New energy vehicle charging management service platform and method based on intelligent internet of things

Abstract

The invention relates to the technical field of intelligent Internet of Things, in particular to a new energy vehicle charging management service platform and method based on intelligent Internet of Things. The method comprises the following steps: collecting user daily behavior data, vehicle external data, charging station external data, vehicle internal data, charging station monitoring data, payment SDK, charging power, road congestion rating and electricity price; inputting vehicle external data and charging station external data to a pre-constructed intelligent scheduling model to output probability distribution of charging station numbers; the probability distribution of the charging station numbers is input to an intelligent scheduling final selection unit to output final charging station numbers; inputting daily behavior data of the user and external data of the vehicle into the adaptive charging model to output a charging strategy; according to the invention, the condition that the adaptive relationship between the vehicle and the charging station is many-to-many butt joint is considered, and one-to-one selection of the accurate butt joint vehicle and the charging station is carried out through the machine learning model.

Description

Technical Field

[0001] The present invention relates to the field of intelligent Internet of Things technology, and more specifically, to a new energy vehicle charging and replacement management service platform and method based on intelligent Internet of Things. Background Art

[0002] Traditional charging and swapping management platforms have weak data collection capabilities and mainly rely on data from a single charging station. Scheduling decisions adopt simple rules with limited considerations. There is a lack of real-time perception and interaction with the infrastructure, and personalized services cannot be provided. The user experience is poor, the system structure is tightly coupled, and the scalability and dynamics are poor. The operation and management process is labor-intensive and inefficient.

[0003] The existing Chinese patent with authorization announcement number CN113313309B discloses a charging method and system for new energy vehicles, including: binding a user terminal and a vehicle, and obtaining information about the vehicle based on the user terminal; obtaining layout information of surrounding charging piles based on the vehicle information; obtaining the operating status of the charging piles based on the layout information and selecting a charging location; after determining the charging location, selecting a charging mode based on user needs and charging; selecting a suitable charging location according to the specific operating status of the vehicle, and selecting a suitable charging mode according to user needs, providing users with a complete charging process service, saving users' time and charging costs, and improving user satisfaction.

[0004] Although the above solution has made certain innovations in the traditional charging and swapping management method, it still belongs to the old framework of traditional charging and swapping, has no system, and has poor scalability and dynamics. It does not interact deeply with the Internet, does not consider the adaptation relationship between vehicles and charging stations for many-to-many docking, and does not consider the joint impact of users' daily behavior and the vehicle's own configuration on charging and swapping. Summary of the invention

[0005] In order to solve the above technical difficulties, an embodiment of the present invention provides a new energy vehicle charging and replacement management method based on intelligent Internet of Things, the method comprising:

[0006] Collect user daily behavior data, vehicle external data, charging station external data, vehicle internal data, charging station monitoring data, payment SDK, charging power, road congestion rating and electricity price;

[0007] Input vehicle external data and charging station external data into the pre-built intelligent scheduling model to output the probability distribution of the charging station number; input the probability distribution of the charging station number into the intelligent scheduling final selection unit to output the final charging station number;

[0008] Input user daily behavior data and vehicle external data into the adaptive charging model to output charging and swapping strategies;

[0009] Adjust charging power according to electricity prices;

[0010] Input vehicle internal data and charging station monitoring data into the pre-built fault diagnosis model to output the fault type;

[0011] Integrate payment SDK for payment;

[0012] Visualize the final charging station number, charging and swapping strategy, adjusted charging power, fault type and payment SDK.

[0013] Furthermore, the method for constructing the intelligent scheduling model includes:

[0014] The intelligent scheduling model uses GNN as the basic model structure. When inputting, the vehicle external data and the charging station external data are abstracted into feature graphs. The abstract settings of the feature graph include node features and edge features. The node features include the vehicle external data and the charging station external data. The edge features are the adaptation relationship between the vehicle and the charging station. The adaptation relationship includes power adaptation, location adaptation, power adaptation and demand adaptation. Power adaptation refers to whether the voltage standard supported by the charging port of the vehicle model is compatible with the output voltage of the charging station corresponding to the charging station number. Location adaptation refers to whether the distance between the vehicle location coordinates and the charging station location coordinates allows driving in for charging. Power adaptation refers to whether the difference between the maximum battery capacity of the vehicle battery and the current power of the vehicle battery exceeds the current total power of the charging station. Demand adaptation refers to whether the vehicle model matches the exclusive vehicle model requirements of the charging station. The basic model structure includes a graph embedding layer, a relationship network layer, a decoding layer and a prediction output layer.

[0015] Furthermore, the graph embedding layer adopts a graph convolution structure, which includes a convolution kernel and a convolution function. The window size of the convolution kernel is k hops, k is a preset value, and the convolution function uses the same function f to calculate and update the central node feature hv for the node features in each window. l+1 , the central node feature refers to the node v as the central node, and the original feature vector of v is hu l , the lth k-hop neighbor of v is u, and the set of all k neighbors is denoted as the set N(v), W l is the weight matrix, and the activation function is the nonlinear activation function σ;

[0016] The relationship network layer adopts a multi-head attention mechanism to update node features. The method for updating node features using the multi-head attention mechanism includes: setting nodes h1 and the next node h2 of h1, taking the node feature h of h1 and h2 as key and value, h includes the feature node of h1 and the node feature of h2, and when processing the edge connecting h1 and h2, mapping the edge feature e to the query vector query; value represents the node feature itself, key includes the starting node feature and the target node feature, and the starting node feature and the target node feature are mapped to key1 and key2 respectively, query, key1 and key2 are dot-product calculated, the result of the dot-product calculation is used as the attention score, the attention score is used as the weight and the dot-product calculation is performed with the value, and all edges whose starting node feature is key1 are summed up, and the result of the summation calculation is used as the edge attention contribution result of h1, and the edge attention contribution result of the previous node h1 of node h2 is spliced ​​with the node feature of h2, that is, a node feature update is completed; all nodes are updated according to the above-mentioned method for updating node features, and finally all updated nodes are output.

