Intelligent parking management system and method for matching charging and parking supply and demand

Abstract

The present application relates to the field of intelligent management, and specifically discloses an intelligent parking management system and method for charging and parking supply and demand matching, which obtains historical charging demand data and real-time traffic status, wherein the real-time traffic status is a congestion level label, and uses data analysis and processing based on deep learning neural networks to perform semantic reasoning analysis of the historical charging demand data and data processing of the real-time traffic status, so as to automatically obtain the recommended value of charging and parking space reservation ratio according to the cross-modal association characteristics of the charging demand and the real-time traffic. This method can more comprehensively understand the charging behavior and traffic conditions, meet the demand peaks in different time periods, adapt to changes in traffic conditions, thereby improving the real-time and accuracy of the system, making parking management more intelligent and efficient, thereby improving the user's service experience.

Description

Technical Field

[0001] The present application relates to the field of intelligent management, and more specifically, to an intelligent parking management system and method for matching charging and parking supply and demand. Background Art

[0002] With the development of economic level, information technology and automobile industry, intelligent networked vehicles have become the development trend of future vehicles, and the proportion of electric vehicles has also increased. However, the low parking space utilization rate and less than 30% of charging parking spaces have caused difficulties in urban charging and parking.

[0003] Parking management usually refers to the process of organizing and controlling vehicle parking. However, the existing charging and parking spaces have no reserved management strategy, facing the problems of supply and demand imbalance and excessive competition, that is, some areas may have excess demand, while other areas may have insufficient supply, resulting in uneven resource utilization, which will cause additional waiting time, affect the operation of parking lot resources, and further cause traffic congestion and economic losses. In addition, the existing parking management system cannot allocate resources and optimize management based on historical data and trends, resulting in inefficient resource utilization and low parking management efficiency.

[0004] Therefore, an optimized intelligent parking management system for matching charging and parking supply and demand is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an intelligent parking management system and method for charging and parking supply and demand matching, which obtains historical charging demand data and real-time traffic status, wherein the real-time traffic status is a congestion level label, and uses data analysis and processing based on deep learning neural networks to perform semantic reasoning analysis of the historical charging demand data and data processing of the real-time traffic status, so as to automatically obtain the recommended value of charging and parking space reservation ratio according to the cross-modal association characteristics of the charging demand and the real-time traffic. This method can more comprehensively understand the charging behavior and traffic conditions to meet the demand peaks in different time periods and adapt to changes in traffic conditions, thereby improving the real-time and accuracy of the system, making parking management more intelligent and efficient, thereby improving the user's service experience.

[0006] According to one aspect of the present application, there is provided an intelligent parking management system for charging and parking supply and demand matching, which includes:

[0007] A charging demand and traffic status data acquisition module, used to acquire historical charging demand data and real-time traffic status, wherein the real-time traffic status is a congestion level label;

[0008] A charging demand data embedding coding module, used for performing low-dimensional embedding coding on each charging demand data in the historical charging demand data to obtain a time series of low-dimensional embedding coding feature vectors of the charging demand data;

[0009] A charging demand context encoding module, used for performing charging demand context encoding based on the full time domain on the time series of the low-dimensional embedding encoding feature vector of the charging demand data to obtain a time series of contextual charging demand data semantic encoding feature vectors;

[0010] A charging demand reasoning module, used for passing the time series of the semantic encoding feature vector of the contextual charging demand data through the charging demand reasoning module based on prior information to obtain a charging demand reasoning semantic feature vector;

[0011] A traffic status one-hot encoding module, used for one-hot encoding the real-time traffic status to obtain a one-hot encoding feature vector of a congestion level label;

[0012] A feature association coding module, used for cross-modal association coding of the charging demand reasoning semantic feature vector and the congestion level label one-hot encoding feature vector to obtain a charging demand reasoning semantic feature matrix under real-time traffic conditions as a charging demand reasoning semantic feature under real-time traffic conditions;

[0013] The reservation ratio recommendation module is used to obtain the recommended value of the charging and parking space reservation ratio based on the semantic characteristics of the charging demand in the real-time traffic state.

[0014] In the above-mentioned intelligent parking management system for charging and parking supply and demand matching, the charging demand context encoding module is used to: embed the charging demand data into a low-dimensional encoding feature vector of the time series through a full-time domain charging demand context encoder based on the Transformer layer to obtain the context charging demand data semantic encoding feature vector of the time series.

[0015] In the above-mentioned intelligent parking management system for charging and parking supply and demand matching, the charging demand reasoning module includes: a position feature determination unit, which is used to determine the context charging demand data semantic encoding feature vector of the last position in the time series of the context charging demand data semantic encoding feature vector; an average distribution information difference calculation unit, which is used to calculate the position difference between the mean feature vector of multiple context charging demand data semantic encoding feature vectors in the time series of the context charging demand data semantic encoding feature vector and the context charging demand data semantic encoding feature vector of the last position to obtain the context charging demand data semantic average distribution information difference feature vector; a random distribution information difference calculation unit, which is used to calculate the context charging demand data semantic encoding feature vector. The positional difference between the contextual charging demand data semantic encoding feature vector at a random position in the time series of the encoding feature vector and the contextual charging demand data semantic encoding feature vector at the last position is used to obtain the contextual charging demand data semantic random distribution information difference feature vector; an inference feature vector construction unit is used to construct a first charging demand inference feature vector and a second charging demand inference feature vector based on the contextual charging demand data semantic average distribution information difference feature vector and the contextual charging demand data semantic random distribution information difference feature vector; and a feature positional addition unit is used to calculate the positional addition between the first charging demand inference feature vector and the second charging demand inference feature vector to obtain the charging demand inference semantic feature vector.

