Battery safety management control method for pure electric unmanned ship

Abstract

The invention relates to the technical field of ship task scheduling, and discloses a pure electric unmanned ship battery safety management control method, which comprises the following steps: step S101, collecting historical task data, constructing and training a random forest model, and obtaining required electric quantity of M tasks; s102, collecting second task feature data of the M tasks and ship feature data of the U ships; s103, constructing a fleet task atlas, constructing an adaptation degree prediction model according to the fleet task atlas, and outputting the adaptation degree between the ship and the task; and step S104, generating a distribution scheme according to the adaptation degree prediction model in combination with an improved geyser optimization algorithm, wherein the distribution scheme represents that the task is distributed to the ship. According to the method, the optimal allocation scheme is automatically generated by constructing the adaptation degree prediction model and combining the improved geyser optimization algorithm, maximum coverage and reasonable allocation of task execution are achieved, and the energy utilization efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of ship task scheduling, and in particular relates to a battery safety management and control method for a pure electric unmanned ship. Background Art

[0002] With the rapid development of automation technology, pure electric unmanned ships have been widely used in the fields of ocean monitoring, cargo transportation, environmental monitoring, etc. However, the battery management problem of pure electric unmanned ships has always been one of the key factors restricting their performance and safety. There are many shortcomings in the current battery management methods, especially in task allocation and energy utilization. These problems not only affect the operating efficiency of unmanned ships, but may also lead to shortened battery life and safety hazards.

[0003] At present, the task allocation of most pure electric unmanned ships still relies on manual experience. Operators decide which ships perform which tasks based on their own experience and judgment. Although this method can meet basic needs to a certain extent, it has obvious limitations. First, it is difficult for manual experience to fully consider all factors, such as the priority of the task, the current status of the ship, and environmental conditions, which can easily lead to unreasonable task allocation. Secondly, manual task allocation lacks systematicity and scientificity, and it is easy to repeat tasks or omit tasks, resulting in waste of resources and low work efficiency. In addition, the task allocation method relying on manual experience cannot monitor the remaining power and health status of the battery in real time, and it is difficult to detect and handle battery abnormalities in a timely manner. Once the battery fails or is over-discharged, it may cause the unmanned ship to fail to operate normally, or even cause a safety accident. Summary of the invention

[0004] The present invention provides a battery safety management and control method for a pure electric unmanned ship, which solves the technical problems in the related technology of reliance on manual experience, unreasonable task allocation, insufficient energy utilization and low operating efficiency.

[0005] The present invention provides a battery safety management and control method for a pure electric unmanned ship, comprising the following steps:

[0006] Step S101, collect historical task data, build and train a random forest model based on the normalized historical task data; and collect the first task feature data and environmental feature data of M tasks in the initial stage, normalize them, input them into the random forest model, and output the required power of M tasks, where M is a custom parameter;

[0007] The historical task data includes: the first task characteristic data of the historical tasks and the environmental characteristic data of the historical tasks, as well as the required power of the corresponding historical tasks. The first task characteristic data includes: task type, task attribute, task status, required equipment and task duration. The environmental characteristic data includes: wind speed, wind direction, temperature, current speed and current direction.

[0008] Step S102, collecting the second task feature data of M tasks and the ship feature data of U ships within the first preset time period T according to the first preset time interval t, and representing the second task feature data of the M tasks and the ship feature data of U ships at each time point by a first feature sequence and a second feature sequence respectively, preprocessing the first feature sequence and the second feature sequence to obtain a third feature sequence and a fourth feature sequence respectively, wherein T and U are custom parameters;

[0009] The ship characteristic data includes: remaining power, power status, current task progress and current task required power; the second task characteristic data includes: task type, task attribute, task status, task duration, required power and priority;

[0010] Step S103, constructing a fleet task map according to the third feature sequence and the fourth feature sequence, and constructing a fitness prediction model according to the fleet task map, wherein the input of the fitness prediction model is the fleet task map, and the output is the fitness between the ship and the task;

[0011] Step S104, generating an allocation plan based on the fitness prediction model combined with the improved geyser optimization algorithm, where the allocation plan represents allocating tasks to ships.

