Charging station utilization rate prediction method, system and device based on deep learning, and medium

Abstract

The invention belongs to the technical field of charging station utilization rate prediction, and discloses a charging station utilization rate prediction method, system and device based on deep learning, and a medium, and the method comprises the steps: obtaining multi-dimensional comprehensive data of charging stations, carrying out the emotion analysis of user comment data in the multi-dimensional comprehensive data of each charging station, and carrying out the emotion analysis of the user comment data; obtaining a corresponding user emotion tendency index; constructing a charging station utilization rate prediction model, wherein the charging station utilization rate prediction model comprises a long short-term memory network layer, an attention module and a convolutional neural network layer which are connected in sequence; and training the charging station utilization rate prediction model based on the multi-dimensional comprehensive data and the user emotional tendency index, and executing a utilization rate prediction task of the to-be-predicted charging station based on the trained charging station utilization rate prediction model. According to the technical scheme, the utilization rate of the charging station can be accurately predicted in real time, and decision support is provided for planning and management of the charging station.

Description

Technical Field

[0001] The present invention belongs to the technical field of charging station utilization prediction, and in particular, relates to a charging station utilization prediction method, system, device and medium based on deep learning. Background Art

[0002] In order to deal with the problem of air pollution caused by the large amount of carbon dioxide produced by traditional fuel vehicles during driving, electric vehicles are seen as an effective solution. Countries around the world are actively promoting this transformation. Charging stations, as key facilities to support the continued growth of the industry, play an important role in promoting the use of electric vehicles. A sound and complete charging infrastructure system is the guarantee of the driving experience of electric vehicles. However, the reality is that the shortage of charging facilities has led to the inability to fully meet the charging demand, which has made many consumers hesitant about electric vehicles and hindered the promotion and development of electric vehicles. The utilization rate of power stations (i.e., the amount of electricity provided by a power station over a period of time) is a key driving factor in the economics of charging stations. Therefore, accurately predicting the utilization rate of charging stations is of great significance for optimizing the layout of charging station networks, improving charging efficiency, and promoting the popularization of electric vehicles. At present, research on the utilization rate of charging stations mainly focuses on using empirical analysis methods to conduct correlation analysis and multivariate linear regression between utilization rate and other factors, and propose management strategies based on this. However, the empirical analysis method using traditional analysis has the following defects when predicting the utilization rate of charging stations: 1. Limitations of traditional models: Traditional prediction methods mostly rely on linear regression or simple time series models, which are difficult to capture the complex dynamic changes of charging station utilization in different time and space; 2. Insufficient multi-dimensional data fusion: Existing research only considers the relationship between the location, price, charging time and charging utilization of charging stations, but ignores the impact of user experience. Summary of the invention

[0003] The purpose of the present invention is to provide a charging station utilization prediction method, system, device and medium based on deep learning to solve the problems existing in the above-mentioned prior art.

[0004] To achieve the above object, the present invention provides a charging station utilization prediction method based on deep learning, comprising:

[0005] Obtain multi-dimensional comprehensive data of the charging station, the multi-dimensional comprehensive data including real-time charging data corresponding to each charging station, supplier service information, user review data, business hours description, additional supplementary description, number of points of interest within a preset range, and utilization rate labels;

[0006] Performing sentiment analysis on the user review data corresponding to each of the charging stations to obtain a corresponding user sentiment tendency index;

[0007] Constructing a charging station utilization prediction model, wherein the charging station utilization prediction model includes a long short-term memory network layer, an attention module, and a convolutional neural network layer connected in sequence;

[0008] The charging station utilization prediction model is trained based on the multi-dimensional comprehensive data and the user sentiment tendency index, and the utilization prediction task of the charging station to be predicted is performed based on the trained charging station utilization prediction model.