[0016] In the above-mentioned intelligent parking management system for charging and parking supply and demand matching, the inference feature vector construction unit includes: a first charging demand inference subunit, used to calculate the product of the natural exponential feature vector of the semantic average distribution information difference feature vector of the context charging demand data as a power and the first Gaussian distribution hyperparameter and the first weight hyperparameter to obtain the first charging demand inference feature vector; and a second charging demand inference subunit, used to calculate the product of the natural exponential feature vector of the semantic random distribution information difference feature vector of the context charging demand data as a power and the second Gaussian distribution hyperparameter and the second weight hyperparameter to obtain the second charging demand inference feature vector.

[0017] In the above-mentioned intelligent parking management system for charging and parking supply and demand matching, the first Gaussian distribution hyperparameter and the second Gaussian distribution hyperparameter are generated by a Gaussian distribution random number function with a mean of 0 and a variance of 1.

[0018] In the above-mentioned intelligent parking management system for charging and parking supply and demand matching, the feature association encoding module is used to: pass the charging demand reasoning semantic feature vector and the congestion level label one-hot encoded feature vector through a cross-modal association encoder based on the CLIP model to obtain the charging demand reasoning semantic feature matrix under the real-time traffic state.

[0019] In the above-mentioned intelligent parking management system for charging and parking supply and demand matching, the reservation ratio recommendation module is used to: pass the charging demand inference semantic feature matrix under the real-time traffic state through the charging and parking space reservation ratio recommender based on the decoder to obtain the charging and parking space reservation ratio recommendation value.

[0020] In the above-mentioned intelligent parking management system for charging and parking supply and demand matching, it also includes a training module for training the full-time domain charging demand context encoder based on the Transformer layer, the charging demand reasoning module based on prior information, the cross-modal association encoder based on the CLIP model, and the charging and parking space reservation ratio recommender based on the decoder.

[0021] In the above-mentioned intelligent parking management system for charging and parking supply and demand matching, the training module includes: a training data acquisition unit, used to acquire training data, the training data including historical training charging demand data and training real-time traffic status, and the real recommended value of the charging and parking space reservation ratio; a training charging demand data encoding unit, used to perform low-dimensional embedding encoding on each training charging demand data in the historical training charging demand data to obtain a time series of low-dimensional embedded encoding feature vectors of training charging demand data; a training charging demand context encoding unit, used to pass the time series of the training charging demand data low-dimensional embedded encoding feature vector through the full-time domain charging demand context encoder based on the Transformer layer to obtain a time series of training context charging demand data semantic encoding feature vectors; a training charging demand inference unit, used to pass the time series of the training context charging demand data semantic encoding feature vector through the charging demand inference module based on prior information to obtain a training charging demand inference semantic feature vector; a training one-hot encoding unit, used to perform one-hot encoding on the training real-time traffic status to obtain a training congestion level label one-hot encoding feature vector; A training cross-modal association encoding unit is used to pass the training charging demand reasoning semantic feature vector and the training congestion level label unique hot encoding feature vector through the cross-modal association encoder based on the CLIP model to obtain a training real-time traffic state charging demand reasoning semantic feature matrix; a feature optimization unit is used to perform feature optimization on the training real-time traffic state charging demand reasoning semantic feature matrix to obtain an optimized training real-time traffic state charging demand reasoning semantic feature matrix; a decoding loss unit is used to pass the optimized training real-time traffic state charging demand reasoning semantic feature matrix through the decoder-based charging and parking space reservation ratio recommender to obtain a decoding loss function value; and a training unit is used to train the Transformer layer-based full-time domain charging demand context encoder, the prior information-based charging demand reasoning module, the CLIP model-based cross-modal association encoder, and the decoder-based charging and parking space reservation ratio recommender based on the decoding loss function value and through gradient descent back propagation, wherein, in each iteration of the model, the training real-time traffic state charging demand reasoning semantic feature matrix is ​​optimized.

[0022] According to another aspect of the present application, a smart parking management method for charging and parking supply and demand matching is provided, which includes:

[0023] Acquire historical charging demand data and real-time traffic status, wherein the real-time traffic status is a congestion level label;

[0024] Performing low-dimensional embedding coding on each charging demand data in the historical charging demand data to obtain a time series of low-dimensional embedding coding feature vectors of the charging demand data;

[0025] Performing charging demand context encoding based on the full time domain on the time series of the low-dimensional embedded coding feature vector of the charging demand data to obtain a time series of contextual charging demand data semantic coding feature vectors;

[0026] Passing the time series of the semantically encoded feature vector of the contextual charging demand data through a charging demand reasoning module based on prior information to obtain a charging demand reasoning semantic feature vector;

[0027] One-hot encoding the real-time traffic status to obtain a one-hot encoded feature vector of a congestion level label;

[0028] Performing cross-modal association encoding on the charging demand reasoning semantic feature vector and the congestion level label one-hot encoding feature vector to obtain a charging demand reasoning semantic feature matrix under real-time traffic conditions as a charging demand reasoning semantic feature under real-time traffic conditions;

[0029] Based on the semantic features of charging demand reasoning under the real-time traffic state, recommended values ​​of charging and parking space reservation ratios are obtained.

[0030] Compared with the prior art, the present application provides an intelligent parking management system and method for charging and parking supply and demand matching, which obtains historical charging demand data and real-time traffic status, wherein the real-time traffic status is a congestion level label, and uses data analysis and processing based on deep learning neural networks to perform semantic reasoning analysis of the historical charging demand data and data processing of the real-time traffic status, so as to automatically obtain the recommended value of charging and parking space reservation ratio according to the cross-modal association characteristics of the charging demand and the real-time traffic. This method can more comprehensively understand the charging behavior and traffic conditions, meet the demand peaks in different time periods, adapt to changes in traffic conditions, thereby improving the real-time and accuracy of the system, making parking management more intelligent and efficient, thereby improving the user's service experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0032] Figure 14 is a block diagram of an intelligent parking management system for matching charging and parking supply and demand according to an embodiment of the present application.

[0033] Figure 2 Schematic diagram of the architecture of an intelligent parking management system for matching charging and parking supply and demand according to an embodiment of the present application.

