An AI-based intelligent charging system for unmanned vessels in a dock

Abstract

The application discloses an intelligent charging system for an unmanned ship dock based on artificial intelligence, and relates to the field of intelligent charging of unmanned ship docks.The system comprises a region division module, a ship type identification module, a charging congestion quantity module, a congestion quantity correction module and a dynamic adjustment module.The region division module is used for dividing the charging region of the unmanned ship dock into several regions.The ship type identification module is used for identifying the type of the unmanned ship to be charged through image recognition technology.The charging congestion quantity module is used for obtaining the quantity of the unmanned ships entering each channel of the dock and calculating the charging congestion quantity of each charging region.The congestion quantity correction module is used for dynamically adjusting the charging congestion quantity of the charging region.The dynamic adjustment module is used for dynamically adjusting the entering channel and charging power of the unmanned ship to be charged based on a distributed Actor-Critic model architecture.The application has the advantages that the entering channel and charging power of the unmanned ship to be charged are intelligently regulated and controlled, and the adaptability and demand of the dock to the charging of different types of unmanned ships are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent charging for unmanned vessel docks, specifically to an intelligent charging system for unmanned vessel docks based on artificial intelligence. Background Technology

[0002] With the accelerated integration of artificial intelligence and autonomous driving technology, building intelligent charging solutions with autonomous adaptation, high efficiency, safety, and collaborative optimization capabilities has become a key issue supporting the large-scale development of the low-altitude economy and intelligent transportation industry. Intelligent charging for unmanned vessels refers to a technical system that enables autonomous adaptation, safe and efficient charging of various types of unmanned vessels through modular interfaces, dynamic power allocation, and standardized communication protocols, while supporting remote monitoring and unmanned operation. Through intelligent power allocation and rapid adjustment of charging channels, the docking time of unmanned vessels can be effectively shortened, and the charging efficiency of unmanned vessel docks can be improved. At the same time, this technology greatly avoids human intervention, reduces labor costs, reduces the risk of human error, and improves the safety and efficiency of the charging process. However, as the number and types of unmanned vessels increase, dynamically adjusting the charging power of unmanned vessel docks and rationally arranging the charging channels of unmanned vessels will become particularly important.

[0003] As the application scenarios of unmanned ships continue to expand, docks, as key infrastructure for their dynamic charging, need to ensure the unification of communication interfaces and protocols between the docks and various types of unmanned ships to ensure that the docks can accommodate intelligent charging of unmanned ships. At the same time, it is necessary to dynamically adjust the charging efficiency of unmanned ships to avoid charging congestion. However, existing technologies lack the ability to identify uncertified unmanned ships, resulting in low universality of unmanned ship charging. In addition, the lack of intelligent allocation of charging channels and dynamic adjustment of charging power leads to uneven utilization of charging piles, excessively long queuing times in some areas, and overall low charging efficiency of unmanned ships. Summary of the Invention

[0004] To address the aforementioned technical problems, an intelligent charging system for unmanned vessels based on artificial intelligence is provided. This technical solution solves the problems mentioned in the background, such as the lack of ability to identify uncertified unmanned vessels, resulting in low universality of charging for unmanned vessels. At the same time, the lack of intelligent allocation of charging channels and dynamic adjustment of charging power leads to uneven utilization of charging piles, excessively long queuing times in some areas, and low overall charging efficiency of unmanned vessels.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] An intelligent charging system for unmanned vessels based on artificial intelligence includes:

[0007] The area division module is used to divide the unmanned vessel dock charging area into several areas according to the types of unmanned vessels currently available in the market, and to charge different types of unmanned vessels respectively.

[0008] The ship type identification module is used to extract the contour features of various types of existing unmanned ships on the market and identify the type of the unmanned ship to be charged through image recognition technology.

[0009] The charging congestion module is used to obtain the number of unmanned ships entering each channel of the dock based on the entry status of the unmanned ships to be charged, and to calculate the charging congestion in each charging area.

[0010] A congestion correction module is used to dynamically adjust the charging congestion in the charging area based on the actual remaining power of the unmanned vessel to be charged.

[0011] The dynamic adjustment module is used to dynamically adjust the entry channel and charging power of the unmanned vessels waiting to be charged based on the charging congestion in each charging area and the type of newly entered unmanned vessels waiting to be charged, using a distributed Actor-Critic model architecture.