[0009] Optionally, before performing sentiment analysis on the user comment data corresponding to each of the charging stations, it also includes performing missing value processing, outlier detection and standardization on the multi-dimensional comprehensive data of each charging station to obtain pre-processed multi-dimensional comprehensive data, and executing a sentiment analysis process and a training process of an initial charging station utilization prediction model based on the pre-processed multi-dimensional comprehensive data.

[0010] Optionally, the sentiment analysis of the user review data corresponding to each of the charging stations specifically includes:

[0011] Based on the Prompt project, the ChatGLM3-6B large language model is used to perform sentiment analysis on each user comment data of the charging station to obtain the sentiment value category corresponding to each user comment data, which includes negative, neutral and positive;

[0012] Based on the sentiment value category corresponding to each user's comment data, the user sentiment tendency index of the corresponding charging station as a whole is calculated.

[0013] Optionally, the training of the charging station utilization prediction model based on the multi-dimensional comprehensive data and the user sentiment tendency index specifically includes:

[0014] Construct an initial charging station utilization prediction model;

[0015] Determine model input data, wherein the model input data includes real-time charging data corresponding to each charging station, supplier service information, user review data, business hours description, additional supplementary description, the number of points of interest within a preset range, and user sentiment tendency indicators;

[0016] The model input data is input into the initial charging station utilization prediction model for classification prediction, and training is performed with the goal of minimizing the loss between the classified initial prediction result and the utilization label corresponding to the model input data to obtain a trained charging station utilization prediction model.

[0017] Optionally, inputting the model input data into the initial charging station utilization prediction model for classification prediction specifically includes:

[0018] Input the time series data in the model input data into the long short-term memory network layer to capture the long-term dependencies in the time series and obtain the time series features;

[0019] Input the time series features into the attention mechanism module to calculate the importance weight of the features and output the weighted time feature vector;

[0020] Input the spatial distribution data in the model input data into the convolutional neural network layer to extract spatial features and obtain spatial features;

[0021] The weighted temporal feature vector and spatial feature are fused to obtain fused features;

[0022] Classification prediction is performed based on the fusion features to obtain a utilization prediction result.

[0023] A charging station utilization prediction system based on deep learning, comprising:

[0024] A data collection module, used to obtain multi-dimensional comprehensive data of charging stations, wherein the multi-dimensional comprehensive data includes real-time charging data corresponding to each charging station, supplier service information, user review data, business hours description, additional supplementary description, number of points of interest within a preset range, and utilization rate labels;

[0025] A sentiment analysis module, used to perform sentiment analysis on the user comment data corresponding to each of the charging stations to obtain a corresponding user sentiment tendency index;

[0026] A model building application module is used to build a charging station utilization prediction model, which includes a long short-term memory network layer, an attention module and a convolutional neural network layer connected in sequence; the charging station utilization prediction model is trained based on the multi-dimensional comprehensive data and user emotional tendency indicators, and the utilization prediction task of the charging station to be predicted is performed based on the trained charging station utilization prediction model.

[0027] An electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute a charging station utilization prediction method based on deep learning.

[0028] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements a charging station utilization prediction method based on deep learning.

[0029] The technical effects of the present invention are:

[0030] In order to improve the accuracy of utilization prediction in the field of charging stations, the present invention utilizes the advantages of large models in natural language processing tasks, designs prompt engineering based on the general large model, and fine-tunes the large model for sentiment analysis to achieve accurate and efficient sentiment scoring of comment texts. On this basis, combined with multi-dimensional information such as spatiotemporal data of charging stations and operator services, a deep learning model based on LSTM+CNN and combined with attention mechanism is constructed, thereby achieving real-time and accurate prediction of charging station utilization, providing decision support for charging station planning and management. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0033] Figure 1 A distribution map of charging stations collected in an embodiment of the present invention;

[0034] Figure 2 is a model structure diagram of a prediction model in an embodiment of the present invention;

[0035] Figure 3 It is the MSE (mean square error) convergence diagram in the embodiment of the present invention;

[0036] Figure 4 This is a prediction flow chart in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] Various exemplary embodiments of the present invention will now be described in detail. This detailed description should not be considered as limiting the present invention, but should be understood as a more detailed description of certain aspects, features, and embodiments of the present invention.