[0034] Figure 3 The block diagram is a charging demand reasoning module in an intelligent parking management system for matching charging and parking supply and demand according to an embodiment of the present application.

[0035] Figure 4 4 is a block diagram of a training module in an intelligent parking management system for matching charging and parking supply and demand according to an embodiment of the present application.

[0036] Figure 5 The figure is a flow chart of an intelligent parking management method for matching charging and parking supply and demand according to an embodiment of the present application. DETAILED DESCRIPTION

[0037] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, but rather these embodiments are provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0038] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0039] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0040] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0041] With the development of economic level, information technology and automobile industry, intelligent networked vehicles have become the development trend of future vehicles, and the proportion of electric vehicles has also increased. However, the low parking space utilization rate and less than 30% of charging parking spaces have caused difficulties in urban charging and parking.

[0042] Parking management usually refers to the process of organizing and controlling vehicle parking. However, the existing charging and parking spaces have no reserved management strategy, facing the problems of supply and demand imbalance and excessive competition, that is, some areas may have excess demand, while other areas may have insufficient supply, resulting in uneven resource utilization, which will cause additional waiting time, affect the operation of parking lot resources, and further cause traffic congestion and economic losses. In addition, the existing parking management system cannot allocate resources and optimize management based on historical data and trends, resulting in inefficient resource utilization and low parking management efficiency.

[0043] Therefore, in response to the above technical problems, the technical concept of the present application is to obtain historical charging demand data and real-time traffic status, wherein the real-time traffic status is a congestion level label, and use data analysis and processing based on deep learning neural networks to perform semantic reasoning analysis of the historical charging demand data and data processing of the real-time traffic status, so as to automatically obtain the recommended value of charging and parking space reservation ratio according to the cross-modal association characteristics of the charging demand and the real-time traffic. This method can more comprehensively understand the charging behavior and traffic conditions to meet the demand peaks in different time periods and adapt to changes in traffic conditions, thereby improving the real-time and accuracy of the system, making parking management more intelligent and efficient, thereby improving the user's service experience.

[0044] Figure 1 4 is a block diagram of an intelligent parking management system for matching charging and parking supply and demand according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of an intelligent parking management system for matching charging and parking supply and demand according to an embodiment of the present application. Figure 1 and Figure 2As shown, according to an embodiment of the present application, the intelligent parking management system 100 for charging and parking supply and demand matching includes: a charging demand traffic status data acquisition module 110, which is used to acquire historical charging demand data and real-time traffic status, wherein the real-time traffic status is a congestion level label; a charging demand data embedding coding module 120, which is used to perform low-dimensional embedding coding on each charging demand data in the historical charging demand data to obtain a time series of low-dimensional embedded coding feature vectors of charging demand data; a charging demand context coding module 130, which is used to perform charging demand context coding based on the full time domain on the time series of the low-dimensional embedded coding feature vectors of the charging demand data to obtain a time series of contextual charging demand data semantic coding feature vectors; a charging demand reasoning module 140, which is used to perform charging demand context coding on the time series of the low-dimensional embedded coding feature vectors of the charging demand data to obtain a time series of contextual charging demand data semantic coding feature vectors; and a charging demand reasoning module 141. 0, used for passing the time series of the semantically encoded feature vector of the contextual charging demand data through a charging demand inference module based on prior information to obtain a charging demand inference semantic feature vector; a traffic state one-hot encoding module 150, used for one-hot encoding the real-time traffic state to obtain a congestion level label one-hot encoded feature vector; a feature association encoding module 160, used for cross-modal association encoding of the charging demand inference semantic feature vector and the congestion level label one-hot encoded feature vector to obtain a charging demand inference semantic feature matrix under the real-time traffic state as the charging demand inference semantic feature under the real-time traffic state; and a reservation ratio recommendation module 170, used for obtaining a recommended value of the charging and parking space reservation ratio based on the charging demand inference semantic feature under the real-time traffic state.

[0045] In an embodiment of the present application, the charging demand traffic status data acquisition module 110 is used to obtain historical charging demand data and real-time traffic status, wherein the real-time traffic status is a congestion level label. It should be understood that the historical charging demand data and the real-time traffic status are key information sources in the intelligent parking management system. Specifically, the historical charging demand data records the user's charging demand in the past period of time, including the frequency of use of charging piles, charging time periods, charging duration and other information. The real-time traffic status provides information on the congestion status of the current roads and traffic networks, usually expressed in congestion level labels, such as unblocked, mild congestion, moderate congestion, severe congestion, etc. Based on this, in order to help the system more accurately predict charging demand and optimize parking space allocation, thereby improving the efficiency of the system and user satisfaction, in the technical solution of the present application, historical charging demand data and real-time traffic status are obtained, wherein the real-time traffic status is a congestion level label.

[0046] In an embodiment of the present application, the charging demand data embedding coding module 120 is used to perform low-dimensional embedding coding on each charging demand data in the historical charging demand data to obtain a time series of low-dimensional embedded coding feature vectors of the charging demand data. Accordingly, considering that each historical charging demand data in the historical charging demand data is high-dimensional data, which contains important data features about the historical charging demand, but also contains a large amount of redundant information accordingly, if the historical charging demand data is directly processed, the calculation time is long and the storage space is also large. Based on this, in order to reduce the noise and redundancy in the historical charging demand data, retain key feature information, and thus improve the efficiency of calculation, in the technical solution of the present application, each charging demand data in the historical charging demand data is low-dimensionally embedded and coded to obtain a time series of low-dimensional embedded coding feature vectors of the charging demand data. It is worth mentioning that the low-dimensional embedded coding is a technology that maps high-dimensional data to low-dimensional space, which is usually implemented by dimensionality reduction algorithms (such as principal component analysis, autoencoders, etc.). In this process, the important features of the data are retained, and the redundant information is eliminated, thereby obtaining a more compact and easier to handle feature representation. That is to say, by performing low-dimensional embedded coding on each charging demand data in the historical charging demand data, the original high-dimensional data can be converted into a more compact low-dimensional low-dimensional embedded coding feature vector representation of the charging demand data. In this way, the system can process the charging demand data more efficiently to discover the potential correlation between each charging demand data, thereby providing a basis and support for subsequent analysis and processing.

