Unmanned ship hydrological data early warning method and system based on big data analysis

Abstract

The invention discloses an unmanned ship hydrological data early warning method and system based on big data analysis, and belongs to the technical field of intelligent monitoring. Comprising the following steps: acquiring historical hydrological big data, establishing a business state model, and constructing a state transition contribution degree evaluation model; acquiring real-time hydrological data acquired by the unmanned ship in the target water area; on the basis of the service state model, substituting the preprocessed real-time hydrological data into the state transition contribution degree evaluation model, and obtaining the contribution degree of the state transition contribution degree evaluation model for driving the current service state to transition to a higher-level service state; generating an integrated cooperation instruction according to the comparison result; and the unmanned ship executes the integrated cooperation instruction. According to the method, the effect of synchronously improving the early warning response efficiency and the communication resource utilization rate is achieved by establishing the service state transition model and the contribution degree evaluation mechanism, and the problems of response delay and resource waste caused by task decision stiffness and data value judgment lag of a traditional unmanned ship are solved.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology, and in particular to an unmanned vessel hydrological data early warning method and system based on big data analysis. Background Technology

[0002] With the deepening of water conservancy informatization, unmanned surface vessel (USV) technology has been widely applied in the field of hydrological monitoring. Traditional hydrological monitoring mainly relies on fixed monitoring stations and manual surveys, which have drawbacks such as limited coverage and slow response speed. USVs, with their advantages of flexibility, maneuverability, and low cost, provide a new technological means for water area monitoring.

[0003] The above-disclosed technical solutions have at least the following technical problems: In traditional methods, unmanned surface vessels (USVs) perform monitoring tasks along preset routes, making it impossible to dynamically adjust monitoring strategies based on real-time hydrological conditions. Furthermore, the indiscriminate data transmission leads to rigid task decision-making and delayed data value assessment, resulting in delayed early warning responses and wasted communication resources. To address these issues, this invention proposes a solution. Summary of the Invention

[0004] This application provides a method and system for early warning of hydrological data from unmanned surface vessels (USVs) based on big data analysis. This solves the problems of delayed early warning response and wasted communication resources caused by rigid decision-making and lagging data value judgment in the prior art. It achieves the effective results of dynamically adjusting monitoring strategies according to real-time hydrological conditions, prioritizing the transmission of key early warning data, and improving early warning efficiency and resource utilization.

[0005] This application provides an unmanned vessel hydrological data early warning method based on big data analysis. The method is applied to an unmanned vessel hydrological data early warning system based on big data analysis, including: acquiring historical hydrological big data, establishing a business state model, and presetting the triggering conditions for transitions between states, and constructing a state transition contribution evaluation model. Acquire real-time hydrological data collected by the unmanned vessel in the target water area, and preprocess the real-time hydrological data; Based on the business state model, the preprocessed real-time hydrological data is substituted into the state transition contribution evaluation model to obtain its contribution to driving the current business state to a higher level business state. The contribution level is compared with a transition threshold that is dynamically adjusted according to the current business status, and an integrated collaborative instruction is generated based on the comparison result. The unmanned vessel executes the integrated collaborative commands and submits key decision-making data to the command center.

[0006] Furthermore, the steps for acquiring historical hydrological big data and establishing an operational status model include: Historical hydrological big data is extracted from the hydrological database, including water level data, flow velocity data, turbidity data, and temperature data. Historical hydrological big data is preprocessed to obtain preprocessed historical hydrological big data. The preprocessed historical hydrological big data was classified using the K-means clustering algorithm, with four cluster centers set to correspond to stable state, alert state, warning state and emergency state. Define state boundaries for each cluster center and determine state affiliation by obtaining the Euclidean distance from each data point to the cluster center; Preset the trigger conditions for transitions between states; A business state model is constructed by mapping historical hydrological big data to these states and recording the transition paths, forming a state transition matrix.

[0007] Furthermore, the steps for constructing a state transition contribution evaluation model include: Extract the state transition matrix from the business state model. This matrix contains the transition probabilities between stable state, alert state, warning state and emergency state. Define a contribution calculation formula for each state; The state transition contribution evaluation model is trained by using logistic regression algorithm, with historical hydrological big data as the training set, multidimensional parameters as input features, and transition occurrence as output feature. To verify the accuracy of the model, historical hydrological big data was divided into training and validation subsets using cross-validation, and the accuracy index of the model on the validation subset was obtained. Adjust the model parameters until the accuracy index is not less than the preset accuracy threshold.