[0017] Furthermore, the decoding layer uses a layer of GRU network to predict the state information of the node at the next time step. The GRU network is composed of a gating mechanism, which includes an update gate zt, a candidate state ht and a forget gate rt; the update gate zt controls the update of the previous state information, the candidate state ht updates the candidate content, and the forget gate rt controls the forgetting of the previous state information; each input content of the time step t includes the node features of the current time step and the previous state information ht-1; the calculation of the previous state information of the next node predicted at the time step t includes:

[0018] Step 1: Calculate the update gate zt and the forget gate rt;

[0019] Step 2: Calculate the candidate state content ht;

[0020] Step 3: After obtaining ht, select and update ht to obtain the final ht;

[0021] The final ht is used as the previous state information of the next node, and steps 1 to 3 are repeated to sequentially predict the state information of the nodes in the entire time series; the time series consists of all time steps;

[0022] The prediction output layer includes a fully connected layer and an output layer; the prediction output layer uses the fully connected layer to take the state information ht as input, calculates the product of ht and the fully connected layer weight matrix WG plus the fully connected layer bias b1, uses softmax as the activation function of the output layer, and the output layer outputs the probability distribution of the charging station number.

[0023] Furthermore, the training method of the intelligent scheduling model includes:

[0024] Use the back-propagation algorithm to train the model parameters, use Adam as the optimizer, input training samples, which include vehicle external data, charging station external data and the probability distribution of the charging station number corresponding to the vehicle external data and charging station external data, input training samples, output the probability distribution of the charging station number, calculate the gradient of the loss function loos with respect to the parameters, use Adam to update each parameter according to gradient descent; repeat the training until loos converges and the training ends.

[0025] Furthermore, the adaptive charging model includes:

[0026] Convert each set of user daily behavior data and vehicle external data into the form of a first feature vector, where the elements of the first feature vector include the location of common charging stations, the length of stay at charging stations, the user's daily mileage, the vehicle model, the vehicle location coordinates, the current battery power of the vehicle, and the maximum cruising range of the vehicle;

[0027] The set of all first eigenvectors is used as the input of an adaptive charging model, the adaptive charging model takes the charging and swapping strategy predicted for each group of user daily behavior data and vehicle external data as output, takes the actual charging and swapping strategy corresponding to each group of user daily behavior data and vehicle external data as the prediction target, and takes minimizing the sum of the first prediction accuracies of all predicted charging and swapping strategies as the training target; the adaptive charging model is trained until the sum of the first prediction accuracies converges, and the training is stopped; the adaptive charging model is any one of a regression model, a decision tree model, or reinforcement learning.

[0028] Furthermore, the fault diagnosis model includes:

[0029] Convert each set of vehicle internal data and charging station monitoring data into the form of a second eigenvector, where the elements of the second eigenvector include the voltage of the vehicle battery, the current of the vehicle battery, the temperature of the vehicle battery, the voltage of the charging pile, the current of the charging pile, and the temperature of the charging pile;

[0030] The set of all second eigenvectors is used as the input of a fault diagnosis model, the fault diagnosis model takes the predicted fault type for each group of vehicle internal data and charging station monitoring data as output, takes the actual fault type corresponding to each group of vehicle internal data and charging station monitoring data as a prediction target, and takes minimizing the sum of the second prediction accuracies of all predicted fault types as a training target; the fault diagnosis model is trained until the sum of the second prediction accuracies reaches convergence, and the training is stopped; the fault diagnosis model is any one of a regression model and a decision tree model.

[0031] Furthermore, the method for adjusting the charging power according to the electricity price includes:

[0032] The charging process is divided into n time slices with a time interval of Δt. Each time slice is te, where te is an integer greater than 0. A heuristic algorithm is used to generate j groups of charging power plans. n and j are preset values. All charging power plans are calculated and sorted, and the charging power in the charging power plan with the lowest charging cost is output.

[0033] Furthermore, the intelligent scheduling final selection unit includes:

[0034] Obtain the road congestion degree through the navigation software, and then obtain the road congestion rating, sort the probability distribution of the charging station number output by the intelligent scheduling model in descending order, select the first vc charging station numbers as the candidate set, retain the ranking number of the charging station number in the candidate set and the probability distribution of the charging station number, set the road congestion rating digitally, set the road congestion rating to hi = (1, 2, 3), use the intrinsic road congestion rating and the ranking number of the charging station number as analysis data, input the analysis data into the trained intelligent scheduling final selection model, and output the final charging station number;

[0035] The specific training process of the final selection model of intelligent scheduling includes:

[0036] Collect the final charging station numbers corresponding to the c groups of analysis data in advance, and convert the analysis data and the corresponding final charging station numbers into a corresponding set of feature vectors;

[0037] Each set of feature vectors is used as the input of the intelligent scheduling final selection model. The intelligent scheduling final selection model uses a set of final charging station numbers corresponding to each set of analysis data as output, and uses the actual final charging station number corresponding to each set of analysis data as the prediction target. The actual final charging station number is the final charging station number corresponding to the analysis data collected in advance; the training goal is to minimize the sum of the prediction errors of all analysis data; wherein the calculation formula of the prediction error is Z K =(a K -μ K ) 2 , where Z K is the prediction error, K is the group number of the eigenvector corresponding to the analysis data, α K is the final charging station number corresponding to the Kth group of analysis data, μ K is the actual final charging station number corresponding to the Kth group of analysis data; the intelligent scheduling final selection model is trained until the sum of the prediction errors reaches convergence and the training is stopped; the intelligent scheduling final selection model is specifically a deep neural network model.