[0047] In an embodiment of the present application, the charging demand context encoding module 130 is used to perform charging demand context encoding based on the full time domain on the time series of the low-dimensional embedded coding feature vector of the charging demand data to obtain a time series of contextual charging demand data semantic coding feature vectors. Specifically, in an embodiment of the present application, the charging demand context encoding module is used to: pass the time series of the low-dimensional embedded coding feature vector of the charging demand data through a full-time domain charging demand context encoder based on the Transformer layer to obtain a time series of contextual charging demand data semantic coding feature vectors. It should be understood that, considering that there is a contextual semantic association relationship of charging demand about each different local time between each low-dimensional embedded coding feature vector of the charging demand data in the time series of the low-dimensional embedded coding feature vector of the charging demand data, and the semantic information of charging demand at these different times has a good effect on the characterization of charging demand in the full time domain. Based on this, in order to better understand and analyze the charging demand on the entire time scale, in the technical solution of the present application, the time series of the low-dimensional embedded coding feature vector of the charging demand data is passed through a full-time domain charging demand context encoder based on the Transformer layer to obtain a time series of contextual charging demand data semantic coding feature vectors. It is worth mentioning that the Transformer model can effectively capture the contextual information in the sequence data, that is, it can model the dependency relationship between different positions in the sequence through the self-attention mechanism, so as to better understand the time series characteristics of the charging demand data. Specifically, the Transformer can capture the contextual semantic association of charging demand between long time spans in the time series of the low-dimensional embedded coding feature vector of the charging demand data, and can better mine the contextual information and semantic features between each low-dimensional embedded coding feature vector of the charging demand data, so as to obtain a time series of contextual charging demand data semantic coding feature vectors containing rich contextual semantic information.

[0048] In an embodiment of the present application, the charging demand reasoning module 140 is used to pass the time series of the contextual charging demand data semantic encoding feature vector through the charging demand reasoning module based on prior information to obtain the charging demand reasoning semantic feature vector. Accordingly, considering that each contextual charging demand data semantic encoding feature vector in the time series of the contextual charging demand data semantic encoding feature vector contains the charging demand data semantic encoding features between different time periods, there are certain differences and differences between these features. Therefore, in order to more accurately infer the contextual charging demand data semantic encoding features of the next time based on the contextual charging demand data semantic encoding features in the time series of the contextual charging demand data semantic encoding feature vector, and facilitate the subsequent recommendation of the charging and parking space reservation ratio, in the technical solution of the present application, the time series of the contextual charging demand data semantic encoding feature vector is passed through the charging demand reasoning module based on prior information to obtain the charging demand reasoning semantic feature vector. It should be understood that the charging demand reasoning module based on prior information calculates the mean of the time series of the input vector and introduces the hyperparameters of the Gaussian distribution function, so as to optimize the time series feature information of the next time according to the previous time series feature information. Specifically, the contextual charging demand data semantic encoding features at the last position in the time series of the contextual charging demand data semantic encoding feature vector are updated by using the mean feature vector and Gaussian distribution hyperparameters of the time series of the contextual charging demand data semantic encoding feature vector, and the feature vector output at the last position can comprehensively and accurately display the charging demand inference semantic feature information obtained based on the prior contextual charging demand data semantic encoding features, thereby accurately representing the semantic information of the charging demand data.

[0049] Figure 3 is a block diagram of a charging demand reasoning module in an intelligent parking management system for charging and parking supply and demand matching according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 3As shown, the charging demand inference module 140 includes: a position feature determination unit 141, which is used to determine the context charging demand data semantic encoding feature vector of the last position in the time series of the context charging demand data semantic encoding feature vector; an average distribution information difference calculation unit 142, which is used to calculate the position difference between the mean feature vector of multiple context charging demand data semantic encoding feature vectors in the time series of the context charging demand data semantic encoding feature vector and the context charging demand data semantic encoding feature vector of the last position to obtain the context charging demand data semantic average distribution information difference feature vector; a random distribution information difference calculation unit 143, which is used to calculate the time series of the context charging demand data semantic encoding feature vector The positional difference between the contextual charging demand data semantic encoding feature vector at a random position in the time sequence and the contextual charging demand data semantic encoding feature vector at the last position is used to obtain the contextual charging demand data semantic random distribution information difference feature vector; an inference feature vector construction unit 144 is used to construct a first charging demand inference feature vector and a second charging demand inference feature vector based on the contextual charging demand data semantic average distribution information difference feature vector and the contextual charging demand data semantic random distribution information difference feature vector; and a feature positional addition unit 145 is used to calculate the positional addition between the first charging demand inference feature vector and the second charging demand inference feature vector to obtain the charging demand inference semantic feature vector.

[0050] More specifically, in an embodiment of the present application, the inference feature vector construction unit includes: a first charging demand inference subunit, which is used to calculate the product of the natural exponential feature vector of the semantic average distribution information difference feature vector of the context charging demand data as a power and the first Gaussian distribution hyperparameter and the first weight hyperparameter to obtain the first charging demand inference feature vector; and a second charging demand inference subunit, which is used to calculate the product of the natural exponential feature vector of the semantic random distribution information difference feature vector of the context charging demand data as a power and the second Gaussian distribution hyperparameter and the second weight hyperparameter to obtain the second charging demand inference feature vector. In particular, the first Gaussian distribution hyperparameter and the second Gaussian distribution hyperparameter are generated by a Gaussian distribution random number function with a mean of 0 and a variance of 1.