[0008] Furthermore, the steps to determine its contribution to driving the transition from the current business state to a higher-level business state include: Receive preprocessed real-time hydrological data and convert the preprocessed real-time hydrological data into a real-time state vector; Load the business status model and determine whether the current business status is one of stable, alert, warning, or emergency. Input the real-time state vector into the state transition contribution evaluation model to obtain the contribution of each component; The formula for calculating the contribution of a component is: ; In the formula, As for contribution level, The weight of the i-th parameter, Let i be the standardized value of the i-th parameter. Let be the threshold of the i-th parameter in the current state.

[0009] Furthermore, the steps for generating integrated collaborative instructions based on the comparison results include: Based on contribution and current business status, the transition threshold is dynamically adjusted according to the current business status. The contribution is compared with the adjusted transition threshold. If the contribution is less than the transition threshold, an instruction to maintain the current task mode is generated, and the data transmission strategy is set to low-frequency transmission of real-time state vector. If the contribution is not less than the transition threshold, an instruction to switch to a higher-level mission mode is generated, and the data transmission strategy is set to high-frequency transmission of real-time status vectors and additional alarm data, while simultaneously integrating the subsequent mission modes of the unmanned vessel. The data transmission strategy integrates real-time state vectors, encapsulates integrated collaborative instructions into structured data packets, and sends integrated collaborative instructions to the unmanned vessel.

[0010] Furthermore, the steps for defining state boundaries for each cluster center and determining state affiliation by obtaining the Euclidean distance from each data point to the cluster center include: Divide historical hydrological big data into K mutually exclusive subsets; Each subset is used as a validation subset in turn, and the remaining K-1 subsets are used as training subsets to train the state transition contribution evaluation model. For each trained state transition contribution evaluation model, its accuracy index on the validation subset is calculated. The accuracy index uses a weighted F1 score, and its calculation formula is as follows: ; In the formula, Let be the proportion of samples in the i-th state. Let be the accuracy for the i-th class of states. Let be the recall rate of the i-th class of states; Accuracy The calculation formula is: ; In the formula, Let be the number of true positive samples in the i-th state. Let be the number of false positive samples in the i-th class. Recall rate The calculation formula is: ; In the formula, Let be the number of false negative samples in the i-th class. The arithmetic mean of the F1 scores obtained from K validations is used as the final accuracy metric for the model.

[0011] Furthermore, the steps for dividing historical hydrological big data into training and validation subsets using cross-validation and obtaining the model's accuracy metrics on the validation subset include: Transmission priority is determined based on the ratio of contribution to the transition threshold, and the formula for calculating the transmission priority is as follows:

[0012] In the formula, For transmission priority, As for contribution level, The transition threshold, As a contribution weighting factor, For information value, As an information value weighting factor, Channel quality factor, Channel quality impact coefficient; Information value The calculation formula is: ; In the formula, The information entropy of the system state before transmission, The information entropy of the system state after transmission. This represents the maximum possible information entropy of the system. Channel quality factor The calculation formula is: ; In the formula, For bit error rate, Packet loss rate; According to transmission priority The numerical range is used to select the corresponding wireless communication protocol and data compression method. When the data exceeds the first priority threshold, it is determined to be urgent data, and the first communication protocol is used in conjunction with lossless compression. when Data that is not greater than the first priority threshold but greater than the second priority threshold is considered important and is compressed using the second communication protocol with lossy compression. when If the data is not greater than the second priority threshold, it is considered regular data and is transmitted using the third communication protocol in conjunction with data aggregation.

[0013] This application provides an unmanned surface vessel (USV) hydrological data early warning system based on big data analysis, which is used to implement an unmanned surface vessel hydrological data early warning method based on big data analysis, including: Model building module, data acquisition module, contribution acquisition module, instruction generation module, instruction execution module; The model building module is used to acquire historical hydrological big data, establish a business status model, preset the triggering conditions for transitions between states, and construct a state transition contribution evaluation model. The data acquisition module is used to acquire real-time hydrological data collected by the unmanned vessel in the target water area and to preprocess the real-time hydrological data. The contribution acquisition module is used to input the preprocessed real-time hydrological data into the state transition contribution evaluation model based on the business state model, and obtain its contribution to driving the current business state to a higher level business state. The instruction generation module is used to compare the contribution level with a transition threshold that is dynamically adjusted according to the current business status, and generate an integrated collaborative instruction based on the comparison result. The instruction execution module is used for the unmanned vessel to execute the integrated collaborative instructions and submit key decision-making data to the command center.