[0012] Furthermore, the task types include: cruising, data collection, transportation and rescue, and the task types are represented by one-hot coding; the task attributes include: ordinary tasks and special tasks, and the task attributes are represented by one-hot coding. When the task attribute is a special task, the task can only be executed by a designated ship; the priorities include: low priority, medium priority and high priority, and the priorities are represented by real number coding; the task status is represented by real number coding, and the task status includes: executed, executing and not executed; the required equipment is represented by one-hot coding; the wind direction and the current direction are both represented by two-dimensional vector coding; the power status includes: standby status and normal status, wherein the standby status indicates that the remaining battery power of the current ship is lower than the first preset threshold, and the ship enters the energy-saving mode, and the ship in the standby state does not perform the task.

[0013] Furthermore, the collected historical task data are normalized using the maximum and minimum normalization method and used as training samples for the random forest model, wherein the first task feature data of the historical tasks and the environmental feature data of the historical tasks are used as sample data for training the random forest model, and the corresponding demand power of the historical tasks are used as sample labels for training the random forest model.

[0014] Further, the first feature sequence and the second feature sequence are preprocessed to obtain a third feature sequence and a fourth feature sequence respectively, including:

[0015] Step S201, use 3 The principle identifies outliers in the first and second feature sequences and replaces them with the mean value;

[0016] Step S202, calculating the power utilization rate of the ship in executing the current task, and splicing it after the sequence unit of the second characteristic sequence;

[0017] Step S203, calculate the urgency of the current task and splice it after the sequence unit of the first feature sequence, wherein the calculation formula of the urgency of the current task is: ,in Indicates the urgency of the current task. Indicates the priority of the current task. Indicates the power consumed by the current task, P indicates the progress of the current task, Indicates the current task duration. Indicates the total task duration, and They represent the remaining power weight coefficient and the task duration weight coefficient respectively. and represent wind speed and ocean current speed respectively, and Represent wind direction and ocean current direction, and Represent the cosine values ​​of wind direction and current direction, represents the temperature influence coefficient, T represents the temperature, represents the smoothing coefficient;

[0018] Step S204, removing the tasks with task types of "executed" and "being executed" in the first feature sequence, and the ships with power status of "standby" in the second feature sequence, to obtain a first feature sequence including J sequence units and a second feature sequence including K sequence units, where J and K are custom parameters;

[0019] Step S205: preprocess the first feature sequence and the second feature sequence using a maximum-minimum normalization method to obtain a third feature sequence and a fourth feature sequence.

[0020] Furthermore, the fleet task graph includes: task nodes, ship nodes, edges, initial features of task nodes and initial features of ship nodes; a mapping relationship is established between task nodes and tasks, a mapping relationship is established between ship nodes and ships, the initial features of task nodes are represented by second task feature data of task nodes with which a mapping relationship is established, and the initial features of ship nodes are represented by ship feature data of ship nodes with which a mapping relationship is established; edge construction methods include: constructing edges between special tasks and designated ships corresponding to the tasks; and establishing connections between ordinary tasks and any ships.

[0021] Further, the fitness prediction model includes: a first hidden layer, an extraction layer, a second hidden layer and a classifier;

[0022] The first hidden layer includes N first units, the nth first unit inputs the initial feature of the node at the nth time point, and outputs the first updated feature, wherein N=T / t, 1≤n≤N;

[0023] The extraction layer is used to extract the first updated feature output by the Nth first unit and use it as the initial feature of the node;

[0024] The second hidden layer inputs the fleet mission graph and outputs a graph update matrix, where each row vector of the graph update matrix represents the second updated feature of a node;

[0025] The second updated feature of each node is input into the classifier, and the classification space of the classifier represents the fitness between the ship and the task, and the above nodes include task nodes and ship nodes.

[0026] Furthermore, the first hidden layer is constructed based on the improved gated recurrent unit. The specific improvements include: introducing an adjustment coefficient in the update gate, adding an attenuation parameter in the reset gate, embedding a residual connection, and adding a trend control gate.

[0027] Furthermore, the calculation formula of the second hidden layer includes: ;

[0028] in, represents the graph update matrix of the second hidden layer output, pile represents the stacking operation, represents the dynamic weighting coefficient, represents the degree matrix, and the elements of the degree matrix represent the number of edges connected to the node. represents the adjacency matrix, the elements of which represent the connection relationship between nodes, W represents the first weight coefficient of the second hidden layer, Represents the index influence coefficient, which is used to control the index term The influence of P represents the feature screening matrix, which is used to screen the features of enhanced task priority and the features of remaining power of the ship. Represents the edge weight adjustment coefficient, which is used to control the weight of the feature screening item. I represents the unit matrix. The element values ​​on the diagonal of the unit matrix are assigned 1, and the values ​​of the other elements are assigned 0. represents the second updated feature of the node, represents the absolute value of the element in the i-th row and j-th column of the initial feature matrix, ReLU represents the ReLU activation function, and the above-mentioned nodes include task nodes and ship nodes.