[0038] It should be understood that the terms described in the present invention are only for describing special embodiments and are not intended to limit the present invention. In addition, for the numerical range in the present invention, it should be understood that each intermediate value between the upper and lower limits of the scope is also specifically disclosed. Each smaller range between the intermediate value in any stated value or stated range and any other stated value or intermediate value in the described range is also included in the present invention. The upper and lower limits of these smaller ranges can be independently included or excluded in the scope.

[0039] It will be apparent to those skilled in the art that various modifications and variations may be made to the specific embodiments of the present invention description without departing from the scope or spirit of the present invention. Other embodiments derived from the present invention description will be apparent to those skilled in the art. The present application description and examples are exemplary only.

[0040] The words “include,” “including,” “have,” “contain,” etc. used in this article are open-ended terms, meaning including but not limited to.

[0041] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0042] like Figure 1 - Figure 4 As shown, in this embodiment, a method for predicting charging station utilization based on deep learning is provided, including: obtaining multi-dimensional comprehensive data of charging stations, the multi-dimensional comprehensive data including real-time charging data, supplier service information, user comment data, business hours description, additional supplementary description, the number of points of interest within a preset range, and utilization labels corresponding to each charging station; performing sentiment analysis on the user comment data corresponding to each of the charging stations to obtain corresponding user sentiment tendency indicators; constructing a charging station utilization prediction model, the charging station utilization prediction model including a long short-term memory network layer, an attention module, and a convolutional neural network layer connected in sequence; training the charging station utilization prediction model based on the multi-dimensional comprehensive data and the user sentiment tendency indicators, and executing the utilization prediction task of the charging station to be predicted based on the trained charging station utilization prediction model.

[0043] In order to improve the accuracy of the utilization prediction in the field of charging stations, this embodiment uses the advantages of large models in natural language processing tasks, designs prompt engineering based on the general large model, and fine-tunes the sentiment analysis of the large model to achieve accurate and efficient sentiment scoring of the comment text. On this basis, combined with the spatiotemporal data of charging stations and multi-dimensional information such as operator services, a deep learning model based on LSTM+CNN and combined with attention mechanism is constructed to achieve real-time prediction of charging station utilization, providing decision support for charging station planning and management.

[0044] This embodiment is implemented by the following steps:

[0045] 1. Data collection: Use Python crawler method to obtain data. Randomly select charging stations from public websites to obtain their location information, real-time charging data over a period of time, service information provided by suppliers, user comments, and the number of POIs (points of interest, used to indicate the prosperity of a certain area) within a 1km range.

[0046] 2. Data preprocessing:

[0047] (1) Cleaning and standardization: missing value processing, outlier detection and standardization are performed on the collected data to ensure data quality.

[0048] (2) Sentiment analysis: This embodiment takes advantage of the large model in natural language processing tasks, designs a prompt project based on the general large model, and fine-tunes the large model for sentiment analysis to achieve accurate and efficient sentiment scoring of the comment text. The specific process includes:

[0049] The collected user comment data is converted into JSON format, and the prompt project is carried out based on the general large model, and fine-tuned in the field of sentiment analysis. The fine-tuned large language model is used to perform sentiment analysis on user comments, and the sentiment is divided into three categories, namely -1 (negative), 0 (neutral), and 1 (positive). The average sentiment value of user comments for each charging station is calculated as the sentiment tendency indicator of the charging station.

[0050] (3) Utilization rate calculation: Calculate the utilization rate of each charging pile at each time. The formula is: Utilization rate = number of charging ports in use / total number of charging ports at the station. Take the average utilization rate of each charging pile to get the utilization rate of the charging station.