[0012] Preferably, the charging congestion module specifically includes:

[0013] An interface data unit is used to obtain the total number of charging interfaces and the number of used charging interfaces in each charging area.

[0014] A power calibration unit is used to obtain the remaining power and target power of the unmanned vessel to be charged according to the communication protocol of the charging interface.

[0015] A charging power setting unit is used to set the charging power of the unmanned vessel in standard mode.

[0016] A pre-charging time unit is used to calculate the pre-charging time of the unmanned vessel to be charged based on the remaining power and charging power of the unmanned vessel to be charged.

[0017] The remaining power estimation unit is used to determine the estimated value of the remaining power of the unmanned ships waiting to be charged in the queue based on the historical charging data of the unmanned ships and the truncated mean formula.

[0018] The entry volume acquisition unit is used to acquire the entry volume of unmanned ships in each channel of the dock based on the entry status of the unmanned ships to be charged.

[0019] A charging congestion acquisition unit is used to establish a charging congestion assessment model for charging areas and calculate the charging congestion in each charging area.

[0020] Preferably, the expression for the charging area congestion assessment model is as follows:

[0021]

[0022] In the formula, S represents the charging congestion level in the charging area, A represents the number of unmanned vessels entering the dock passage, G represents the total number of charging interfaces in the charging area, g represents the number of charging interfaces already in use in the charging area, and T represents the number of charging interfaces already in use. Q-i The pre-charging time for the i-th unmanned vessel entering the dock passage is given, where max{1, Gg} is the value to prevent... The denominator is zero. In particular, the above formula is a dimensionless calculation.

[0023] Preferably, the congestion correction module specifically includes:

[0024] Error data unit, the error data unit is used to obtain error data between the actual value and the estimated value of the remaining power of the unmanned ship to be charged and the error data of the charging congestion in the charging area by using test experiments, and to establish an error dataset.

[0025] The mapping function unit is used to establish a mapping function between the remaining power of the unmanned vessel to be charged and the charging congestion error in the charging area based on the error dataset.

[0026] The compensation coefficient unit is used to determine the compensation coefficient of the charging congestion in the charging area based on the error of the charging congestion value in the charging area.

[0027] A congestion correction unit is used to establish a dynamic adjustment model for charging congestion in the charging area, dynamically adjust the charging congestion in the charging area, and improve the calculation accuracy of charging congestion in the charging area.

[0028] Preferably, the mapping function between the remaining battery power of the unmanned vessel to be charged and the error in charging congestion in the charging area is:

[0029] ΔS = αΔC + β(ΔC) 2

[0030] In the formula, ΔS is the error of the charging congestion value in the charging area, ΔC is the error of the remaining power of the unmanned boat to be charged, α is the linear compensation coefficient, and β is the quadratic linear compensation coefficient, which is used for secondary correction when the error is large. α and β can be determined by solving the error dataset based on the least squares method.

[0031] The expression for the compensation coefficient for charging congestion in the charging area is as follows:

[0032]

[0033] In the formula, λ is the compensation coefficient for charging congestion in the charging area;

[0034] The expression for the dynamic adjustment model of charging congestion in the charging area is:

[0035] S 真 =λS

[0036] In the formula, S 真 The charging congestion level in the charging area is dynamically adjusted.

[0037] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0038] This invention provides an intelligent charging system for unmanned aerial vehicles (UAVs) docks based on artificial intelligence. According to the types of UAVs currently available in the market, the dock charging area is divided into several zones, and a one-to-one correspondence is established between each UAV type and the charging zone and independent channel. Image recognition technology guides the identified UAV types to the charging channels in each zone. Secondly, this solution uses a charging congestion module to obtain the charging congestion situation in each charging zone, and a congestion correction module dynamically adjusts the charging congestion level to ensure accuracy. Finally, based on the charging congestion level of each charging zone and the types of newly arriving UAVs, an intelligent charging model for the dock is constructed based on a distributed Actor-Critic model architecture. This model comprehensively evaluates and dynamically adjusts the entry channels and charging power of the UAVs to be charged, thereby intelligently controlling the entry channels and charging power of the UAVs to be charged, improving the dock's adaptability and demand for charging different types of UAVs. Attached Figure Description

[0039] Figure 1 This is a structural block diagram of an intelligent charging system for an unmanned ship dock based on artificial intelligence, according to the present invention.