[0038] A new energy vehicle charging and swapping management service platform based on intelligent IoT, which includes:

[0039] Collection module: used to collect user daily behavior data, vehicle external data, charging station external data, vehicle internal data, charging station monitoring data, payment SDK, charging power, road congestion rating and electricity price;

[0040] Intelligent scheduling module: The intelligent scheduling module includes an intelligent scheduling model and an intelligent scheduling final selection unit, which is used to input vehicle external data and charging station external data into the pre-built intelligent scheduling model to output the probability distribution of the charging station number; the probability distribution of the charging station number is input into the intelligent scheduling final selection unit to output the final charging station number;

[0041] Adaptive charging module: used to input user daily behavior data and vehicle external data into the adaptive charging model to output charging and swapping strategies;

[0042] Charging power adjustment module: used to adjust the charging power according to the electricity price;

[0043] Intelligent fault diagnosis module: used to input vehicle internal data and charging station monitoring data into the pre-built fault diagnosis model to output the fault type;

[0044] Integrated payment management module: used to integrate payment SDK for payment;

[0045] Data visualization module: used to visualize the final charging station number, charging and swapping strategy, adjusted charging power, fault type and payment SDK.

[0046] The technical effects and advantages of the new energy vehicle charging and swapping management service platform and method based on intelligent Internet of Things of the present invention are as follows:

[0047] Comprehensively collect all kinds of vehicle-mounted and charging station data to broaden the system's perception capabilities; establish different models to handle specific tasks such as scheduling selection, charging and swapping strategies, fault diagnosis, etc., which are systematic in structure and improve efficiency and accuracy; taking into account the situation that the adaptation relationship between vehicles and charging stations is many-to-many docking, the adaptation relationship between vehicles and charging stations is abstracted into a graph to represent many-to-many docking when inputting the machine learning model, and then the machine learning model is used to accurately dock vehicles and charging stations One-to-one selection; realize data interaction through the Internet to complete intelligent real-time monitoring; fully consider real-time situations such as congestion, so that scheduling decisions are flexible and adaptable; comprehensively consider user daily behavior data and vehicle external data, and use machine learning algorithms to select automated charging and swapping strategies to make management services more humane and intelligent; integrate functional modules such as authentication payment, and the architecture is complete and convenient for practical application; the model training process is scientific and reasonable, and multiple methods such as regression and classification are used to improve learning effects; the system has a clear hierarchical design structure, high module decoupling and strong dynamics; it can collect feedback data in real time to continuously optimize each module to ensure efficient operation of the system; the visual interface is friendly and easy to operate, which enhances user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a connection diagram of the new energy vehicle charging and swapping management service platform based on intelligent Internet of Things in Example 1;

[0049] Figure 2 This is a connection diagram of the new energy vehicle charging and swapping management service platform based on intelligent Internet of Things in Example 2;

[0050] Figure 3 This is a flow chart of the new energy vehicle charging and battery replacement management method based on intelligent Internet of Things of the present invention;

[0051] Figure 4 is a schematic diagram of an electronic device of the present invention;

[0052] Figure 5 Schematic diagram of the storage medium of the present invention. DETAILED DESCRIPTION

[0053] 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.

[0054] Embodiment 1:

[0055] See also Figure 1 As shown, the new energy vehicle charging and swapping management service platform based on intelligent Internet of Things described in this embodiment includes:

[0056] Collection module: used to collect user daily behavior data, vehicle external data, charging station external data, vehicle internal data, charging station monitoring data, payment SDK and charging station data;

[0057] Intelligent scheduling module: used to input vehicle external data and charging station external data into the pre-built intelligent scheduling model to output the charging station number;

[0058] Adaptive charging module: used to input user daily behavior data and vehicle external data into the adaptive charging model to output charging and swapping strategies;

[0059] Intelligent fault diagnosis module: used to input vehicle internal data and charging station monitoring data into the pre-built fault diagnosis model to output the fault type;

[0060] Integrated payment management module: used to integrate payment SDK for payment;

[0061] Data visualization module: used to visualize charging station numbers, charging and swapping strategies, fault types, and payment SDKs.

[0062] Users' daily behavior data include the locations of commonly used charging stations, the length of stay at charging stations, and the users' daily mileage; vehicle external data includes vehicle model, vehicle location coordinates, vehicle battery current charge, and vehicle maximum range; vehicle internal data includes vehicle battery voltage, vehicle battery current, and vehicle battery temperature; charging station monitoring data includes charging pile voltage, charging pile current, and charging pile temperature; charging station external data includes charging station number, charging station name, charging station location coordinates, current total charging station charge, and number of available charging piles.

[0063] The locations of commonly used charging stations, the user's daily mileage, the charging station location coordinates and the vehicle location coordinates are obtained by GPS or other navigation software. The vehicle location coordinates and the charging station location coordinates affect the distance from the vehicle to the charging station. The closer the distance from the vehicle to the charging station, the greater the probability that the corresponding charging station number will be selected by the intelligent scheduling model as the charging station for vehicle charging. The location of commonly used charging stations is the charging and swapping interface area of ​​the available charging piles that are used most frequently by users, which is used to determine the user's charging and swapping preference. The charging and swapping interface area of ​​the available charging piles includes an automatic battery swapping interface and an automatic charging interface. If the location of commonly used charging stations is the automatic battery swapping interface of the charging pile, the automatic battery swapping interface is activated. If the location of commonly used charging stations is the automatic charging interface of the charging pile, the automatic charging interface is activated. The user's daily mileage determines the charging and swapping needs. The more mileage the user travels every day, the more The larger the distance the vehicle travels, the wider the area it passes through on a daily basis, and the more charging and battery swapping times are required; the number of charging times is related to the vehicle's maximum cruising range. When the vehicle's maximum cruising range is less than the user's daily mileage, one more charging or battery swapping will be required; the charging station residence time accumulates from the moment charging or battery swapping begins until the charging or battery swapping ends; the charging station residence time and the number of charging times affect the charging and battery swapping strategy selected when the charging station activates the automatic charging interface. The charging and battery swapping strategy includes the charging mode and the number of scheduled charging times. The charging modes include fast charging, slow charging and scheduled charging; when the charging station residence time exceeds the preset value A, slow charging is selected, and the number of scheduled charging times is 0; when the charging station residence time is lower than the preset value B, fast charging is selected, and the number of scheduled charging times is 0; when the charging station residence time is between A and B, scheduled charging is selected, and the number of scheduled charging times depends on the number of charging times.