[0014] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: By introducing a business state model and a state transition mechanism, the unmanned surface vessel (USV) can dynamically adjust its monitoring tasks based on real-time hydrological data, solving the problem of rigid task decision-making in traditional systems and improving the flexibility and response speed of flood prevention early warning. Furthermore, during the operation of this state transition mechanism, the contribution evaluation model quantifies the data value in real time, automatically identifying key data and prioritizing its transmission, effectively solving the problem of delayed data value judgment and reducing the waste of communication bandwidth. Moreover, during the execution of the data transmission strategy, differentiated transmission frequencies and compression methods are adopted based on the magnitude of contribution, ensuring both real-time feedback of key information and optimizing the utilization efficiency of communication resources, thereby constructing an adaptive and highly efficient intelligent early warning closed-loop system. Through the above technical solutions, the intelligence level of unmanned surface vessel hydrological monitoring is significantly improved, ensuring timely and accurate decision-making in flood prevention early warning. Attached Figure Description

[0015] Figure 1 Flowchart of an unmanned vessel hydrological data early warning method based on big data analysis provided in this application embodiment; Figure 2 A schematic diagram of the structure of an unmanned vessel hydrological data early warning system based on big data analysis provided in this application embodiment. Detailed Implementation

[0016] This application provides a method and system for early warning of hydrological data from unmanned vessels based on big data analysis. This solves the problems of rigid decision-making and delayed data value judgment in the prior art for unmanned vessel missions. By establishing a business status model to evaluate the contribution of data and dynamically generating integrated instructions, it realizes the adaptive adjustment of monitoring tasks and the intelligent allocation of communication resources.

[0017] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0018] like Figure 1 As shown, this application provides an unmanned vessel hydrological data early warning method based on big data analysis. The method is applied to an unmanned vessel hydrological data early warning system based on big data analysis, including: acquiring historical hydrological big data, establishing a business state model including stable state, alert state, early warning state and emergency state, and presetting the triggering conditions for transitions between each state, and constructing a state transition contribution evaluation model. The system acquires real-time hydrological data collected by unmanned surface vessels in the target water area and preprocesses the real-time hydrological data. The preprocessing includes data cleaning, data alignment, data standardization, and vector construction, and the real-time hydrological data is processed into a real-time state vector containing multi-dimensional parameters such as water level, flow velocity, and turbidity. Based on the business state model, the preprocessed real-time hydrological data, i.e. the real-time state vector, is substituted into the state transition contribution evaluation model to obtain its contribution to driving the current business state to a higher level business state. Real-time hydrological data with a contribution of not less than a preset contribution threshold are listed as key data. The contribution is compared with the transition threshold that is dynamically adjusted according to the current business status, and an integrated collaborative instruction is generated based on the comparison result. This instruction simultaneously determines the subsequent task mode of the unmanned vessel and the data transmission strategy of the real-time state vector. The unmanned vessel executes the integrated collaborative instructions and submits the key decision-making data transmitted back through the data transmission strategy to the command center. The command center confirms or corrects the current business status and then feeds back the updated business status to the unmanned vessel, forming a closed loop of decision-making and execution.