[0040] Figure 2 The present invention extracts the contour features of various types of existing unmanned vessels in the market and uses image recognition technology to identify the type of unmanned vessel to be charged. Flowchart.

[0041] Figure 3 The present invention provides a flowchart for obtaining the number of unmanned ships entering each channel of the dock based on the entry status of the unmanned ships to be charged, and calculating the charging congestion in each charging area.

[0042] Figure 4This invention provides a flowchart for dynamically adjusting the charging congestion level in a charging area based on the actual remaining battery power of the unmanned vessel to be charged. Detailed Implementation

[0043] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0044] Reference Figure 1 As shown, an intelligent charging system for unmanned vessels based on artificial intelligence includes:

[0045] The area division module is used to divide the unmanned vessel dock charging area into several areas according to the types of unmanned vessels currently available in the market, and to charge different types of unmanned vessels respectively.

[0046] The ship type identification module is used to extract the contour features of various types of existing unmanned ships on the market and identify the type of the unmanned ship to be charged through image recognition technology.

[0047] The charging congestion module is used to obtain the number of unmanned ships entering each channel of the dock based on the entry status of the unmanned ships to be charged, and to calculate the charging congestion in each charging area.

[0048] A congestion correction module is used to dynamically adjust the charging congestion in the charging area based on the actual remaining power of the unmanned vessel to be charged.

[0049] The dynamic adjustment module is used to dynamically adjust the entry channel and charging power of the unmanned vessels waiting to be charged based on the charging congestion in each charging area and the type of newly entered unmanned vessels waiting to be charged, using a distributed Actor-Critic model architecture.

[0050] Reference Figure 1 As shown, the charging congestion module specifically includes:

[0051] An interface data unit is used to obtain the total number of charging interfaces and the number of used charging interfaces in each charging area.

[0052] A power calibration unit is used to obtain the remaining power and target power of the unmanned vessel to be charged according to the communication protocol of the charging interface.

[0053] A charging power setting unit is used to set the charging power of the unmanned vessel in standard mode.

[0054] A pre-charging time unit is used to calculate the pre-charging time of the unmanned vessel to be charged based on the remaining power and charging power of the unmanned vessel to be charged.

[0055] The remaining power estimation unit is used to determine the estimated value of the remaining power of the unmanned ships waiting to be charged in the queue based on the historical charging data of the unmanned ships and the truncated mean formula.

[0056] The entry volume acquisition unit is used to acquire the entry volume of unmanned ships in each channel of the dock based on the entry status of the unmanned ships to be charged.

[0057] A charging congestion acquisition unit is used to establish a charging congestion assessment model for charging areas and calculate the charging congestion in each charging area.

[0058] Reference Figure 1 As shown, the congestion correction module specifically includes:

[0059] Error data unit, the error data unit is used to obtain error data between the actual value and the estimated value of the remaining power of the unmanned ship to be charged and the error data of the charging congestion in the charging area by using test experiments, and to establish an error dataset.

[0060] The mapping function unit is used to establish a mapping function between the remaining power of the unmanned vessel to be charged and the charging congestion error in the charging area based on the error dataset.

[0061] The compensation coefficient unit is used to determine the compensation coefficient of the charging congestion in the charging area based on the error of the charging congestion value in the charging area.

[0062] A congestion correction unit is used to establish a dynamic adjustment model for charging congestion in the charging area, dynamically adjust the charging congestion in the charging area, and improve the calculation accuracy of charging congestion in the charging area.