[0064] The vehicle model and charging station number are obtained from the Internet. According to the vehicle model, the voltage standard supported by the vehicle charging port, the maximum battery capacity of the vehicle battery, and the parameters in the linear relationship between the vehicle model, the current power of the vehicle battery, and the maximum cruising range of the vehicle can be retrieved; according to the charging station number, the number of available charging piles at the charging station, the exclusive vehicle model requirements of the charging station, the current total power of the charging station, and the number of available charging piles that affect the probability of users entering the charging station to charge can be retrieved. Based on the current power prediction of the vehicle battery, the charging station with the largest number of available charging piles will be given priority.

[0065] The current charge of the vehicle battery, the maximum cruising range of the vehicle, the voltage, current and battery temperature of the vehicle battery are obtained by the vehicle battery sensor, wherein the current charge of the vehicle battery and the maximum cruising range of the vehicle are in a linear relationship, and the parameters in the linear relationship are known from the vehicle model. The current charge of the vehicle battery and the maximum cruising range of the vehicle affect the probability that the charging station number is selected by the intelligent scheduling model as the charging station for charging the vehicle. When the current charge of the vehicle battery and the maximum cruising range of the vehicle cannot guarantee that the vehicle reaches the charging station, the charging station number corresponding to the charging station will not be regarded as an option for the charging station for charging the vehicle by the intelligent scheduling model; the voltage, current and battery temperature of the vehicle battery are the basis for the fault diagnosis model to judge the fault type. For example, abnormal vehicle battery voltage can indicate that the fault source is inside the vehicle battery. For example, large differences in single cell voltages may involve rupture faults. Excessive current of the vehicle battery during charging and discharging may cause overheating or short circuit of the battery pole, thereby causing mechanical faults of the battery such as internal short circuit; long-term excessive current consumption indicates that the service life is approaching, which may cause the battery capacity of the vehicle battery to decay; excessive battery temperature is prone to cause chemical reaction out of control. If it is too high, it may cause fire safety faults.

[0066] The charging station monitoring data is obtained by the sensors inside the charging pile. The charging station monitoring data has an impact on the diagnosis of fault types. For example, if the charging pile voltage is abnormal, it may be due to power supply or connection problems, causing a charging power drop fault; if the charging pile current exceeds the limit, it may be due to poor contact of the charging port or reduced insulation performance, leading to circuit short circuit and other faults; if the charging pile temperature exceeds the limit, there may be a circuit board overheating fault caused by excessive charging power.

[0067] Fault types include abnormal vehicle battery voltage, vehicle battery current exceeding the limit, battery capacity attenuation of the vehicle battery, battery temperature exceeding the limit, abnormal charging pile voltage and charging pile temperature exceeding the limit.

[0068] The method for constructing the intelligent scheduling model includes:

[0069] The intelligent scheduling model uses GNN as the basic model structure. When inputting, the vehicle external data and the charging station external data are abstracted into feature graphs. The abstract settings of the feature graph include node features and edge features. The node features include the vehicle external data and the charging station external data. The edge features are the adaptation relationship between the vehicle and the charging station. The adaptation relationship includes power adaptation, location adaptation, power adaptation and demand adaptation. Power adaptation refers to whether the voltage standard supported by the charging port of the vehicle model is compatible with the output voltage of the charging station corresponding to the charging station number. Location adaptation refers to whether the distance between the vehicle location coordinates and the charging station location coordinates allows driving in for charging. Power adaptation refers to whether the difference between the maximum battery capacity of the vehicle battery and the current power of the vehicle battery exceeds the current total power of the charging station. Demand adaptation refers to whether the vehicle model matches the exclusive vehicle model requirements of the charging station. The basic model structure includes a graph embedding layer, a relationship network layer, a decoding layer and a prediction output layer.

[0070] The graph embedding layer adopts a graph convolution structure, which includes a convolution kernel and a convolution function. The window size of the convolution kernel is k hops, where k is a preset value. The convolution function uses the same function f to calculate and update the central node feature hv for the node features in each window. l+1 , the central node feature refers to the node v as the central node, and the original feature vector of v is hu l , the lth k-hop neighbor of v is u, and the set of all k-hop neighbors is denoted as the set N(v), W l is the weight matrix, the activation function is the nonlinear activation function σ, and the formula for calculating and updating the central node features includes:

[0071] f = hu l ×W l ;

[0072] hv l+1 =σ(∑{u∈N(v)}f);

[0073] The relationship network layer adopts a multi-head attention mechanism to update node features. The method of updating node features using the multi-head attention mechanism includes: setting node h1 and the next node h2 of h1, taking the node feature h of h1 and h2 as key and value, h includes the feature node of h1 and the node feature of h2, and when processing the edge connecting h1 and h2, mapping the edge feature e to the query vector query; value represents the node feature itself, which is used as the weight value of attention calculation, that is, taking the node feature itself as a reference value, which is used as the weight value of attention calculation, the key includes the starting node feature and the target node feature, and the starting node feature is The query, key1 and target node features are mapped to key1 and key2 respectively, and the dot product of query, key1 and key2 is calculated. The result of the dot product calculation is used as the attention score. The attention score is used as the weight and the dot product calculation is performed with the value. The sum of all edges whose starting node feature is key1 is calculated, and the result of the sum calculation is used as the edge attention contribution result of h1. The edge attention contribution result of the previous node h1 of node h2 is concatenated with the node feature of h2 to complete a node feature update. All nodes are updated according to the above method of updating node features, and finally all updated nodes are output.