[0019] Furthermore, the steps for acquiring historical hydrological big data and establishing an operational status model include: Historical hydrological big data is extracted from the hydrological database. The historical hydrological big data includes water level data, flow velocity data, turbidity data and temperature data. The water level data is acquired by an ultrasonic water level sensor, the flow velocity data is collected by an electromagnetic flow meter, the turbidity data is collected by an optical turbidity sensor, and the temperature data is collected by a thermistor temperature sensor. Historical hydrological big data is preprocessed to obtain preprocessed historical hydrological big data. The preprocessing includes cleaning operations, removing missing values ​​and outliers, processing missing values ​​by mean imputation and detecting outliers by Z-score method. The preprocessed historical hydrological big data was classified using the K-means clustering algorithm, with four cluster centers set to correspond to stable state, alert state, warning state and emergency state. The steps for classifying preprocessed historical hydrological big data are as follows: First, four initial cluster centers were randomly selected, each corresponding to a historical hydrological data point, to serve as the initial representatives of the stable state, alert state, early warning state, and emergency state. Secondly, each preprocessed historical hydrological data point is assigned to the nearest cluster center using the Euclidean distance formula, the mathematical expression of which is: ; In the formula, Let represent the Euclidean distance between the i-th hydrological data point and the j-th cluster center, and n represent the feature dimension of the hydrological data point (such as the number of features such as water level, flow rate, and rainfall). This represents the component of the i-th hydrological data point in the k-th feature dimension. This represents the component of the j-th cluster center on the k-th feature dimension; Then, update the coordinates of each cluster center. The coordinates of the new cluster center are the average of the feature dimension components of all data points in the cluster (that is, for each feature dimension, obtain the average of all data points in the cluster in that dimension, and use it as the component of the new center in that dimension). Repeat the assignment and update steps until the change in the cluster center position is less than the preset threshold of 0.001 (i.e., the change in each feature dimension component of all cluster centers is less than 0.001) or the maximum number of iterations of 100 is reached, and the algorithm converges; finally, the cluster to which each data point belongs corresponds to its classification status (stable, alert, warning, emergency).

[0020] Define state boundaries for each cluster center, and determine the state affiliation by obtaining the Euclidean distance from each data point to the cluster center; The triggering conditions for transitions between states are preset, and the specific triggering conditions for transitions between states are as follows: From stable to alert status: water level exceeds the historical average water level by 15%, or flow velocity exceeds the historical average flow velocity by 20%, or turbidity exceeds the historical average turbidity by 25%.

[0021] From alert status to warning status: water level exceeds the historical average water level by 25% and flow velocity exceeds the historical average flow velocity by 30%, or turbidity exceeds the historical average turbidity by 50% and flow velocity exceeds the historical average flow velocity by 25%.

[0022] From warning status to emergency status: water level exceeds the historical average water level by 50%, or flow velocity exceeds the historical average flow velocity by 80%, or turbidity exceeds the historical average turbidity by 100%.

[0023] A business state model is constructed by mapping historical hydrological big data to these states and recording the transition paths to form a state transition matrix. This matrix records the transition probability from each state to the next state. The steps to build a business state model are as follows: Using Markov chain modeling technology, we first extract state sequences from historical hydrological big data. These states refer to stable state (all parameters are within the normal range), alert state (a single parameter exceeds the threshold), warning state (multiple parameters exceed the threshold in a correlated manner), and emergency state (extreme danger mode). Secondly, obtain the number of transitions from each state to the next state, for example, count the number of times a state transitions from a stable state to a vigilant state. Then, form a state transition matrix, which is a 4x4 square matrix where each row represents the current state, each column represents the next state, and the elements are transition probabilities, obtained by dividing the number of transitions by the total number of times the current state is reached. For example, the mathematical expression for the matrix elements is: ; in, This represents the probability of transitioning from state i to state j. This represents the number of times a transition from state i to state j occurs. This represents the total number of times state i occurs; Finally, the row sum of the verification matrix is ​​1 (i.e., for any state i, we have...). This ensures probabilistic integrity and is used to predict future state transitions.

[0024] The business state model is stored in memory for later real-time data comparison. Throughout the process, data sampling is set to a fixed interval to maintain data consistency.

[0025] Furthermore, the steps for constructing a state transition contribution evaluation model include: Extract the state transition matrix from the business state model. This matrix contains the transition probabilities between stable state, alert state, warning state and emergency state. A contribution calculation formula is defined for each state, and the influence weights of multi-dimensional parameters such as water level, flow velocity, turbidity and temperature are integrated by weighted summation. The influence weights are obtained from historical hydrological big data through principal component analysis algorithm. The contribution calculation formula for each state is defined as follows: ; In the formula, As for contribution level, , , , The weights for water level, flow velocity, turbidity, and temperature are respectively. , , , These are the standardized water level, flow velocity, turbidity, and temperature values, respectively. The mathematical expressions for the contribution functions of each parameter are: Water level contribution function: ; Flow rate contribution function: ; Turbidity contribution function: ; Temperature contribution function: ; in, The threshold values ​​for each parameter under the corresponding state.