[0051] In the embodiment of the present application, preferably, the charging demand reasoning module is used to: use the charging demand reasoning module based on prior information to process the time series of the context charging demand data semantic encoding feature vector with the following reasoning formula to obtain the charging demand reasoning semantic feature vector; wherein the reasoning formula is:

[0052]

[0053] Among them, p1 and p2 are weight hyperparameters, f1 and f2 are hyperparameters that use the Gaussian distribution random number function with a mean of 0 and a variance of 1 as the coefficients of the Gaussian distribution function, and V i is the contextual charging demand data semantic encoding feature vector at the last position in the time series of the contextual charging demand data semantic encoding feature vector, V rand is the contextual charging demand data semantic encoding feature vector at a random position in the time series of the contextual charging demand data semantic encoding feature vector, V r is the mean feature vector of the time series of the semantic encoding feature vector of the contextual charging demand data, V c is the charging demand reasoning semantic feature vector.

[0054] In the embodiment of the present application, the traffic state unique hot encoding module 150 is used to uniquely encode the real-time traffic state to obtain a unique hot encoding feature vector of a congestion level label. It should be understood that, considering that the real-time traffic state refers to a congestion level label, the level label contains multiple data types, such as "mild congestion", "moderate congestion", and "severe congestion" levels, which are usually discrete categorical data, and machine learning algorithms usually require input of numerical data. Based on this, in order to more clearly understand and analyze each congestion data in the real-time traffic state and make it suitable for the input of the machine learning model, in the technical solution of the present application, the real-time traffic state is uniquely encoded to obtain a unique hot encoding feature vector of a congestion level label. Specifically, using unique hot encoding can avoid the model from misinterpreting the numerical size relationship between the congestion level labels as a meaningful order or size relationship. That is, each level is encoded as a unique hot vector, in which only one element is 1 and the rest are 0, so that the real-time traffic state can be converted into a unique hot encoding feature vector of a congestion level label that can be directly processed by the machine learning model, thereby improving the accuracy and performance of the model.

[0055] In an embodiment of the present application, the feature association coding module 160 is used to perform cross-modal association coding on the charging demand reasoning semantic feature vector and the congestion level label unique hot encoding feature vector to obtain a charging demand reasoning semantic feature matrix under real-time traffic conditions as a charging demand reasoning semantic feature under real-time traffic conditions. Specifically, in an embodiment of the present application, the feature association coding module is used to: pass the charging demand reasoning semantic feature vector and the congestion level label unique hot encoding feature vector through a cross-modal association encoder based on the CLIP model to obtain a charging demand reasoning semantic feature matrix under the real-time traffic conditions. Accordingly, it is considered that the charging demand reasoning semantic feature vector represents semantic information about charging demand, such as characteristics of the user's charging demand degree, time, location, and charging equipment type. The congestion level label unique hot encoding feature vector represents information about the traffic congestion level. At the same time, there is a mutual correlation between the charging demand reasoning semantic feature vector and the congestion level label unique hot encoding feature vector. Therefore, in order to better understand the mutual correlation between the charging demand reasoning semantic features and the congestion level features, so as to realize the semantic interaction between the charging demand and the congestion level, in the technical solution of the present application, the charging demand reasoning semantic feature vector and the congestion level label one-hot encoded feature vector are passed through a cross-modal association encoder based on the CLIP model to obtain the charging demand reasoning semantic feature matrix under real-time traffic conditions. It is worth mentioning that the CLIP model can map the corresponding charging demand reasoning semantic feature vector and the congestion level label one-hot encoded feature vector into a shared semantic space by learning the semantic consistency between them, so that the charging demand reasoning semantic feature vector and the congestion level label one-hot encoded feature vector can be effectively associated to obtain a richer semantic representation, so as to better understand and utilize the semantic association information of charging demand under real-time traffic conditions.

[0056] In an embodiment of the present application, the reserved ratio recommendation module 170 is used to obtain the recommended value of the reserved ratio of charging and parking spaces based on the semantic features of the charging demand reasoning under the real-time traffic state. Specifically, in an embodiment of the present application, the reserved ratio recommendation module is used to: pass the semantic feature matrix of the charging demand reasoning under the real-time traffic state through the charging and parking space reserved ratio recommender based on the decoder to obtain the recommended value of the reserved ratio of charging and parking spaces. That is, the semantic features of the charging demand reasoning under the real-time traffic state obtained by cross-modal association coding using the charging demand reasoning semantic feature vector and the congestion level label unique hot encoding feature vector are decoded and processed, so as to automatically obtain the recommended value of the reserved ratio of charging and parking spaces according to the cross-modal association features of the charging demand and the real-time traffic. Through this method, the charging behavior and traffic conditions can be more comprehensively understood to meet the demand peaks in different time periods and adapt to changes in traffic conditions, thereby improving the real-time and accuracy of the system, making parking management more intelligent and efficient, thereby improving the user's service experience.

[0057] It is worth mentioning that those skilled in the art should be aware that before applying a deep neural network model for inference, the deep neural network model must be trained so that the deep neural network can implement specific functional capabilities.

[0058] Specifically, in the technical solution of the present application, the intelligent parking management system for charging and parking supply and demand matching also includes a training module for training the full-time domain charging demand context encoder based on the Transformer layer, the charging demand reasoning module based on prior information, the cross-modal association encoder based on the CLIP model, and the charging and parking space reservation ratio recommender based on the decoder.