[0074] The decoding layer uses a layer of GRU network to predict the state information of the node at the next time step. The GRU network consists of a gating mechanism, which includes an update gate zt, a candidate state ht, and a forget gate rt. The update gate zt controls the update of the previous state information, the candidate state ht is the candidate update content, and the forget gate rt controls the forgetting of the previous state information. Each input content of time step t includes the node features of the current time step and the previous state information ht-1. The calculation of the previous state information of the next node predicted at time step t includes:

[0075] Step 1: Calculate the update gate zt and the forget gate rt. The formula for calculating the update gate zt and the forget gate rt is:

[0076] zt=σ(Wz×xt+Uz×(ht-1));

[0077] rt = σ(Wr × xt + Ur × (ht-1));

[0078] In the formula, xt is the node feature input at time step t; Uz is the update gate matrix in the GRU network, Ur is the forget gate matrix in the GRU network; Wz is the update gate weight matrix, Wr is the forget gate weight matrix;

[0079] Step 2: Calculate the candidate state content ht. The formula for calculating the candidate state content ht is:

[0080] ht=tanh(W×xt+U(rtΘ(ht-1)));

[0081] Where U is the candidate state matrix in the GRU network; W is the candidate state weight matrix; Θ represents the Hadamard product, element-by-element multiplication; tanh is a nonlinear activation function;

[0082] Step 3: After obtaining ht, select and update ht to obtain the final ht. The formula for selecting and updating ht is:

[0083] ht=(1-zt)Θ(ht-1)+(ztΘht);

[0084] The final ht is used as the previous state information of the next node, and steps 1 to 3 are repeated to sequentially predict the state information of the nodes in the entire time series; the time series consists of all time steps;

[0085] It should be noted that the time step of the intelligent scheduling model is more of a unit used to express the order. Since the state information of the node is not single and requires a period of time to store, the time between starting to process the previous node and starting to process the next node can be regarded as a time step; the time steps are arranged in sequence to form a time series.

[0086] The prediction output layer includes a fully connected layer and an output layer; the prediction output layer uses the fully connected layer to take the state information ht as input, calculates the product of ht and the fully connected layer weight matrix WG plus the fully connected layer bias b1, uses softmax as the activation function of the output layer, and the output layer outputs the probability distribution of the charging station number, and selects the charging station number with the highest probability as the final output of the prediction output layer.

[0087] The training method of the intelligent scheduling model includes:

[0088] Use the back propagation algorithm to train the model parameters, the optimizer is adam, and the training samples are input. The training samples include the vehicle external data, the charging station external data and the probability distribution of the charging station number corresponding to the vehicle external data and the charging station external data. Input the training samples, output the probability distribution of the charging station number, and calculate the gradient of the loss function to the parameters. The formula of the loss function loos is:

[0089] loos = -∑yg × log2ygp;

[0090] Where yg is the probability distribution of the charging station number in the training sample; ygp is the probability distribution of the charging station number output by the model;

[0091] Use adam to update each parameter according to gradient descent; repeat the training until loos converges and the training ends.

[0092] It is particularly noteworthy that the charging station numbers corresponding to the vehicle external data and charging station external data in the training samples are the probability distributions of the charging station numbers manually labeled based on the vehicle external data and the charging station external data; the parameters are all parameters of the model.

[0093] The adaptive charging model includes:

[0094] Convert each set of user daily behavior data and vehicle external data into the form of a first feature vector, where the elements of the first feature vector include the location of common charging stations, the length of stay at charging stations, the user's daily mileage, the vehicle model, the vehicle location coordinates, the current battery power of the vehicle, and the maximum cruising range of the vehicle;

[0095] The set of all first feature vectors is used as the input of the adaptive charging model, the adaptive charging model uses the charging and swapping strategy predicted for each set of user daily behavior data and vehicle external data as the output, uses the actual charging and swapping strategy corresponding to each set of user daily behavior data and vehicle external data as the prediction target, and uses minimizing the sum of the first prediction accuracies of all predicted charging and swapping strategies as the training target;

[0096] Among them, the calculation formula of the first prediction accuracy is: M P =(N P -S P ) 2 , where M P is the first prediction accuracy, N P is the predicted charging and swapping strategy corresponding to the daily behavior data of the Pth group of users and the external data of the vehicle, S P is the actual charging and swapping strategy corresponding to the Pth group of user daily behavior data and vehicle external data, where P is the number of the user daily behavior data and the vehicle external data; the adaptive charging model is trained until the sum of the first prediction accuracies reaches convergence and the training is stopped; the adaptive charging model is any one of a regression model, a decision tree model or reinforcement learning.

[0097] The fault diagnosis model comprises:

[0098] Convert each set of vehicle internal data and charging station monitoring data into the form of a second eigenvector, where the elements of the second eigenvector include the voltage of the vehicle battery, the current of the vehicle battery, the temperature of the vehicle battery, the voltage of the charging pile, the current of the charging pile, and the temperature of the charging pile;

[0099] The set of all second feature vectors is used as input of a fault diagnosis model, wherein the fault diagnosis model uses the fault type predicted for each set of vehicle internal data and charging station monitoring data as output, uses the actual fault type corresponding to each set of vehicle internal data and charging station monitoring data as a prediction target, and uses minimizing the sum of the second prediction accuracies of all predicted fault types as a training target;

[0100] Among them, the calculation formula for the second prediction accuracy is: i =(β i -Ω i ) 2 , among which, ga i is the second prediction accuracy, β i is the predicted fault type corresponding to the i-th group of vehicle internal data and charging station monitoring data, Ω i is the actual fault type corresponding to the i-th group of vehicle internal data and charging station monitoring data; the fault diagnosis model is trained until the sum of the second prediction accuracy reaches convergence and the training is stopped; the fault diagnosis model is any one of a regression model and a decision tree model.

[0101] The integrated payment SDK is used to pay fees including:

[0102] Select a payment platform, enable the merchant function on the payment platform, obtain the merchant number, download the payment SDK corresponding to the merchant number, call the interface in the payment SDK, generate the interface of the order SDK, return the payment SDK, display the payment SDK on the platform, scan the payment SDK for successful payment, call the order query interface of the order SDK, verify that the order status is payment completed, and after the verification is completed, the order SDK will be changed to an open state.