[0063] This can be explained by the fact that, with the continuous expansion of unmanned surface vessels (USVs) application scenarios, docks, as key infrastructure for dynamic charging, require unified communication interfaces and protocols to ensure the docks can accommodate intelligent charging for various types of USVs. Simultaneously, the charging efficiency of USVs needs to be dynamically adjusted to avoid charging congestion. Therefore, this solution divides the USV dock charging area into several zones based on the existing types of USVs in the market, establishing a one-to-one correspondence between USV types, charging zones, and independent channels. Then, through image recognition technology, the identified USV types are guided to the appropriate charging channels in each zone. Secondly, this solution obtains the charging congestion status of each charging area by setting a charging congestion module, and dynamically adjusts the charging congestion in each charging area through a congestion correction module to ensure the accuracy of the charging congestion data. Finally, based on the charging congestion in each charging area and the type of newly arrived unmanned vessels waiting to be charged, an intelligent charging model for the unmanned vessel dock is constructed based on a distributed Actor-Critic model architecture. This model comprehensively evaluates and dynamically adjusts the entry channels and charging power of the unmanned vessels waiting to be charged, thereby intelligently controlling the entry channels and charging power of the unmanned vessels waiting to be charged, and improving the dock's adaptability and demand for charging different types of unmanned vessels.

[0064] The method of dividing the unmanned vessel dock charging area into several zones based on the existing types of unmanned vessels in the market, and charging different types of unmanned vessels separately, specifically includes:

[0065] Based on the types of unmanned ships currently available in the market, the unmanned ship dock charging area is divided into several zones. The division of charging zones is mainly based on the type of unmanned ship, charging interface, and communication protocol.

[0066] Based on the division of the dock charging area for unmanned vessels, independent passages leading to each charging area are set up.

[0067] Based on the existing types of unmanned vessels in the market, establish a one-to-one correspondence between each type of unmanned vessel and its charging area and independent passageway.

[0068] The explanation is that the heterogeneity of charging interfaces and communication protocols means the system can only recognize pre-certified vessel types. Multiple types of unmanned vessels experience charging difficulties or charging anomalies due to incompatibility between their communication interfaces and protocols. Specifically, the system cannot obtain charging data from the unmanned vessel being charged, including remaining power, target power, and charging efficiency. Therefore, to ensure the charging safety of unmanned vessels, this solution divides the dock charging area according to the types of unmanned vessels currently on the market, setting up independent channels to each charging area and establishing a one-to-one correspondence between different unmanned vessel types and charging areas and channels. This improves the adaptability of unmanned vessels to charging interfaces and communication protocols, further enhancing charging safety. As the unmanned vessel type certification becomes more detailed, the dock charging service becomes more universally applicable.

[0069] Reference Figure 2 As shown, the extraction of contour features of various types of existing unmanned vessels in the market, and the identification of the type of unmanned vessel to be charged using image recognition technology, specifically includes:

[0070] The system collects multi-angle visual data of the external contour features of various types of unmanned vessels and establishes an unmanned vessel recognition image set, which includes the external contour features of unmanned vessels under different scenes and lighting conditions.

[0071] Based on the grayscale formula, the images in the unmanned vessel recognition image set are processed into grayscale.

[0072] Based on filtering algorithms, noise reduction and smoothing are performed on images in the grayscale unmanned vessel recognition image set.

[0073] Based on the image recognition technology of Canny operator edge detection, the external contour features of the unmanned vessel to be charged are extracted, and a database of unmanned vessel types to be charged is constructed.

[0074] An image acquisition device is set up to obtain visual data of the unmanned vessel to be charged, and the type of the unmanned vessel to be charged is identified based on the database of unmanned vessel types.

[0075] It can be explained that in existing technologies, the limited authentication methods using preset IDs or tags are insufficient to identify third-party unmanned vessels not included in the authentication system. This makes it difficult for third-party unmanned vessels to enter the unmanned vessel dock for charging, or to effectively allocate charging channels for them. Therefore, this solution collects multi-angle visual data of the external contour features of various types of unmanned vessels to establish an unmanned vessel identification image set. It then uses Canny operator edge detection image recognition technology to extract the external contour features of the unmanned vessel to be charged and constructs a database of unmanned vessel types to be charged. Based on the external contour of the unmanned vessel to be charged collected by the image acquisition device and the database of unmanned vessel types to be charged, the type of the unmanned vessel to be charged is comprehensively determined by comparing point sets, thereby completing the identification of the type of unmanned vessel to be charged. The database of unmanned vessel types to be charged includes: the point set coordinates or feature vectors of the contour of the unmanned vessel to be charged and associated metadata, such as object category, size, and allowable error range.