[0059] Figure 4 is a block diagram of a training module in an intelligent parking management system for charging and parking supply and demand matching according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 4As shown, the training module 200 includes: a training data acquisition unit 210, which is used to acquire training data, and the training data includes historical training charging demand data and training real-time traffic status, as well as the actual recommended values ​​of the charging and parking space reservation ratios; a training charging demand data encoding unit 220, which is used to perform low-dimensional embedding encoding on each training charging demand data in the historical training charging demand data to obtain a time series of low-dimensional embedded encoding feature vectors of the training charging demand data; a training charging demand context encoding unit 230, which is used to pass the time series of the low-dimensional embedded encoding feature vectors of the training charging demand data through the full-time domain charging demand context encoder based on the Transformer layer to obtain a time series of semantic encoding feature vectors of the training context charging demand data; a training charging demand inference unit 240, which is used to pass the time series of the semantic encoding feature vectors of the training context charging demand data through the charging demand inference module based on prior information to obtain a training charging demand inference semantic feature vector; a training one-hot encoding unit 250, which is used to perform one-hot encoding on the training real-time traffic status to obtain a training congestion level label one-hot encoding feature vector; training cross-modal association coding Unit 260 is used to pass the training charging demand reasoning semantic feature vector and the training congestion level label unique hot encoded feature vector through the cross-modal association encoder based on the CLIP model to obtain the training real-time traffic state charging demand reasoning semantic feature matrix; feature optimization unit 270 is used to perform feature optimization on the training real-time traffic state charging demand reasoning semantic feature matrix to obtain the optimized training real-time traffic state charging demand reasoning semantic feature matrix; decoding loss unit 280 is used to pass the optimized training real-time traffic state charging demand reasoning semantic feature matrix through the decoder-based charging and parking space reservation ratio recommender to obtain a decoding loss function value; and, training unit 290 is used to train the Transformer layer-based full-time domain charging demand context encoder, the prior information-based charging demand reasoning module, the CLIP model-based cross-modal association encoder and the decoder-based charging and parking space reservation ratio recommender based on the decoding loss function value and through gradient descent back propagation, wherein, in each iteration of the model, the training real-time traffic state charging demand reasoning semantic feature matrix is ​​optimized.

[0060] Specifically, in the technical solution of the present application, in each iteration process of the model, the charging demand reasoning semantic feature matrix under the training real-time traffic state is optimized to obtain an optimized charging demand reasoning semantic feature matrix under the training real-time traffic state, including: dividing each eigenvalue of the charging demand reasoning semantic feature matrix under the training real-time traffic state by the maximum eigenvalue of the charging demand reasoning semantic feature matrix under the training real-time traffic state to obtain the charging demand reasoning semantic interaction representation matrix under the training real-time traffic state; dividing the eigenvalue mean of the charging demand reasoning semantic feature matrix under the training real-time traffic state by the eigenvalue standard deviation of the charging demand reasoning semantic feature matrix under the training real-time traffic state to obtain a statistically significant difference corresponding to the charging demand reasoning semantic feature matrix under the training real-time traffic state. Calculate the interaction value; subtract the statistical interaction value from each eigenvalue of the semantic interaction representation matrix of charging demand reasoning under the training real-time traffic state, and then calculate the logarithm of each position to obtain the semantic interaction information representation matrix of charging demand reasoning under the training real-time traffic state; add the statistical interaction value to each eigenvalue of the semantic interaction representation matrix of charging demand reasoning under the training real-time traffic state, and then multiply it by a predetermined weight hyperparameter to obtain the semantic interaction representation matrix pattern representation matrix of charging demand reasoning under the training real-time traffic state; and add the semantic interaction information representation matrix of charging demand reasoning under the training real-time traffic state to the pattern representation matrix of the semantic interaction representation matrix of charging demand reasoning under the training real-time traffic state to obtain the optimized semantic feature matrix of charging demand reasoning under the training real-time traffic state.

[0061] That is, in order to improve the feature expression effect of the cross-modal global association representation of the training semantic feature matrix of charging demand reasoning under real-time traffic conditions based on the data time series encoding semantics and the state one-hot encoding semantics in the complex feature representation dimension, in the above steps, by using a short sequence of interactive representations of probability statistical features represented by means of means and standard values ​​and interactive representations of distributions of eigenvalues ​​and maximum eigenvalues ​​as latent variable features as sub-manifold latent motifs under the complex manifold network of the training semantic feature matrix of charging demand reasoning under real-time traffic conditions, the latent motif feature information pattern and feature distribution pattern of the obtained semantic feature matrix of charging demand reasoning under real-time traffic conditions based on the training semantic interactive information representation matrix of charging demand reasoning under real-time traffic conditions and the training semantic interactive representation matrix of charging demand reasoning under real-time traffic conditions are used as its A global structure inference unit, through the point addition operation of the semantic interaction information representation matrix of the charging demand reasoning under the training real-time traffic state and the semantic interaction representation matrix pattern representation matrix of the charging demand reasoning under the training real-time traffic state, reconstructs the complex manifold structure of the semantic feature matrix of the charging demand reasoning under the training real-time traffic state in the form of a global structure potential motif dictionary based on connection, so as to improve the generation and evolution understanding ability of the model for the manifold structure corresponding to the features under the complex feature representation dimension in the iteration process, and improve the feature expression effect of the complex feature representation dimension of the semantic feature matrix of the charging demand reasoning under the training real-time traffic state in the model iteration process, so as to improve the accuracy of the charging and parking space reservation ratio recommendation value obtained by the charging and parking space reservation ratio recommender based on the decoder of the semantic feature matrix of the charging demand reasoning under the training real-time traffic state. Through this method, the charging behavior and traffic conditions can be understood more comprehensively to meet the demand peaks in different time periods and adapt to the changes in traffic conditions, thereby improving the real-time performance and accuracy of the system, making parking management more intelligent and efficient, thereby improving the user's service experience.

[0062] In summary, the intelligent parking management system 100 for charging and parking supply and demand matching based on the embodiment of the present application is explained, which obtains historical charging demand data and real-time traffic status, wherein the real-time traffic status is a congestion level label, and uses data analysis and processing based on deep learning neural networks to perform semantic reasoning analysis of the historical charging demand data and data processing of the real-time traffic status, so as to automatically obtain the recommended value of charging and parking space reservation ratio according to the cross-modal association characteristics of the charging demand and the real-time traffic. This method can more comprehensively understand the charging behavior and traffic conditions, meet the demand peaks in different time periods, adapt to changes in traffic conditions, thereby improving the real-time and accuracy of the system, making parking management more intelligent and efficient, thereby improving the user's service experience.