[0103] Example 2

[0104] See also Figure 2 As shown, this embodiment further improves the design on the basis of embodiment 1. The difference is that when the intelligent scheduling module is running, embodiment 1 directly selects the charging station number with the highest probability as the final output of the prediction output layer. In the actual operation process, the road congestion situation will also have a strong impact on the selection of the charging station. When the road congestion situation is serious, even if the charging station number with the highest probability is output in the intelligent scheduling module, it is not applicable in this case, so it is necessary to adjust the final output of the intelligent scheduling module; at the same time, the electricity price will fluctuate at different time points, with peak electricity consumption during the day and low electricity consumption at night. The electricity price during the day is higher than that at night, so it is necessary to monitor the power grid price in real time and adjust the charging power during the charging stage to maximize the saving of users' charging costs; in summary, the new energy vehicle charging and swapping management service platform based on intelligent Internet of Things provided in this embodiment also includes a collection module, an intelligent scheduling module and a charging power adjustment module;

[0105] Collection module: used to collect user daily behavior data, vehicle external data, charging station external data, vehicle internal data, charging station monitoring data, payment SDK and charging station data, charging power, road congestion rating and electricity price;

[0106] The road congestion rating is based on the three-level classification of the road congestion level given by the navigation software. 1 is normal, and the navigation software does not give a congestion warning; 2 is congested, and the navigation software reminds that there are many vehicles on the road ahead; 3 is congested, and the navigation software warns that the road ahead is congested;

[0107] The electricity price is obtained in real time by the Internet Energy Exchange, and the charging power is obtained by the vehicle battery sensor. The charging power is adjusted with the fluctuation of the electricity price. When the electricity price is high, the charging power is relatively reduced, and when the electricity price is low, the charging power is relatively increased.

[0108] The intelligent scheduling module includes an intelligent scheduling model and an intelligent scheduling final selection unit;

[0109] Charging power adjustment module: used to adjust the charging power according to the electricity price;

[0110] The charging power adjustment module comprises:

[0111] The charging process is divided into n time slices, with a time interval of Δt. Each time slice is te, where te is an integer greater than 0. A heuristic algorithm is used to generate j groups of charging power plans, where j is a preset value. All charging power plans are calculated and sorted, and the charging power of the charging power plan with the lowest charging cost is output;

[0112] Methods for calculating the charging cost of the charging power scheme include:

[0113] The current charging power remains unchanged and is set to P0. The charging increment ΔR in each time slice is calculated based on the power P0 and the time interval Δt:

[0114] ΔR=P0×Δt / U;

[0115] Where U is the voltage of the vehicle battery;

[0116] Accumulated charge increment Ri:

[0117]

[0118] The electricity price corresponding to each time slice te is set to re te ; k is the number of the time slice; ΔR k is the charge increment in the kth time slice; R0 is the initial charge increment;

[0119] Calculate the estimated charging cost fy: fy = P0 × Δt × re te ×Ri.

[0120] It is particularly noted that there are various ways to solve the heuristic algorithm, such as randomness, greedy selection, heuristic evaluation function, camping search and hill climbing algorithm. In view of the random generation of solutions based on the Internet in this implementation scenario, the heuristic algorithm itself is not described in this embodiment. The heuristic algorithm is different from the precise algorithm and cannot guarantee the optimal solution, but it has high dynamic and computational efficiency. The reason why the precise algorithm is not used in this embodiment is that the embodiment 1 has already screened the data and filtered out a large number of irrelevant solutions. In view of the number of users, the number of charging stations and the server computing power, the heuristic algorithm is selected to save computing resources.

[0121] The intelligent scheduling final selection unit includes:

[0122] Obtain the road congestion degree through the navigation software, and then obtain the road congestion rating, sort the probability distribution of the charging station number output by the intelligent scheduling model in descending order, select the first vc charging station numbers as the candidate set, retain the ranking number of the charging station number in the candidate set and the probability distribution of the charging station number, set the road congestion rating digitally, set the road congestion rating to hi = (1, 2, 3), use the intrinsic road congestion rating and the ranking number of the charging station number as analysis data, input the analysis data into the trained intelligent scheduling final selection model, and output the final charging station number;

[0123] The specific training process of the final selection model of intelligent scheduling includes:

[0124] Collect the final charging station numbers corresponding to the c groups of analysis data in advance, and convert the analysis data and the corresponding final charging station numbers into a corresponding set of feature vectors;

[0125] Each set of feature vectors is used as the input of the intelligent scheduling final selection model. The intelligent scheduling final selection model uses a set of final charging station numbers corresponding to each set of analysis data as output, and uses the actual final charging station number corresponding to each set of analysis data as the prediction target. The actual final charging station number is the final charging station number corresponding to the analysis data collected in advance; the training goal is to minimize the sum of the prediction errors of all analysis data; wherein the calculation formula of the prediction error is Z K =(a K -μ K ) 2 , where Z K is the prediction error, K is the group number of the eigenvector corresponding to the analysis data, α K is the final charging station number corresponding to the Kth group of analysis data, μ K is the actual final charging station number corresponding to the Kth group of analysis data; the final selection model of intelligent scheduling is trained until the sum of the prediction errors reaches convergence and the training is stopped;

[0126] The final selection model of the above intelligent scheduling is specifically a deep neural network model.

[0127] Embodiment 3:

[0128] like Figure 3 As shown, the part not described in detail in this embodiment is described in Example 1 and Example 2, and a new energy vehicle charging and replacement management method based on intelligent Internet of Things is provided:

[0129] Collect user daily behavior data, vehicle external data, charging station external data, vehicle internal data, charging station monitoring data, payment SDK, charging power, road congestion rating and electricity price;

[0130] Input vehicle external data and charging station external data into the pre-built intelligent scheduling model to output the probability distribution of the charging station number; input the probability distribution of the charging station number into the intelligent scheduling final selection unit to output the final charging station number;

[0131] Input user daily behavior data and vehicle external data into the adaptive charging model to output charging and swapping strategies;

[0132] Adjust charging power according to electricity prices;

[0133] Input vehicle internal data and charging station monitoring data into the pre-built fault diagnosis model to output the fault type;

[0134] Integrate payment SDK for payment;

[0135] Visualize the final charging station number, charging and swapping strategy, adjusted charging power, fault type and payment SDK.