[0029] Furthermore, the steps of generating an allocation plan based on the fitness prediction model combined with the improved geyser optimization algorithm include:

[0030] Step S301, randomly create a starting population with Q individuals that meet the constraint conditions, where Q is a custom parameter;

[0031] The individuals of the initial population are encoded by the vector Indicates that Indicates that the 1st task to the Jth task are assigned to the To vessels;

[0032] Step S302, constructing an objective function in combination with a fitness prediction model;

[0033] Step S303, calculating the fitness value of each individual in the initial population according to the objective function;

[0034] Step S304, using the roulette wheel selection method to select half of the individuals with higher fitness values ​​in the population as channels, and the selection probability is proportional to the fitness value of the individual;

[0035] Step S305, calculating the Euclidean distance between each individual and other individuals, and selecting y individuals with the closest Euclidean distance as neighbors of the current individual;

[0036] Step S306, updating the position of the individual according to the individual with the largest current fitness value and the individual's neighbors;

[0037] Step S307, calculating the pressure on the individual and the probability of selecting a new channel, and updating the position of the individual when selecting a new channel in combination with the Cauchy distribution;

[0038] Step S308, the current number of iterations is increased by 1, and if it is determined that the current number of iterations reaches the maximum number of iterations, the current individual code is output as the optimal allocation solution, otherwise, the process returns to step S303, where the maximum number of iterations is a custom parameter.

[0039] Furthermore, the objective function in step S302 is: ;

[0040] Where F represents the individual fitness value, represents the i-th task and the ship assigned to it The degree of fit between It represents the remaining power of the e-th ship after removing the power required by the task assigned to the ship. and Represent the urgency weight coefficient and distance penalty coefficient respectively, Indicates the urgency of the task. Indicates the distance from the current position of the ship to the mission location.

[0041] The beneficial effects of the present invention are as follows: by constructing a fleet task map and then constructing a fitness prediction model, the present invention can automatically evaluate the matching degree between each task and the ship, and then complete the task allocation in a systematic and scientific manner, avoiding the inefficiency and resource waste caused by relying on manual experience, and improving the operating efficiency of the unmanned ship, achieving maximum coverage and rational allocation of task execution;

[0042] The present invention generates allocation schemes by modeling multiple features of tasks and ships and combining them with an improved geyser optimization algorithm, which can ensure the optimal resource adaptation under different task requirements;

[0043] The present invention uses an intelligent power status monitoring mechanism to track the remaining power and health status of the ship's battery in real time, ensuring that the system always takes the battery status into consideration during mission execution. This real-time monitoring and dynamic management mechanism can effectively avoid over-discharge, extend battery life, and improve energy efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of a battery safety management and control method for a pure electric unmanned ship of the present invention;

[0045] Figure 2 is a flow chart of preprocessing the first feature sequence and the second feature sequence of the present invention;

[0046] Figure 3 is a flow chart of the improved geyser optimization algorithm of the present invention. DETAILED DESCRIPTION

[0047] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0048] like Figure 1-Figure 3 As shown, a battery safety management and control method for a pure electric unmanned ship includes the following steps:

[0049] Step S101, collect historical task data, build and train a random forest model based on the normalized historical task data; and collect the first task feature data and environmental feature data of M tasks in the initial stage, normalize them, input them into the random forest model, and output the required power of M tasks, where M is a custom parameter;

[0050] The historical task data includes: the first task characteristic data of the historical tasks and the environmental characteristic data of the historical tasks, as well as the required power of the corresponding historical tasks. The first task characteristic data includes: task type, task attribute, task status, required equipment and task duration. The environmental characteristic data includes: wind speed, wind direction, temperature, current speed and current direction.