[0051] 3. Feature construction: The types and quantity of services provided by service providers at each charging station, the number of user reviews, the comment sentiment score, business hours, and the number of surrounding POIs are taken as independent variables; the utilization rate of each charging station at different times is taken as the dependent variable to construct a complete data set.

[0052] This embodiment not only considers structured data when constructing features: location information, real-time charging data, the number of POIs within 1km of the charging station, business hours data, etc., but also includes unstructured data such as: service types and quantities, user comment information, site supplementary information, etc. By combining multi-dimensional information, it is possible to comprehensively analyze the key factors affecting the performance of charging stations.

[0053] 4. Model training and testing:

[0054] (1) Data division: 80% of the collected real-time charging data is used as a training set and 20% as a test set.

[0055] (2) Model construction: A hybrid network model combining LSTM+CNN and attention mechanism is established. The LSTM layer processes time series data, and the attention mechanism is introduced after the LSTM layer to identify and focus on factors that have a greater impact on the prediction results. The CNN part is used to extract features from spatially distributed data.

[0056] This embodiment combines the long short-term memory network (LSTM) and the convolutional neural network (CNN), and introduces the attention mechanism to construct a hybrid network model, which can effectively process spatiotemporal data and make the model focus on the factors that have the greatest impact on the prediction results, making the prediction accurate and reliable.

[0057] 5. Performance evaluation: In order to evaluate the prediction performance of the model, mean square error (MSE) and R 2 MSE is measured by two indicators: R and MSE. MSE is a quantitative measure of the model's prediction error by calculating the average square of the error between the model's predicted value and the actual value. The specific calculation process is shown in Formula 1. 2 The calculation of (coefficient of determination) is shown in Formula 2, where SSR represents the regression sum of squares and SST represents the total sum of squares, which reflects the degree of correlation between the model predicted value and the actual value.

[0058]

[0059] R 2 =SSR / SST (2)

[0060] The specific implementation process of this embodiment will be described in detail through the following steps to realize the charging station utilization prediction method based on LSTM+CNN combined with attention mechanism.

[0061] (1) Data collection: This example randomly selects 948 charging stations in the United States as research samples. Figure 1 As shown, each green dot represents a charging station whose data is collected, and the detailed information of these 948 charging stations is collected through the PlugShare website (PlugShare is one of the world's largest electric vehicle owner communities, providing users with a comprehensive platform for finding, evaluating and sharing charging station experiences), covering the location data of the stations, nearly 2.3 million real-time charging data updated every half hour within two months, the types and quantity of services provided by charging providers, user review information, business hours, additional supplementary instructions, and the number of POIs within 1km around the charging station.

[0062] (2) Feature engineering and data preprocessing: The collected data is cleaned, standardized and encoded to facilitate model training and prediction. Specifically, the ratio of the number of charging ports in use to the total number of charging ports at the time of collection of each charging pile is calculated as the utilization rate of the charging pile at that time, and the utilization rate of the charging pile is further averaged to obtain the overall utilization rate of the charging station; in terms of comment sentiment analysis, the collected user comment data is converted into Json format data, and the Prompt project is carried out with the help of the ChatGLM3-6B large language model (a series of Chinese-English bilingual dialogue models jointly developed by Tsinghua University and Zhipu AI) to perform sentiment analysis on user comments. The sentiment value is divided into three levels: -1 represents negative, 0 represents neutral, and 1 represents positive. The idea of ​​Prompt construction is as follows: 1. Clear task instructions: instruct the model to generate the sentiment value of the comment text, and the sentiment value is specified to be selected between -1, 0, and 1; 2. Determine the input content: provide user comment data in English text as context; 3. Classification output format: require the format of the output classification result. Below is a Prompt example I built: I need to classify the sentiment of a paragraph, where positive sentiment is represented by 1, neutral sentiment is represented by 0, and negative sentiment is represented by -1. Analyze 'Comment information', which of the three categories does this paragraph belong to, give the numerical value corresponding to the classification result, and do not need to explain the reason. The final output format is: "Sentiment value is:". In order to verify the accuracy of the model analysis, a part of the data is extracted to compare and analyze the sentiment value generated by the large model with the sentiment value annotated manually. The results show that the sentiment value predicted by the large language model is consistent with the sentiment value annotated manually by 94%, which proves the accuracy of the large model analysis. Based on the sentiment analysis of the large model, the average sentiment value of all user comments on each charging station is further calculated as the overall user sentiment tendency indicator of the charging station. The types and quantity of services provided by the charging supplier, the number of user comments, the sentiment score of the comments, the business hours description, the additional supplementary description, and the number of POIs within 1km of the charging station are taken as independent variables, and the utilization rate of the charging station at different times is taken as the dependent variable.