[0136] Example 4

[0137] See also Figure 4 As shown, according to another aspect of the present application, an electronic device 500 is also provided. The electronic device 500 may include one or more processors and one or more memories. Among them, the memory stores a computer readable code, and when the computer readable code is executed by one or more processors, the new energy vehicle charging and replacement management method based on intelligent Internet of Things as described above can be executed.

[0138] The system according to the embodiment of the present application can also be used by Figure 4 The electronic device architecture shown in FIG. Figure 4As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, ROM 503, RAM 504, a communication port 505 connected to a network, an input / output 506, a hard disk 507, etc. The storage device in the electronic device 500, such as ROM 503 or hard disk 507, may store the new energy vehicle charging and replacement management method based on intelligent Internet of Things provided in this application. Furthermore, the electronic device 500 may also include a user interface 508. Of course, Figure 4 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 4 One or more components of an electronic device are shown.

[0139] Example 5

[0140] See also Figure 5 As shown, it is a computer-readable storage medium according to an embodiment of the present application. Computer-readable instructions are stored on the computer-readable storage medium. When the computer-readable instructions are executed by the processor, the new energy vehicle charging and replacement management method based on intelligent Internet of Things according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0141] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0142] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A new energy vehicle charging and replacement management method based on intelligent Internet of Things, characterized in that: Methods include: Collect user daily behavior data, vehicle external data, charging station external data, vehicle internal data, charging station monitoring data, payment SDK, charging power, road congestion rating and electricity price; Input vehicle external data and charging station external data into the pre-built intelligent scheduling model to output the probability distribution of the charging station number; input the probability distribution of the charging station number into the intelligent scheduling final selection unit to output the final charging station number; Input user daily behavior data and vehicle external data into the adaptive charging model to output charging and swapping strategies; Adjust charging power according to electricity prices; Input vehicle internal data and charging station monitoring data into the pre-built fault diagnosis model to output the fault type; Integrate payment SDK for payment; Visualize the final charging station number, charging and swapping strategy, adjusted charging power, fault type and payment SDK.

2. The method for managing charging and replacing new energy vehicles based on intelligent Internet of Things according to claim 1 is characterized in that: The construction method of the intelligent scheduling model includes: The intelligent scheduling model uses GNN as the basic model structure. When inputting, the vehicle external data and the charging station external data are abstracted into feature graphs. The abstract settings of the feature graph include node features and edge features. The node features include the vehicle external data and the charging station external data. The edge features are the adaptation relationship between the vehicle and the charging station. The adaptation relationship includes power adaptation, location adaptation, power adaptation and demand adaptation. Power adaptation refers to whether the voltage standard supported by the charging port of the vehicle model is compatible with the output voltage of the charging station corresponding to the charging station number. Location adaptation refers to whether the distance between the vehicle location coordinates and the charging station location coordinates allows driving in for charging. Power adaptation refers to whether the difference between the maximum battery capacity of the vehicle battery and the current power of the vehicle battery exceeds the current total power of the charging station. Demand adaptation refers to whether the vehicle model matches the exclusive vehicle model requirements of the charging station. The basic model structure includes a graph embedding layer, a relationship network layer, a decoding layer and a prediction output layer.

3. The method for managing charging and replacing new energy vehicles based on intelligent Internet of Things according to claim 2 is characterized in that: The graph embedding layer adopts a graph convolution structure, which includes a convolution kernel and a convolution function. The window size of the convolution kernel is k hops, where k is a preset value. The convolution function uses the same function f to calculate and update the central node feature hv for the node features in each window. l+1 , the central node feature refers to the node v as the central node, and the original feature vector of v is hu l , the lth k-hop neighbor of v is u, and the set of all k-hop neighbors is denoted as the set N(v), W l is the weight matrix, and the activation function is the nonlinear activation function σ; The relationship network layer adopts a multi-head attention mechanism to update node features. The method for updating node features using the multi-head attention mechanism includes: setting nodes h1 and the next node h2 of h1, taking the node feature h of h1 and h2 as key and value, h includes the feature node of h1 and the node feature of h2, and when processing the edge connecting h1 and h2, mapping the edge feature e to the query vector query; value represents the node feature itself, key includes the starting node feature and the target node feature, and the starting node feature and the target node feature are mapped to key1 and key2 respectively, query, key1 and key2 are dot-product calculated, the result of the dot-product calculation is used as the attention score, the attention score is used as the weight and the dot-product calculation is performed with the value, and all edges whose starting node feature is key1 are summed up, and the result of the summation calculation is used as the edge attention contribution result of h1, and the edge attention contribution result of the previous node h1 of node h2 is spliced ​​with the node feature of h2, that is, a node feature update is completed; all nodes are updated according to the above-mentioned method for updating node features, and finally all updated nodes are output.

4. The method for managing charging and replacing new energy vehicles based on intelligent Internet of Things according to claim 2 is characterized in that: The decoding layer uses a layer of GRU network to predict the state information of the node at the next time step. The GRU network is composed of a gating mechanism, which includes an update gate zt, a candidate state ht and a forget gate rt. The update gate zt controls the update of the previous state information, the candidate state ht updates the candidate content, and the forget gate rt controls the forgetting of the previous state information. Each input content of the time step t includes the node features of the current time step and the previous state information ht-1. Predicting the previous state information of the next node at time step t includes: Step 1: Calculate the update gate zt and the forget gate rt; Step 2: Calculate the candidate state content ht; Step 3: After obtaining ht, select and update ht to obtain the final ht; The final ht is used as the previous state information of the next node, and steps 1 to 3 are repeated to sequentially predict the state information of the nodes in the entire time series; the time series consists of all time steps; The prediction output layer includes a fully connected layer and an output layer; the prediction output layer uses the fully connected layer to take the state information ht as input, calculates the product of ht and the fully connected layer weight matrix WG plus the fully connected layer bias b1, uses softmax as the activation function of the output layer, and the output layer outputs the probability distribution of the charging station number.

