present charging cycle; (1.4) Solving the problem using the sparrow search algorithm. Q t As decision variables, with Given an optimization problem with an objective function that satisfies the first model constraint, we obtain: ,according to To allocate the optimal charging power to each electric vehicle; The second-layer optimization strategy includes the following: (2.1) Second decision variable as follows: ； wherein, w t is a matrix of decision variables, denotes a vehicle j at time i is charged, denotes a vehicle j at time i is not charged ; (2.2) The second objective function The second objective function is The average of the user satisfaction of the number of vehicles participating in the current charging period, which is calculated as follows: The average of the user satisfaction of the number of vehicles participating in the current charging period, which is calculated as follows: ； ； ； In the formula, representing the vehicle The amount of electricity / charge that can be obtained within the charging time window, which refers to the time to its predicted departure time , obtained from electric vehicle travel time prediction data; representing the vehicle The target demand for electricity; is the charging completion degree of the vehicle , wherein is the charging power or the amount of electricity corresponding to the charging power per unit time step; is the single-vehicle satisfaction, determined by the threshold rule: when take , otherwise take ; (2.3) Second model constraints include the following: (2.3.1) The energy acquisition constraint is: ； (2.3.2) The total amount of distributed power constraint is: ； In the formula, is the time i the basic load value at the time, is the time i the total charging load of the electric vehicles not participating in the dispatch at the time (2.4) The sparrow search algorithm is used to solve the optimization problem with as the decision variable, with as the objective function, and satisfying the second model constraint, to obtain . According to , the charging or non-charging decision of each electric vehicle at each time step is obtained.

Images