[0051] Step S102, collecting the second task feature data of M tasks and the ship feature data of U ships within the first preset time period T according to the first preset time interval t, and representing the second task feature data of the M tasks and the ship feature data of U ships at each time point by a first feature sequence and a second feature sequence respectively, preprocessing the first feature sequence and the second feature sequence to obtain a third feature sequence and a fourth feature sequence respectively, wherein T and U are custom parameters;

[0052] The ship characteristic data includes: remaining power, power status, current task progress and current task required power; the second task characteristic data includes: task type, task attribute, task status, task duration, required power and priority;

[0053] Step S103, constructing a fleet task map according to the third feature sequence and the fourth feature sequence, and constructing a fitness prediction model according to the fleet task map, wherein the input of the fitness prediction model is the fleet task map, and the output is the fitness between the ship and the task;

[0054] Step S104, generating an allocation plan based on the fitness prediction model combined with the improved geyser optimization algorithm, where the allocation plan represents allocating tasks to ships.

[0055] In one embodiment of the present invention, the task types include: cruising, data collection, transportation and rescue, and the task types are represented by one-hot coding, for example, when the task type is cruising, it is represented by [1, 0, 0, 0]; the task attributes include: ordinary tasks and special tasks, and the task attributes are represented by one-hot coding, for example, when the task attribute is an ordinary task, it is represented by [1, 0], and when the task attribute is a special task, the task can only be performed by a designated ship; the priority includes: low priority, medium priority and high priority, and the priority is represented by real number codes 1 to 3, for example, when the priority is medium priority, it is represented by 2; the task status is represented by real number codes 1 to 3, and the task status includes: executed, executing and not executed, for example, when the priority is medium priority, it is represented by 2; the required equipment is represented by one-hot coding, for example, the required equipment includes: radar, sonar, camera, satellite phone and high-intensity flashlight, and when the required equipment is radar and sonar, it is represented by [1, 1, 0, 0, 0]; the wind direction and the current direction are both represented by two-dimensional vector coding, and the mathematical expression is: , ,in Indicates the sine value of wind direction or current direction, Represents the cosine value of wind direction or current direction, Cd represents the angle of wind direction or current direction; the power status includes: standby status and normal status, wherein the standby status indicates that the current remaining battery power of the vessel is lower than the first preset threshold value, the vessel enters the energy-saving mode, the vessel in the standby status does not perform the mission, and the return and communication capabilities of the vessel are guaranteed. Preferably, the first preset threshold value is set to 20%.

[0056] In one embodiment of the present invention, the system provides a power status monitoring mechanism that can check the power status of the vessel at a second preset time interval, timely update the power status characteristics of the vessel, and ensure the accuracy of the power status; when the remaining power of the vessel is lower than the first preset threshold, the vessel will automatically switch to standby mode, and the system will issue a notification to prompt the current vessel to enter energy-saving mode.

[0057] In one embodiment of the present invention, the collected historical task data is normalized using the maximum and minimum normalization method and used as training samples for the random forest model, wherein the first task feature data of the historical tasks and the environmental feature data of the historical tasks are used as sample data for training the random forest model, and the corresponding power demand of the historical tasks are used as sample labels for training the random forest model.

[0058] In one embodiment of the present invention, preprocessing the first feature sequence and the second feature sequence to obtain a third feature sequence and a fourth feature sequence respectively includes:

[0059] Step S201, use 3 The principle is to identify outliers in the first feature sequence and the second feature sequence and replace them with the average value. Specifically, the mean and standard deviation of the features in the first feature sequence and the second feature sequence are calculated, and the normal data range is determined as: , detect outliers outside the range, replace the detected outliers with the average value to ensure the stability and accuracy of the data, where is the mean, is the standard deviation;

[0060] Step S202, calculate the power utilization rate of the ship in executing the current task, and splice it after the sequence unit of the second characteristic sequence, wherein the calculation formula of the power utilization rate is: , where R represents the power utilization rate, Indicates the power consumed by the current task. Indicates the remaining power. The higher the power utilization rate, the more the remaining power of the ship is required to meet the current mission.

[0061] Step S203, calculate the urgency of the current task and splice it after the sequence unit of the first feature sequence, wherein the calculation formula of the urgency of the current task is: ,in Indicates the urgency of the current task. Indicates the priority of the current task. Indicates the power consumed by the current task, P indicates the progress of the current task, Indicates the current task duration. Indicates the total task duration, and They represent the remaining power weight coefficient and the task duration weight coefficient respectively. and represent wind speed and ocean current speed respectively, and Represent wind direction and ocean current direction, and Represent the cosine values ​​of wind direction and current direction, represents the temperature influence coefficient, T represents the temperature, represents the smoothing coefficient;

[0062] Step S204, removing the tasks with task types of "executed" and "being executed" in the first feature sequence, and the ships with power status of "standby" in the second feature sequence, to obtain a first feature sequence including J sequence units and a second feature sequence including K sequence units, where J and K are custom parameters;

[0063] Step S205, using the maximum and minimum normalization method to pre-process the first feature sequence and the second feature sequence to obtain a third feature sequence and a fourth feature sequence.