[0063] As described above, the intelligent parking management system 100 for charging and parking supply and demand matching according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with an intelligent parking management algorithm for charging and parking supply and demand matching. In a possible implementation, the intelligent parking management system 100 for charging and parking supply and demand matching according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the intelligent parking management system 100 for charging and parking supply and demand matching can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the intelligent parking management system 100 for charging and parking supply and demand matching can also be one of the many hardware modules of the wireless terminal.

[0064] Alternatively, in another example, the intelligent parking management system 100 for matching charging and parking supply and demand and the wireless terminal may also be separate devices, and the intelligent parking management system 100 for matching charging and parking supply and demand may be connected to the wireless terminal via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0065] Optionally, in another embodiment of the present application, a charging & parking space allocation scheme based on mixed proportion planning is also provided. Specifically, firstly, by mining real-time traffic status, parking lot geographical features and historical charging parking demand and other information, based on geographic information analysis, machine learning and other methods, the future charging parking demand of the parking lot is estimated; secondly, using queuing theory, statistics, Lagrange multiplier method, etc., with minimizing the waiting time of available parking spaces for non-charging and charging vehicles as the objective function, satisfying parking demand, reserving charging & parking spaces for charging vehicles and other constraints, based on constraint programming methods, problem modeling, optimization and solution are performed to determine the future charging parking demand that can be estimated, minimize the waiting time of available parking spaces for the two types of vehicles, and (approximately) the optimal reserved charging & parking space proportion; finally, based on the mining and characterization of user time-money features, using monetary incentive mechanisms, the reserved parking fees for non-charging vehicles are increased, thereby increasing the proportion of charging vehicles, and realizing flexible adjustment of charging and non-charging parking resources.

[0066] In particular, in order to address the problem that charging vehicles have to wait for available charging parking spaces for a long time during peak charging demand periods due to the traditional non-reserved charging & parking space management mechanism, a flexible charging & parking space reservation service mechanism is adopted, and a constraint programming method is used to perform (approximately) optimal charging & parking space reservation ratio scheduling. A monetary incentive mechanism is used to increase the charges for non-charging vehicles in reserved charging & parking spaces, thereby achieving flexible reservation and enhancing the ability of parking lots to fully meet charging & parking needs and flexibly dispatch parking resources.

[0067] Figure 5 FIG. 1 is a flow chart of an intelligent parking management method for matching charging and parking supply and demand according to an embodiment of the present application. Figure 5 As shown, according to the embodiment of the present application, the intelligent parking management method for charging and parking supply and demand matching includes: S110, obtaining historical charging demand data and real-time traffic status, wherein the real-time traffic status is a congestion level label; S120, performing low-dimensional embedding coding on each charging demand data in the historical charging demand data to obtain a time series of low-dimensional embedded coding feature vectors of charging demand data; S130, performing charging demand context coding based on the full time domain on the time series of low-dimensional embedded coding feature vectors of charging demand data to obtain a time series of contextual charging demand data semantic coding feature vectors; S140, encoding the contextual charging demand data semantic coding feature vectors of ... The time series of the semantically encoded feature vector of the demand data is passed through a charging demand inference module based on prior information to obtain a charging demand inference semantic feature vector; S150, the real-time traffic status is one-hot encoded to obtain a congestion level label one-hot encoded feature vector; S160, the charging demand inference semantic feature vector and the congestion level label one-hot encoded feature vector are cross-modally associated encoded to obtain a charging demand inference semantic feature matrix under the real-time traffic status as the charging demand inference semantic feature under the real-time traffic status; and, S170, based on the charging demand inference semantic feature under the real-time traffic status, a recommended value of the charging and parking space reservation ratio is obtained.

[0068] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned intelligent parking management method for matching charging and parking supply and demand have been described in the above reference. Figures 1 to 4 The description of the intelligent parking management system for matching charging and parking supply and demand has been introduced in detail, and therefore, its repeated description will be omitted.

[0069] The above descriptions of various implementations of the present disclosure are exemplary and not exhaustive. The disclosure is not limited to the disclosed implementations, and many modifications and variations are obvious to those of ordinary skill in the art without departing from the scope and spirit of the described implementations. The terms used herein are selected to best explain the principles of the implementations, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the various practical methods disclosed herein.