[0076] Reference Figure 3 As shown, the step of obtaining the number of unmanned vessels entering each channel of the dock based on the entry status of the unmanned vessels to be charged, and calculating the charging congestion in each charging area specifically includes:

[0077] Obtain the total number of charging ports and the number of charging ports in use for each charging area;

[0078] According to the communication protocol of the charging interface, obtain the remaining power and target power of the unmanned vessel to be charged;

[0079] Set the charging power of the unmanned ship in standard mode;

[0080] Calculate the pre-charging time of the unmanned vessel to be charged based on its remaining power and charging power.

[0081] Based on the historical charging data of unmanned ships, and using the truncated mean formula, the estimated remaining battery power of unmanned ships waiting to be charged in the queue is determined.

[0082] Based on the entry status of the unmanned ships waiting to be charged, obtain the entry volume of unmanned ships in each channel of the dock.

[0083] Establish a charging congestion assessment model for charging areas and calculate the charging congestion in each charging area;

[0084] The expression for the charging congestion assessment model in the charging area is as follows:

[0085]

[0086] In the formula, S represents the charging congestion level in the charging area, A represents the number of unmanned vessels entering the dock passage, G represents the total number of charging interfaces in the charging area, g represents the number of charging interfaces already in use in the charging area, and T represents the number of charging interfaces already in use. Q-i The pre-charging time for the i-th unmanned vessel entering the dock passage is given, where max{1, Gg} is the value to prevent... The denominator is zero. In particular, the above formula is a dimensionless calculation.

[0087] It can be explained that obtaining the charging congestion level in each charging area is crucial for dynamically adjusting the entry channels and charging power of unmanned surface vessels (USVs) waiting to be charged. Especially with the continuous expansion of USV application scenarios and the increase in the number and types of USVs, the demand for dockside charging is growing. By optimizing the path planning for USVs entering charging channels and dynamically adjusting the charging power allocation strategy, the efficient and coordinated operation of the dockside intelligent charging system can be achieved, significantly improving the utilization rate of charging resources and the overall operational efficiency of the USV fleet. This is an important guarantee for the healthy development of USV dockside intelligent charging. Therefore, this solution establishes a charging congestion assessment model to calculate the charging congestion level in each charging area, thereby providing important data support for the subsequent dynamic adjustment of the entry channels and charging power of USVs waiting to be charged.

[0088] The expression for calculating the pre-charging time of the unmanned vessel to be charged is as follows:

[0089]

[0090] In the formula, T BZ T represents the pre-charging time of the unmanned vessel under standard conditions. CC T is the pre-charging time for the unmanned ship to be charged during the constant current phase. CV T is the pre-charging time for the unmanned ship to be charged during the constant voltage phase. Z C represents the total pre-charging time of the unmanned vessel awaiting charging in a dynamic state. m C represents the target charge level, i.e., the charge level when fully charged. d C is the estimated remaining battery power of the unmanned boat awaiting charging. s P represents the charge at the end of the constant current phase. max P represents the maximum charging power during the constant current phase. avg η is the average power during the constant voltage phase, η is the charging efficiency of the current type of unmanned vessel to be charged, and δ is a correction coefficient, which is an empirical coefficient. During the constant voltage phase, the charging power gradually decreases. If the average power is used directly to calculate the time, the actual time required will be underestimated. Therefore, δ needs to be introduced to match the actual charging time by amplifying the calculated value. It is usually taken as 1.2 to 1.5.

[0091] In the actual charging process, the charging power may change nonlinearly, such as constant current followed by constant voltage during fast charging of lithium batteries. Therefore, the charging time under different transition states is determined according to the intelligent charging requirements of the unmanned ship dock.

[0092] Among them, T Q ∈{T ZB T Z The selection can be achieved through piecewise functions. For example, a boundary threshold can be set based on the number of unmanned vessels entering each channel. When the threshold is exceeded, the pre-charging time under dynamic conditions, i.e., T, can be used. Z The value of T is used; otherwise, the precharge time under standard conditions is used, i.e., T. BZ The value;

[0093] The formula for the truncated mean is:

[0094]

[0095] In the formula, C y The remaining battery level of the unmanned boats waiting to be charged is estimated, where n is the sample size of historical charging data, k is the number of extreme data values ​​removed, and c is the estimated remaining battery level. i The remaining battery power value of the i-th unmanned vessel in the historical charging data;

[0096] Since it is impossible to accurately obtain the remaining battery power of the unmanned ships waiting to be charged in the queue, it is necessary to introduce the estimated value of the remaining battery power of the unmanned ships waiting to be charged in the queue when calculating the charging congestion in each charging area, so as to estimate the charging congestion in the charging area. Therefore, this solution adopts the truncated mean method to obtain the estimated value.