[0064] In one embodiment of the present invention, the fleet task graph includes: task nodes, ship nodes, edges, initial features of task nodes and initial features of ship nodes; a mapping relationship is established between task nodes and tasks, a mapping relationship is established between ship nodes and ships, the initial features of task nodes are represented by second task feature data of task nodes with which a mapping relationship is established, and the initial features of ship nodes are represented by ship feature data of ship nodes with which a mapping relationship is established; edge construction methods include: constructing edges between special tasks and designated ships corresponding to the tasks; and establishing connections between ordinary tasks and any ships.

[0065] In one embodiment of the present invention, the fitness prediction model includes: a first hidden layer, an extraction layer, a second hidden layer and a classifier;

[0066] The first hidden layer includes N first units, the nth first unit inputs the initial feature of the node at the nth time point, and outputs the first updated feature, where N=T / t, 1≤n≤N;

[0067] The extraction layer is used to extract the first updated feature output by the Nth first unit and use it as the initial feature of the node;

[0068] The second hidden layer inputs the fleet mission graph and outputs a graph update matrix, where each row vector of the graph update matrix represents the second updated feature of a node;

[0069] The second updated feature of each node is input into the classifier, and the classification space of the classifier represents the fitness between the ship and the task, and the above nodes include task nodes and ship nodes.

[0070] In one embodiment of the present invention, the first hidden layer is constructed based on the improved gated recurrent unit, and the specific improvements include: introducing an adjustment coefficient in the update gate, adding an attenuation parameter in the reset gate, embedding a residual connection, and adding a trend control gate.

[0071] Specifically, the formula of the nth first unit of the first hidden layer includes: ; ; ; ; ;

[0072] in, Represents the initial features of the node at the nth time point of the current input, and represents the first updated features of the nth and n-1th first unit outputs, , , and They represent the update gate, reset gate, trend control gate and candidate hidden state of the first unit of the nth hidden layer, respectively. , and They respectively represent the first weight coefficient, the second weight coefficient and the bias parameter of the update gate of the nth first unit of the first hidden layer, represents the modulation coefficient, , and They respectively represent the first weight coefficient, the second weight coefficient and the bias parameter of the reset gate of the nth first unit of the first hidden layer, represents the attenuation parameter, , and They respectively represent the first weight coefficient, the second weight coefficient and the bias parameter of the trend control gate of the nth first unit of the first hidden layer, , and represent the first weight coefficient, the second weight coefficient and the bias parameter of the candidate hidden state of the nth first unit of the first hidden layer, respectively, represents the weight coefficient of the residual connection, represents point-wise multiplication, and sigmoid represents the sigmoid activation function.

[0073] Specifically, the adjustment coefficient can dynamically adjust the influence of the hidden state at the previous moment, making the model more flexible in balancing the proportion of historical state and current task or ship state input, thereby enhancing the adaptability to complex tasks; the attenuation parameter is added to the reset gate to reduce the interference of short-term fluctuations on the model, so that the model can more accurately retain the long-term trend information of features such as ship power and task progress, avoid overreaction caused by short-term changes, and improve the stability of the model. On this basis, the model adopts residual connection so that the key information in the task and ship characteristics can be directly transmitted to the subsequent layers, effectively avoiding the loss of important information when dealing with short-term fluctuations, thereby improving the model's expressiveness and anti-interference ability in complex time series. In addition, the newly added trend control gate can specifically capture the long-term trend of the input data, so that the model can learn the change pattern of ship and task characteristics over a long period of time. This gated design enables the model to effectively focus on the long-term trends of tasks and ships when dealing with short-term fluctuations, significantly improving the overall prediction effect.