[0063] (3) Divide the data into training set and test set: 80% of the collected real-time charging data is used as the training set and 20% as the test set. The model is trained using the training set data, and the model is verified and evaluated using the test set data.

[0064] (4) Model construction: A combination of long short-term memory network (LSTM) and convolutional neural network (CNN) is adopted, and the attention mechanism is introduced to form a hybrid network model, such as Figure 2As shown in the figure. The LSTM layer is used to process and predict time series data, and the hidden layer dimension is set to 64. The CNN part is responsible for extracting features from spatially distributed data, including convolutional layers, pooling layers, and fully connected layers. The convolutional layer has 64 input channels, 32 output channels, and the convolution kernel size is set to 2. The pooling kernel size in the pooling layer is 2, and the stride is 1. The input dimension of the fully connected layer is 32, and the output dimension is 1. The number of training rounds is 150.

[0065] (5) Training and testing: 80% of the collected real-time charging data is used as a training set and 20% as a test set. The model is trained using the training set data, and the model is verified and evaluated using the test set data.

[0066] (6) Performance evaluation: by calculating the MSE and R of the prediction results 2 values ​​to evaluate the predictive performance of the model.

[0067] According to the performance of the hybrid network model in the experiment, this embodiment obtained an MSE of 0.0430 and R 2 The result is a value of 0.4486. Figure 3 The figure shows the change of MSE with the iteration of the model. The MSE of 0.0430 indicates that the mean square difference between the actual value and the predicted value is relatively small, which means that the predicted result of the model is close to the actual value. 2 The value of 0.4486 indicates that the model explains about 44.86% of the data variability, and some variability is not captured by the model. There are also some factors that affect the utilization of charging stations that are not captured, such as weather changes, maintenance and operation of charging stations, policies and subsidies, etc. Overall, these results provide valuable references for charging station operators and point out the direction of future research.

[0068] A charging station utilization prediction system based on deep learning, comprising:

[0069] A data collection module, used to obtain multi-dimensional comprehensive data of charging stations, wherein the multi-dimensional comprehensive data includes real-time charging data corresponding to each charging station, supplier service information, user review data, business hours description, additional supplementary description, number of points of interest within a preset range, and utilization rate labels;

[0070] A sentiment analysis module, used to perform sentiment analysis on the user comment data corresponding to each of the charging stations to obtain a corresponding user sentiment tendency index;

[0071] A model building application module is used to build a charging station utilization prediction model, which includes a long short-term memory network layer, an attention module and a convolutional neural network layer connected in sequence; the charging station utilization prediction model is trained based on the multi-dimensional comprehensive data and user emotional tendency indicators, and the utilization prediction task of the charging station to be predicted is performed based on the trained charging station utilization prediction model.

[0072] An electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute a charging station utilization prediction method based on deep learning.

[0073] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements a charging station utilization prediction method based on deep learning.