[0097] Reference Figure 4 As shown, the dynamic adjustment of charging congestion in the charging area based on the actual remaining battery power of the unmanned vessel to be charged specifically includes:

[0098] Using test experiments, error data were obtained on the difference between the actual and estimated remaining battery power of the unmanned vessel to be charged and the error data on the charging congestion in the charging area, and an error dataset was established.

[0099] Based on the error dataset, a mapping function is established between the remaining battery power of the unmanned vessel to be charged and the error of charging congestion in the charging area.

[0100] Determine the compensation coefficient for charging congestion in the charging area based on the error in the charging congestion value.

[0101] Establish a dynamic adjustment model for charging congestion in charging areas to dynamically adjust the charging congestion in charging areas and improve the calculation accuracy of charging congestion in charging areas.

[0102] The mapping function between the remaining battery power of the unmanned vessel to be charged and the charging congestion error in the charging area is as follows:

[0103] ΔS = αΔC + β(ΔC) 2

[0104] In the formula, ΔS is the error of the charging congestion value in the charging area, ΔC is the error of the remaining power of the unmanned boat to be charged, α is the linear compensation coefficient, and β is the quadratic linear compensation coefficient, which is used for secondary correction when the error is large. α and β can be determined by solving the error dataset based on the least squares method.

[0105] The expression for the compensation coefficient for charging congestion in the charging area is as follows:

[0106]

[0107] In the formula, λ is the compensation coefficient for charging congestion in the charging area;

[0108] The expression for the dynamic adjustment model of charging congestion in the charging area is:

[0109] S 真 =λS

[0110] In the formula, S 真 The charging congestion level in the charging area is dynamically adjusted.

[0111] This can be explained by the fact that the unmanned vessel dock cannot accurately obtain the remaining power of the unmanned vessels waiting to be charged in the queue. The remaining power of the unmanned vessels waiting to be charged can only be accurately obtained when charging is in progress. Therefore, it is necessary to correct the calculated charging congestion level in the charging area to ensure its accuracy. Thus, this solution obtains the error data between the actual value and the estimated value of the remaining power of the unmanned vessels waiting to be charged, establishes a mapping function between the remaining power of the unmanned vessels waiting to be charged and the error of the charging congestion level in the charging area, thereby reflecting the error data of the charging congestion level in the charging area, determining the compensation coefficient of the charging congestion level in the charging area, and establishing a dynamic adjustment model for the charging congestion level in the charging area to achieve dynamic adjustment of the charging congestion level in the charging area.

[0112] The specific steps of dynamically adjusting the entry channel and charging power of unmanned vessels waiting to be charged based on the charging congestion level in each charging area and the type of newly arrived unmanned vessels waiting to be charged, using a distributed Actor-Critic model architecture, include:

[0113] Based on the identification results of the new unmanned vessels waiting to be charged and the calculation results of the charging congestion in the charging area, establish the data input volume of the same sequence.

[0114] The charging congestion calculation results include: the value of charging congestion in the charging area and the data of each parameter in the dynamic adjustment model of charging congestion in the charging area;

[0115] Set channel labels for each charging area, and use the channel label of the unmanned vessel to be charged and the charging power adjustment amount as the output;

[0116] Based on the distributed Actor-Critic model architecture, an intelligent charging model for unmanned ships in docks is constructed to comprehensively evaluate and dynamically adjust the entry channel and charging power of unmanned ships waiting to be charged.