[0074] In one embodiment of the present invention, the calculation formula of the second hidden layer includes: ;

[0075] in, represents the graph update matrix of the second hidden layer output, pile represents the stacking operation, represents the dynamic weighting coefficient, represents the degree matrix, and the elements of the degree matrix represent the number of edges connected to the node. represents the adjacency matrix, the elements of which represent the connection relationship between nodes, W represents the first weight coefficient of the second hidden layer, Represents the index influence coefficient, which is used to control the index term The influence of P represents the feature screening matrix, which is used to screen the features of enhanced task priority and the features of remaining power of the ship. Represents the edge weight adjustment coefficient, which is used to control the weight of the feature screening item. I represents the unit matrix. The element values ​​on the diagonal of the unit matrix are assigned 1, and the values ​​of the other elements are assigned 0. represents the second updated feature of the node, represents the absolute value of the element in the i-th row and j-th column of the initial feature matrix, ReLU represents the ReLU activation function, and the above-mentioned nodes include task nodes and ship nodes.

[0076] In one embodiment of the present invention, the step of generating an allocation scheme according to the fitness prediction model combined with the improved geyser optimization algorithm includes:

[0077] Step S301, randomly create a starting population with Q individuals that meet the constraint conditions, where Q is a custom parameter;

[0078] The constraints include: special tasks can only be assigned to designated ships; each task can only be undertaken by one ship; the power of each ship after the task is assigned must be higher than the first preset threshold;

[0079] The individuals of the initial population are encoded by the vector Indicates that Indicates that the 1st task to the Jth task are assigned to the To ships, such as , indicating that the fifth task is assigned to the fourth ship.

[0080] Step S302, constructing an objective function in combination with a fitness prediction model;

[0081] Step S303, calculating the fitness value of each individual in the initial population according to the objective function;

[0082] Step S304, using the roulette wheel selection method to select half of the individuals with higher fitness values ​​in the population as channels, the selection probability is proportional to the fitness value of the individual, and the formula for the selection probability is: ,in represents the probability of selecting the current channel, and Represents an individual and individuals The fitness value of

[0083] Step S305, calculate the Euclidean distance between each individual and other individuals, and select y individuals with the closest Euclidean distance as neighbors of the current individual. The neighbors of these individuals are usually individuals that are close to the current individual in the search space. The purpose of selecting neighbors is to use local information to guide the search direction of the individual. The distance formula is: ,in Represents an individual With other individuals The Euclidean distance of and Respectively represent the i-th element value of individual a and individual b, that is, the ship number to which the i-th task of individual a and individual b is assigned. Preferably, y is set to 5;

[0084] Step S306, based on the individual with the largest current fitness value and the individual's neighbors, the individual's position is updated. The purpose of the position update is to guide the individual to move closer to a better solution. The calculation formula for the individual position update is: ,in represents the updated position of the individual, represents the location of the neighbors of an individual, Represents a random number between 0 and 1. Indicates the individual position with the largest current fitness value, Indicates the current individual's position;

[0085] Step S307, calculating the pressure on the individual and the probability of selecting a new channel, and updating the position of the individual when selecting a new channel in combination with the Cauchy distribution; the calculation formula for the pressure on the individual is: ,in Represents an individual pressure, Indicates the current iteration number, and Represents the minimum and maximum fitness values ​​in the current population; the formula for the probability of an individual selecting a new channel is: ,in represents the probability of an individual selecting a new channel; the formula for updating the individual position when selecting a new channel is: ,in represents the updated position of the individual when a new channel is selected, represents the individual with the largest current fitness value, represents a random number drawn from a Cauchy distribution with location parameter 0 and scale parameter 1;

[0086] Specifically, the pressure of individuals will guide them to move to specific channels. The purpose of individuals choosing new channels is to increase the diversity of search and prevent the algorithm from converging to the local optimal solution too early. Integrating the Cauchy distribution into the individual position update can effectively utilize the long-tail characteristics of the Cauchy distribution, increase the global search ability of the algorithm, and jump out of the local optimal solution.

[0087] Step S308, the current number of iterations is increased by 1, and if it is determined that the current number of iterations reaches the maximum number of iterations, the current individual code is output as the optimal allocation solution, otherwise, the process returns to step S303, wherein the maximum number of iterations is a custom parameter, and preferably, the maximum number of iterations is set to 200.

[0088] In one embodiment of the present invention, the objective function in step S302 is: ;

[0089] Where F represents the individual fitness value, represents the i-th task and the ship assigned to it The degree of fit between It represents the remaining power of the e-th ship after removing the power required by the task assigned to the ship. and Represent the urgency weight coefficient and distance penalty coefficient respectively, Indicates the urgency of the task. Indicates the distance from the current position of the ship to the mission location; by maximizing the matching degree and the remaining power, the efficiency of mission execution can be improved, energy consumption can be reduced, and the battery life can be extended. Real-time monitoring of the remaining power of the ship can avoid battery damage caused by excessive discharge, thereby achieving safe battery management.