Claims

1. An intelligent parking management system for charging and parking supply and demand matching, characterized in that: include: A charging demand and traffic status data acquisition module, used to acquire historical charging demand data and real-time traffic status, wherein the real-time traffic status is a congestion level label; A charging demand data embedding coding module, used for performing low-dimensional embedding coding on each charging demand data in the historical charging demand data to obtain a time series of low-dimensional embedding coding feature vectors of the charging demand data; A charging demand context encoding module, used for performing charging demand context encoding based on the full time domain on the time series of the low-dimensional embedding encoding feature vector of the charging demand data to obtain a time series of contextual charging demand data semantic encoding feature vectors; A charging demand reasoning module, used for passing the time series of the semantic encoding feature vector of the contextual charging demand data through the charging demand reasoning module based on prior information to obtain a charging demand reasoning semantic feature vector; A traffic status one-hot encoding module, used for one-hot encoding the real-time traffic status to obtain a one-hot encoding feature vector of a congestion level label; A feature association coding module, used for cross-modal association coding of the charging demand reasoning semantic feature vector and the congestion level label one-hot encoding feature vector to obtain a charging demand reasoning semantic feature matrix under real-time traffic conditions as a charging demand reasoning semantic feature under real-time traffic conditions; The reservation ratio recommendation module is used to obtain the recommended value of the charging and parking space reservation ratio based on the semantic features of the charging demand inference under the real-time traffic state. The charging demand context encoding module is used to: embed the charging demand data into a low-dimensional encoding feature vector of the time series through a full-time domain charging demand context encoder based on the Transformer layer to obtain the context charging demand data semantic encoding feature vector of the time series; The charging demand reasoning module includes: a position feature determination unit, configured to determine the contextual charging demand data semantic encoding feature vector at the last position in the time series of the contextual charging demand data semantic encoding feature vector; an average distribution information difference calculation unit, used for calculating the position difference between the mean feature vector of multiple contextual charging demand data semantic encoding feature vectors in the time series of the contextual charging demand data semantic encoding feature vector and the contextual charging demand data semantic encoding feature vector at the last position to obtain the contextual charging demand data semantic average distribution information difference feature vector; a random distribution information difference calculation unit, used for calculating the position difference between the context charging demand data semantic encoding feature vector at a random position in the time series of the context charging demand data semantic encoding feature vector and the context charging demand data semantic encoding feature vector at the last position to obtain a context charging demand data semantic random distribution information difference feature vector; An inference feature vector construction unit, configured to construct a first charging demand inference feature vector and a second charging demand inference feature vector based on the contextual charging demand data semantic average distribution information difference feature vector and the contextual charging demand data semantic random distribution information difference feature vector; a feature position addition unit, configured to calculate a position addition between the first charging demand reasoning feature vector and the second charging demand reasoning feature vector to obtain the charging demand reasoning semantic feature vector; The inference feature vector construction unit comprises: A first charging demand inference subunit, configured to calculate the product of a natural exponential feature vector whose semantic average distribution information difference feature vector of the contextual charging demand data is a power, a first Gaussian distribution hyperparameter, and a first weight hyperparameter to obtain the first charging demand inference feature vector; A second charging demand inference subunit, configured to calculate the product of a natural exponential feature vector of a semantic random distribution information difference feature vector of the contextual charging demand data with a second Gaussian distribution hyperparameter and a second weight hyperparameter to obtain the second charging demand inference feature vector; The feature association encoding module is used to: pass the charging demand reasoning semantic feature vector and the congestion level label one-hot encoding feature vector through a cross-modal association encoder based on the CLIP model to obtain the charging demand reasoning semantic feature matrix under the real-time traffic state; The reservation ratio recommendation module is used to: pass the charging demand reasoning semantic feature matrix under the real-time traffic state through the charging and parking space reservation ratio recommender based on the decoder to obtain the charging and parking space reservation ratio recommendation value; It also includes a training module for training the Transformer layer-based full-time domain charging demand context encoder, the prior information-based charging demand reasoning module, the CLIP model-based cross-modal association encoder, and the decoder-based charging and parking space reservation ratio recommender; The training module comprises: A training data acquisition unit, used to acquire training data, wherein the training data includes historical training charging demand data and training real-time traffic status, and actual recommended values ​​of the charging and parking space reservation ratios; a training charging demand data encoding unit, configured to perform low-dimensional embedding encoding on each training charging demand data in the historical training charging demand data to obtain a time series of low-dimensional embedding encoding feature vectors of the training charging demand data; A training charging demand context encoding unit, used to embed the training charging demand data into a low-dimensional encoding feature vector of the time series through the full-time domain charging demand context encoder based on the Transformer layer to obtain a time series of semantic encoding feature vectors of the training context charging demand data; A training charging demand reasoning unit, configured to pass the time series of the semantically encoded feature vector of the training context charging demand data through the charging demand reasoning module based on prior information to obtain a training charging demand reasoning semantic feature vector; A training one-hot encoding unit, used for performing one-hot encoding on the training real-time traffic status to obtain a one-hot encoding feature vector of a training congestion level label; A training cross-modal association encoding unit is used to pass the training charging demand reasoning semantic feature vector and the training congestion level label one-hot encoding feature vector through the cross-modal association encoder based on the CLIP model to obtain a training charging demand reasoning semantic feature matrix under real-time traffic conditions; A feature optimization unit, used for performing feature optimization on the training real-time traffic state charging demand reasoning semantic feature matrix to obtain an optimized training real-time traffic state charging demand reasoning semantic feature matrix; A decoding loss unit, used for passing the optimized training real-time traffic state charging demand reasoning semantic feature matrix through the decoder-based charging and parking space reservation ratio recommender to obtain a decoding loss function value; A training unit is used to train the full-time domain charging demand context encoder based on the Transformer layer, the charging demand reasoning module based on prior information, the cross-modal association encoder based on the CLIP model, and the charging and parking space reservation ratio recommender based on the decoder based on the decoding loss function value and through back propagation of gradient descent, wherein in each iteration of the model, the charging demand reasoning semantic feature matrix under the training real-time traffic state is optimized.

2. The intelligent parking management system for charging and parking supply and demand matching according to claim 1 is characterized in that: The first Gaussian distribution hyperparameter and the second Gaussian distribution hyperparameter are generated by a Gaussian distribution random number function with a mean of 0 and a variance of 1.

3. The management method of the intelligent parking management system for charging and parking supply and demand matching according to claim 1, characterized in that: include: Acquire historical charging demand data and real-time traffic status, wherein the real-time traffic status is a congestion level label; Performing low-dimensional embedding coding on each charging demand data in the historical charging demand data to obtain a time series of low-dimensional embedding coding feature vectors of the charging demand data; Performing charging demand context encoding based on the full time domain on the time series of the low-dimensional embedded coding feature vector of the charging demand data to obtain a time series of contextual charging demand data semantic coding feature vectors; Passing the time series of the semantically encoded feature vector of the contextual charging demand data through a charging demand reasoning module based on prior information to obtain a charging demand reasoning semantic feature vector; One-hot encoding the real-time traffic status to obtain a one-hot encoded feature vector of a congestion level label; Performing cross-modal association encoding on the charging demand reasoning semantic feature vector and the congestion level label one-hot encoding feature vector to obtain a charging demand reasoning semantic feature matrix under real-time traffic conditions as a charging demand reasoning semantic feature under real-time traffic conditions; Based on the semantic features of charging demand reasoning under the real-time traffic state, recommended values ​​of charging and parking space reservation ratios are obtained.

Images