5. The method for managing charging and replacing new energy vehicles based on intelligent Internet of Things according to claim 2 is characterized in that: The training methods of the intelligent scheduling model include: Use the back-propagation algorithm to train the model parameters, use Adam as the optimizer, input training samples, which include vehicle external data, charging station external data and the probability distribution of the charging station number corresponding to the vehicle external data and charging station external data, input training samples, output the probability distribution of the charging station number, calculate the gradient of the loss function loos with respect to the parameters, use Adam to update each parameter according to gradient descent; repeat the training until loos converges and the training ends.

6. The method for managing charging and replacing new energy vehicles based on intelligent Internet of Things according to claim 1 is characterized in that: Adaptive charging models include: Convert each set of user daily behavior data and vehicle external data into the form of a first feature vector, where the elements of the first feature vector include the location of common charging stations, the length of stay at charging stations, the user's daily mileage, the vehicle model, the vehicle location coordinates, the current battery power of the vehicle, and the maximum cruising range of the vehicle; The set of all first eigenvectors is used as the input of an adaptive charging model, the adaptive charging model takes the charging and swapping strategy predicted for each group of user daily behavior data and vehicle external data as output, takes the actual charging and swapping strategy corresponding to each group of user daily behavior data and vehicle external data as the prediction target, and takes minimizing the sum of the first prediction accuracies of all predicted charging and swapping strategies as the training target; the adaptive charging model is trained until the sum of the first prediction accuracies converges, and the training is stopped; the adaptive charging model is any one of a regression model, a decision tree model, or reinforcement learning.

7. The method for managing charging and replacing new energy vehicles based on intelligent Internet of Things according to claim 1 is characterized in that: The fault diagnosis model includes: Convert each set of vehicle internal data and charging station monitoring data into the form of a second eigenvector, where the elements of the second eigenvector include the voltage of the vehicle battery, the current of the vehicle battery, the temperature of the vehicle battery, the voltage of the charging pile, the current of the charging pile, and the temperature of the charging pile; The set of all second eigenvectors is used as the input of a fault diagnosis model, the fault diagnosis model takes the predicted fault type for each group of vehicle internal data and charging station monitoring data as output, takes the actual fault type corresponding to each group of vehicle internal data and charging station monitoring data as a prediction target, and takes minimizing the sum of the second prediction accuracies of all predicted fault types as a training target; the fault diagnosis model is trained until the sum of the second prediction accuracies reaches convergence, and the training is stopped; the fault diagnosis model is any one of a regression model and a decision tree model.

8. The method for managing charging and replacing new energy vehicles based on intelligent Internet of Things according to claim 1 is characterized in that: Methods for adjusting charging power according to electricity prices include: The charging process is divided into n time slices with a time interval of Δt. Each time slice is te, where te is an integer greater than 0. A heuristic algorithm is used to generate j groups of charging power plans. n and j are preset values. All charging power plans are calculated and sorted, and the charging power in the charging power plan with the lowest charging cost is output.

9. The method for managing charging and replacing new energy vehicles based on intelligent Internet of Things according to claim 1 is characterized in that: The final selection unit of intelligent scheduling includes: Obtain the road congestion degree through the navigation software, and then obtain the road congestion rating, sort the probability distribution of the charging station number output by the intelligent scheduling model in descending order, select the first vc charging station numbers as the candidate set, retain the ranking number of the charging station number in the candidate set and the probability distribution of the charging station number, set the road congestion rating digitally, set the road congestion rating to hi = (1, 2, 3), use the intrinsic road congestion rating and the ranking number of the charging station number as analysis data, input the analysis data into the trained intelligent scheduling final selection model, and output the final charging station number; The specific training process of the final selection model of intelligent scheduling includes: Pre-collecting the final charging station numbers corresponding to the c groups of analysis data, and converting the analysis data and the corresponding final charging station numbers into a corresponding set of feature vectors; Each set of feature vectors is used as the input of the intelligent scheduling final selection model. The intelligent scheduling final selection model uses a set of final charging station numbers corresponding to each set of analysis data as output, and uses the actual final charging station number corresponding to each set of analysis data as the prediction target. The actual final charging station number is the final charging station number corresponding to the analysis data collected in advance; the training goal is to minimize the sum of the prediction errors of all analysis data; wherein the calculation formula of the prediction error is Z K =(a K -μ K ) 2 , where Z K is the prediction error, K is the group number of the eigenvector corresponding to the analysis data, α K is the final charging station number corresponding to the Kth group of analysis data, μ K is the actual final charging station number corresponding to the Kth group of analysis data; the intelligent scheduling final selection model is trained until the sum of the prediction errors reaches convergence and the training is stopped; the intelligent scheduling final selection model is specifically a deep neural network model.

10. A new energy vehicle charging and swapping management service platform based on intelligent IoT, characterized in that: Implementing the new energy vehicle charging and replacement management method based on intelligent Internet of Things as described in any one of claims 1 to 9, the service platform includes: Collection module: used to collect user daily behavior data, vehicle external data, charging station external data, vehicle internal data, charging station monitoring data, payment SDK, charging power, road congestion rating and electricity price; Intelligent scheduling module: The intelligent scheduling module includes an intelligent scheduling model and an intelligent scheduling final selection unit, which is used to input vehicle external data and charging station external data into the pre-built intelligent scheduling model to output the probability distribution of the charging station number; the probability distribution of the charging station number is input into the intelligent scheduling final selection unit to output the final charging station number; Adaptive charging module: used to input user daily behavior data and vehicle external data into the adaptive charging model to output charging and swapping strategies; Charging power adjustment module: used to adjust the charging power according to the electricity price; Intelligent fault diagnosis module: used to input vehicle internal data and charging station monitoring data into the pre-built fault diagnosis model to output the fault type; Integrated payment management module: used to integrate payment SDK for payment; Data visualization module: used to visualize the final charging station number, charging and swapping strategy, adjusted charging power, fault type and payment SDK.

Images