[0090] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A battery safety management and control method for a pure electric unmanned ship, characterized in that: The following steps are involved: Step S101, collect historical task data, build and train a random forest model based on the normalized historical task data; and collect the first task feature data and environmental feature data of M tasks in the initial stage, normalize them, input them into the random forest model, and output the required power of M tasks, where M is a custom parameter; The historical task data includes: the first task characteristic data of the historical tasks and the environmental characteristic data of the historical tasks, as well as the required power of the corresponding historical tasks. The first task characteristic data includes: task type, task attribute, task status, required equipment and task duration. The environmental characteristic data includes: wind speed, wind direction, temperature, current speed and current direction. Step S102, collecting the second task feature data of M tasks and the ship feature data of U ships within the first preset time period T according to the first preset time interval t, and representing the second task feature data of the M tasks and the ship feature data of U ships at each time point by a first feature sequence and a second feature sequence respectively, preprocessing the first feature sequence and the second feature sequence to obtain a third feature sequence and a fourth feature sequence respectively, wherein T and U are custom parameters; The ship characteristic data includes: remaining power, power status, current task progress and current task required power; the second task characteristic data includes: task type, task attribute, task status, task duration, required power and priority; Step S103, constructing a fleet task map according to the third feature sequence and the fourth feature sequence, and constructing a fitness prediction model according to the fleet task map, wherein the input of the fitness prediction model is the fleet task map, and the output is the fitness between the ship and the task; Step S104, generating an allocation plan based on the fitness prediction model combined with the improved geyser optimization algorithm, where the allocation plan represents allocating tasks to ships.

2. A battery safety management and control method for a pure electric unmanned ship according to claim 1, characterized in that: The task types include: cruising, data collection, transportation and rescue, and the task types are represented by one-hot coding; the task attributes include: ordinary tasks and special tasks, and the task attributes are represented by one-hot coding. When the task attribute is a special task, the task can only be executed by a designated ship; the priorities include: low priority, medium priority and high priority, and the priorities are represented by real number coding; the task status is represented by real number coding, and the task status includes: executed, executing and not executed; the required equipment is represented by one-hot coding; the wind direction and the current direction are both represented by two-dimensional vector coding; the power status includes: standby status and normal status, wherein the standby status indicates that the remaining battery power of the current ship is lower than the first preset threshold, and the ship enters the energy-saving mode, and the ship in the standby state does not execute the task.

3. A battery safety management and control method for a pure electric unmanned ship according to claim 1, characterized in that: The collected historical task data are normalized using the maximum and minimum normalization method and used as training samples for the random forest model. The first task feature data of the historical tasks and the environmental feature data of the historical tasks are used as sample data for training the random forest model, and the corresponding power demand of the historical tasks are used as sample labels for training the random forest model.

4. A battery safety management and control method for a pure electric unmanned ship according to claim 1, characterized in that: Preprocessing the first feature sequence and the second feature sequence to obtain a third feature sequence and a fourth feature sequence respectively includes: Step S201, use 3 The principle identifies outliers in the first and second feature sequences and replaces them with the mean value; Step S202, calculating the power utilization rate of the ship in executing the current task, and splicing it after the sequence unit of the second characteristic sequence; Step S203, calculate the urgency of the current task and splice it after the sequence unit of the first feature sequence, wherein the calculation formula of the urgency of the current task is: ,in Indicates the urgency of the current task. Indicates the priority of the current task. Indicates the power consumed by the current task, P indicates the progress of the current task, Indicates the current task duration. Indicates the total task duration, and They represent the remaining power weight coefficient and the task duration weight coefficient respectively. and represent wind speed and ocean current speed respectively, and Represent wind direction and ocean current direction, and Represent the cosine values ​​of wind direction and current direction, represents the temperature influence coefficient, T represents the temperature, represents the smoothing coefficient; Step S204, removing the tasks with task types of "executed" and "being executed" in the first feature sequence, and the ships with power status of "standby" in the second feature sequence, to obtain a first feature sequence including J sequence units and a second feature sequence including K sequence units, where J and K are custom parameters; Step S205, using the maximum and minimum normalization method to pre-process the first feature sequence and the second feature sequence to obtain a third feature sequence and a fourth feature sequence.