[0074] The above is only a preferred specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A charging station utilization prediction method based on deep learning, characterized in that: include: Obtain multi-dimensional comprehensive data of the charging station, the multi-dimensional comprehensive data including real-time charging data corresponding to each charging station, supplier service information, user review data, business hours description, additional supplementary description, number of points of interest within a preset range, and utilization rate labels; Performing sentiment analysis on the user review data corresponding to each of the charging stations to obtain a corresponding user sentiment tendency index; Constructing a charging station utilization prediction model, wherein the charging station utilization prediction model includes a long short-term memory network layer, an attention module, and a convolutional neural network layer connected in sequence; The charging station utilization prediction model is trained based on the multi-dimensional comprehensive data and the user sentiment tendency index, and the utilization prediction task of the charging station to be predicted is performed based on the trained charging station utilization prediction model.

2. The method for predicting charging station utilization rate based on deep learning according to claim 1, characterized in that: Before performing sentiment analysis on the user review data corresponding to each of the charging stations, it also includes performing missing value processing, outlier detection and standardization on the multi-dimensional comprehensive data of each charging station to obtain pre-processed multi-dimensional comprehensive data, and executing a sentiment analysis process and a training process of an initial charging station utilization prediction model based on the pre-processed multi-dimensional comprehensive data.

3. The method for predicting charging station utilization rate based on deep learning according to claim 1, characterized in that: The sentiment analysis of the user review data corresponding to each charging station specifically includes: Based on the Prompt project, the ChatGLM3-6B large language model is used to perform sentiment analysis on each user comment data of the charging station to obtain the sentiment value category corresponding to each user comment data, which includes negative, neutral and positive; Based on the sentiment value category corresponding to each user's comment data, the user sentiment tendency index of the corresponding charging station as a whole is calculated.

4. The method for predicting charging station utilization rate based on deep learning according to claim 1, characterized in that: The training of the charging station utilization prediction model based on the multi-dimensional comprehensive data and the user sentiment tendency index specifically includes: Construct an initial charging station utilization prediction model; Determine model input data, wherein the model input data includes real-time charging data corresponding to each charging station, supplier service information, user review data, business hours description, additional supplementary description, the number of points of interest within a preset range, and user sentiment tendency indicators; The model input data is input into the initial charging station utilization prediction model for classification prediction, and training is performed with the goal of minimizing the loss between the classified initial prediction result and the utilization label corresponding to the model input data to obtain a trained charging station utilization prediction model.

5. A method for predicting charging station utilization rate based on deep learning according to claim 4, characterized in that: The step of inputting the model input data into the initial charging station utilization rate prediction model for classification prediction specifically includes: Input the time series data in the model input data into the long short-term memory network layer to capture the long-term dependencies in the time series and obtain the time series features; Input the time series features into the attention mechanism module to calculate the importance weight of the features and output the weighted time feature vector; Input the spatial distribution data in the model input data into the convolutional neural network layer to extract spatial features and obtain spatial features; The weighted temporal feature vector and spatial feature are fused to obtain fused features; Classification prediction is performed based on the fusion features to obtain a utilization prediction result.

6. A charging station utilization prediction system based on deep learning, characterized in that: include: A data collection module, used to obtain multi-dimensional comprehensive data of charging stations, wherein the multi-dimensional comprehensive data includes real-time charging data corresponding to each charging station, supplier service information, user review data, business hours description, additional supplementary description, number of points of interest within a preset range, and utilization rate labels; A sentiment analysis module, used to perform sentiment analysis on the user comment data corresponding to each of the charging stations to obtain a corresponding user sentiment tendency index; A model building application module is used to build a charging station utilization prediction model, which includes a long short-term memory network layer, an attention module and a convolutional neural network layer connected in sequence; the charging station utilization prediction model is trained based on the multi-dimensional comprehensive data and user emotional tendency indicators, and the utilization prediction task of the charging station to be predicted is performed based on the trained charging station utilization prediction model.

7. An electronic device, characterized in that: It includes a memory and a processor, the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute a charging station utilization prediction method based on deep learning according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that: It stores a computer program, which, when executed by a processor, implements a charging station utilization prediction method based on deep learning as described in any one of claims 1 to 5.

Images