[0117] It can be explained that the distributed Actor-Critic model architecture is a deep reinforcement learning algorithm that combines asynchronous parallel training with the Actor-Critic framework. It aims to accelerate the training process and improve policy stability through distributed computing. Its core architecture includes parallel worker nodes generating experience data and a global parameter server aggregating learning. This solution uses the identification results of newly entering unmanned vessels waiting to be charged and the calculation results of charging congestion in the charging area as inputs, and the channel label of the unmanned vessel entering and the charging power adjustment amount as outputs. Based on the distributed Actor-Critic model architecture, an intelligent charging model for unmanned vessel docks is constructed. The network architecture of the intelligent charging model for unmanned vessel docks includes:

[0118] A graph neural network is used to process input data, modeling charging piles and unmanned boats in each charging area as nodes of a graph, with connections as edges. Through a three-layer attention mechanism, it outputs 256-dimensional node embeddings and generates a global state vector by mean pooling of all node embeddings.

[0119] The Actor network outputs the selection probability of each channel through a fully connected layer and a Softmax function, and outputs the charging power adjustment amount through a fully connected layer and a Sigmoid function.

[0120] The Critic network, through a fully connected layer, outputs a scalar to predict the long-term cumulative reward of the current state.

[0121] The reward function expression is:

[0122]

[0123] In the formula, R t To predict the reward value for the current state, ω1, ω2, ω3, and ω4 are weighting coefficients, f is the number of unmanned ships that have completed charging, F is the total number of unmanned ships that have been charged, and T is the weighting coefficient. evg The average waiting time for unmanned ships, T 总The total waiting time for unmanned ships in the queue. Let L be the charging power fluctuation rate of the unmanned vessel, and L be the conflict allocation indication function, where...

[0124] In summary, the advantages of this invention are: intelligent control of the entry channel and charging power of the unmanned vessel to be charged, thereby improving the dock's adaptability and ability to meet the charging needs of different types of unmanned vessels.

[0125] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence-based unmanned ship dock intelligent charging system, characterized in that, The application relates to an unmanned ship charging area dynamic adjustment method and device. The application comprises the following: A region division module is used for dividing the unmanned ship dock charging area into several regions according to the types of existing unmanned ships in the market, and different types of unmanned ships are charged respectively; A ship type identification module is used for extracting the contour features of various types of existing unmanned ships in the market, and the type of the unmanned ship to be charged is identified through image recognition technology; A charging congestion amount module is used for obtaining the amount of unmanned ships entering each channel of the dock according to the entering condition of the unmanned ship to be charged, and calculating the charging congestion amount of each charging region; A congestion amount correction module is used for dynamically adjusting the charging congestion amount of the charging region according to the actual condition of the residual power of the unmanned ship to be charged; 2. The intelligent charging system for an unmanned ship dock based on artificial intelligence according to claim 1, characterized in that, A dynamic adjustment module is used for dynamically adjusting the entering channel and charging power of the unmanned ship to be charged according to the charging congestion amount of each charging region and the type of the new unmanned ship to be charged, and based on a distributed Actor-Critic model architecture. The unmanned ship dock charging area is divided into several regions according to the types of existing unmanned ships in the market, and different types of unmanned ships are charged respectively, and the method specifically comprises the following steps: The unmanned ship dock charging area is divided into several regions according to the types of existing unmanned ships in the market, wherein the charging region division is mainly performed through the types of unmanned ships, charging interfaces and communication protocols; According to the division result of the unmanned ship dock charging region, independent channels leading to each charging region are respectively set; 3. The AI-based unmanned ship dock intelligent charging system according to claim 2, wherein According to the types of existing unmanned ships in the market, a one-to-one correspondence relationship between various types of unmanned ships and charging regions and independent channels is established. The contour features of various types of existing unmanned ships in the market are extracted, and the type of the unmanned ship to be charged is identified through image recognition technology, and the method specifically comprises the following steps: A system collects multi-angle visual data of the external contour features of multiple types of unmanned ships, and establishes an unmanned ship identification image set, wherein the unmanned ship identification image set contains the external contour features of unmanned ships under different scenes and light conditions; Based on a gray-scale formula, the images in the unmanned ship identification image set are subjected to gray-scale processing; Based on a filtering algorithm, the images in the gray-scaled unmanned ship identification image set are subjected to noise reduction and smoothing processing; Based on the image recognition technology of Canny operator edge detection, the external contour features of the unmanned ship to be charged are extracted, and a type database of the unmanned ship to be charged is constructed; 4. The intelligent charging system for an unmanned ship dock based on artificial intelligence according to claim 3, characterized in that, An image acquisition device is arranged, visual data of the unmanned ship to be charged are acquired, and the type of the unmanned ship to be charged is identified based on the type database of the unmanned ship to be charged. The charging congestion amount module specifically comprises: An interface data unit is used for obtaining the total number of charging interfaces and the number of used charging interfaces in each charging region; An electric quantity calibration unit is used for obtaining the residual power and target power of the unmanned ship to be charged according to the communication protocol of the charging interface; The charging power setting unit is configured to set the charging power of the unmanned ship in a standard mode; The pre-charging time unit is configured to calculate the pre-charging time of the unmanned ship to be charged according to the remaining power of the unmanned ship to be charged and the charging power; The remaining power estimation unit is configured to determine the estimated value of the remaining power of the unmanned ship to be charged in the queue based on the truncated mean formula according to the historical charging data of the unmanned ship; The driving-in amount acquisition unit is configured to acquire the driving-in amount of the unmanned ship in each channel of the dock according to the driving-in condition of the unmanned ship to be charged; The charging congestion amount acquisition unit is configured to establish a charging area charging congestion amount evaluation model and calculate the charging congestion amount of each charging area.