5. A battery safety management and control method for a pure electric unmanned ship according to claim 1, characterized in that: The fleet task graph includes: task nodes, ship nodes, edges, initial features of task nodes and initial features of ship nodes; a mapping relationship is established between task nodes and tasks, a mapping relationship is established between ship nodes and ships, the initial features of task nodes are represented by second task feature data of task nodes with which a mapping relationship is established, and the initial features of ship nodes are represented by ship feature data of ship nodes with which a mapping relationship is established; edge construction methods include: building edges between special tasks and designated ships corresponding to the tasks; and building connections between ordinary tasks and any ships.

6. A battery safety management and control method for a pure electric unmanned ship according to claim 1, characterized in that: The fitness prediction model includes: a first hidden layer, an extraction layer, a second hidden layer and a classifier; The first hidden layer includes N first units, the nth first unit inputs the initial feature of the node at the nth time point, and outputs the first updated feature, wherein N=T / t, 1≤n≤N; The extraction layer is used to extract the first updated feature output by the Nth first unit and use it as the initial feature of the node; The second hidden layer inputs the fleet mission graph and outputs a graph update matrix, where each row vector of the graph update matrix represents the second updated feature of a node; The second updated feature of each node is input into the classifier, and the classification space of the classifier represents the fitness between the ship and the task, and the above nodes include task nodes and ship nodes.

7. A battery safety management and control method for a pure electric unmanned ship according to claim 6, characterized in that: The first hidden layer is built based on the improved gated recurrent unit. The specific improvements include: introducing an adjustment coefficient in the update gate, adding an attenuation parameter in the reset gate, embedding a residual connection, and adding a trend control gate.

8. A battery safety management and control method for a pure electric unmanned ship according to claim 6, characterized in that: The calculation formula for the second hidden layer includes: ； in, represents the graph update matrix of the second hidden layer output, pile represents the stacking operation, represents the dynamic weighting coefficient, represents the degree matrix, and the elements of the degree matrix represent the number of edges connected to the node. represents the adjacency matrix, the elements of which represent the connection relationship between nodes, W represents the first weight coefficient of the second hidden layer, Represents the index influence coefficient, which is used to control the index term The influence of P represents the feature screening matrix, which is used to screen the features of enhanced task priority and the features of remaining power of the ship. Represents the edge weight adjustment coefficient, which is used to control the weight of the feature screening item. I represents the unit matrix. The element values ​​on the diagonal of the unit matrix are assigned 1, and the values ​​of the other elements are assigned 0. represents the second updated feature of the node, represents the absolute value of the element in the i-th row and j-th column of the initial feature matrix, ReLU represents the ReLU activation function, and the above-mentioned nodes include task nodes and ship nodes.

9. A battery safety management and control method for a pure electric unmanned ship according to claim 1, characterized in that: The steps of generating an allocation plan based on the fitness prediction model combined with the improved geyser optimization algorithm include: Step S301, randomly create a starting population with Q individuals that meet the constraint conditions, where Q is a custom parameter; The individuals of the initial population are encoded by the vector Indicates that Indicates that the 1st task to the Jth task are assigned to the To vessels; Step S302, constructing an objective function in combination with a fitness prediction model; Step S303, calculating the fitness value of each individual in the initial population according to the objective function; Step S304, using the roulette wheel selection method to select half of the individuals with higher fitness values ​​in the population as channels, and the selection probability is proportional to the fitness value of the individual; Step S305, calculating the Euclidean distance between each individual and other individuals, and selecting y individuals with the closest Euclidean distance as neighbors of the current individual; Step S306, updating the position of the individual according to the individual with the largest current fitness value and the individual's neighbors; Step S307, calculating the pressure on the individual and the probability of selecting a new channel, and updating the position of the individual when selecting a new channel in combination with the Cauchy distribution; Step S308, the current number of iterations is increased by 1, and if it is determined that the current number of iterations reaches the maximum number of iterations, the current individual code is output as the optimal allocation solution, otherwise, the process returns to step S303, where the maximum number of iterations is a custom parameter.

10. A battery safety management and control method for a pure electric unmanned ship according to claim 9, characterized in that: The objective function in step S302 is: ； Where F represents the individual fitness value, represents the i-th task and the ship assigned to it The degree of fit between It represents the remaining power of the e-th ship after removing the power required by the task assigned to the ship. and Represent the urgency weight coefficient and distance penalty coefficient respectively, Indicates the urgency of the task. Indicates the distance from the current position of the ship to the mission location.

Images