5. The AI-based unmanned ship dock intelligent charging system according to claim 4, wherein The expression of the charging area charging congestion amount evaluation model is: In the formula, S is the charging congestion value of the charging area, A is the number of unmanned ships entering the dock channel, G is the total number of charging interfaces in the charging area, g is the number of charging interfaces used in the charging area, T Q-i is the pre-charging time of the i-th unmanned ship entering the dock channel to be charged, wherein max{1, G-g} is to prevent The denominator is zero, and the above formula is a dimensionless calculation.

6. The AI-based unmanned ship dock intelligent charging system according to claim 5, wherein, The congestion amount correction module specifically includes: The error data unit is configured to acquire error data of the actual value and the estimated value of the remaining power of the unmanned ship to be charged and error data of the charging congestion amount of the charging area by using a test experiment, and establish an error data set; The mapping function unit is configured to establish a mapping function of the error between the remaining power of the unmanned ship to be charged and the charging congestion amount of the charging area according to the error data set; The compensation coefficient unit is configured to determine the compensation coefficient of the charging congestion amount of the charging area according to the error of the charging congestion amount value of the charging area; The congestion amount correction unit is configured to establish a charging area charging congestion amount dynamic adjustment model, dynamically adjust the charging congestion amount of the charging area, and improve the calculation accuracy of the charging congestion amount of the charging area.

7. The AI-based unmanned ship dock intelligent charging system according to claim 6, wherein, The mapping function of the error between the remaining power of the unmanned ship to be charged and the charging congestion amount of the charging area is: ΔS = aΔC + β(ΔC) 2 In the formula, ΔS is the error of the charging congestion amount value of the charging area, ΔC is the error of the remaining power of the unmanned ship to be charged, α is a linear compensation coefficient, and β is a quadratic linear compensation coefficient for quadratic correction when the error is large, wherein α and β can be determined by solving the error data set based on the least square method; The expression of the compensation coefficient of the charging congestion amount of the charging area is: In the formula, λ is the compensation coefficient of the charging congestion amount of the charging area; The expression of the charging area charging congestion amount dynamic adjustment model is: S 真 = λS In the formula, S 真 is the congestion amount of the charging region after dynamic adjustment.

8. The AI-based unmanned ship dock intelligent charging system according to claim 7, wherein, The dynamic adjustment of the driving-in channel and the charging power of the unmanned ship to be charged based on the distributed Actor-Critic model architecture according to the charging congestion amount of each charging area and the type of the new unmanned ship to be charged specifically includes: According to the type identification result of the new unmanned ship to be charged and the charging congestion amount calculation result, a simultaneous data input amount is established; The charging congestion amount calculation result data includes the value of the charging congestion amount of the charging area and the data of each parameter in the charging congestion amount dynamic adjustment model of the charging area; The channel labels of each charging area are set, and the channel label and the charging power adjustment amount of the unmanned ship to be charged are taken as the output amount; Based on the distributed Actor-Critic model architecture, an unmanned ship dock intelligent charging model is constructed, and the driving-in channel and charging power of the to-be-charged unmanned ship are comprehensively evaluated and dynamically adjusted.

Images