[0026] The state transition contribution evaluation model is trained by using logistic regression algorithm, with historical hydrological big data as the training set, multidimensional parameters as input features, and transition occurrence as output feature. The steps for training a state transition contribution evaluation model using the logistic regression algorithm are as follows: First, prepare the training dataset by using the multidimensional parameters of historical hydrological big data as the input feature vector and the occurrence of the transition (1 for occurrence, 0 for non-occurrence) as the output label. Next, initialize the model parameters, including setting the weight vector to zero and the bias to 0; then calculate the loss function (cross-entropy loss) using the gradient descent optimization algorithm, whose mathematical expression is: ; Where y is the true label and p is the sigmoid prediction probability; The mathematical expression for updating the weights is: ; in The gradient is the derivative of the loss with respect to the weights. Repeat the iteration until the loss change is less than 0.001 or the preset number of iterations of 1000 is reached; finally, save the optimized weights to predict the probability of transitions to new data.

[0027] Set the algorithm parameters, including the learning rate as a preset small value, the regularization strength as a preset coefficient, and the number of iterations as a preset integer value, and optimize the model convergence. To verify the accuracy of the model, historical hydrological big data was divided into training and validation subsets using cross-validation, and the accuracy index of the model on the validation subset was obtained. Adjust the model parameters until the accuracy index is not less than the preset accuracy threshold; Adjusting model parameters includes: Learning rate adjustment: A learning rate decay strategy is adopted, with the initial learning rate set to 0.1 and decayed to 0.9 times the original rate every 100 iterations.

[0028] Regularization strength adjustment: Using L2 regularization, the optimal regularization coefficient is found in the range [0.001,0.01,0.1,1] through grid search.

[0029] Determine the number of iterations: Set an early stopping mechanism, and stop training when the validation set loss no longer decreases after 10 consecutive iterations.

[0030] An integrated sensor data input interface ensures that data from ultrasonic water level sensors, electromagnetic flow meters, optical turbidity sensors, and thermistor temperature sensors can be directly input into the model. Store the completed state transition contribution evaluation model in the file system for real-time evaluation.

[0031] Furthermore, the steps to determine its contribution to driving the transition from the current business state to a higher-level business state include: The system receives pre-processed real-time hydrological data, which is collected at a fixed sampling frequency by ultrasonic level sensors, electromagnetic flow meters, optical turbidity sensors, and thermistor temperature sensors on the unmanned vessel. The pre-processed real-time hydrological data is converted into a real-time state vector, which includes water level, flow velocity, turbidity, and temperature components. The steps to convert preprocessed real-time hydrological data into real-time state vectors are as follows: The values ​​of four parameters—water level, flow rate, turbidity, and temperature—are extracted from the preprocessed data.

[0032] Using the same standardized parameters as historical data, the parameter values ​​are converted to the [0,1] range.

[0033] The standardized values ​​are assembled into a four-dimensional vector in a fixed order (water level, flow velocity, turbidity, temperature).

[0034] Load the business status model and determine whether the current business status is one of stable, alert, warning, or emergency. Input the real-time state vector into the state transition contribution evaluation model to obtain the contribution of each component; The formula for calculating the contribution of a component is: ; In the formula, Contribution, dimensionless. The weight of the i-th parameter is obtained through principal component analysis and satisfies... , Let i be the standardized value of the i-th parameter. Let be the threshold of the i-th parameter in the current state.

[0035] Furthermore, the steps for generating integrated collaborative instructions based on the comparison results include: Based on contribution and current business status, the transition threshold is dynamically adjusted according to the current business status. A first threshold is set for a stable state (low, such as 0.3, triggering a state transition when the contribution reaches 0.3), a second threshold is set for an alert state (medium, such as 0.5, triggering a state transition when the contribution reaches 0.5), and a third threshold is set for an early warning state (high, such as 0.7, triggering a state transition when the contribution reaches 0.7). The contribution is compared with the adjusted transition threshold. If the contribution is less than the transition threshold, an instruction to maintain the current task mode is generated, and the data transmission strategy is set to transmit the real-time state vector at a low frequency (e.g., once every ten minutes). If the contribution is not less than the transition threshold, an instruction to switch to a higher-level mission mode is generated, and the data transmission strategy is set to high frequency (e.g., data is transmitted once per minute) to transmit real-time status vectors and additional alarm data. At the same time, the subsequent mission modes of the unmanned vessel are integrated, including adjusting the navigation path, increasing sampling points, and activating emergency response. The data transmission strategy integrates real-time state vectors, including the selection of wireless communication protocols and data compression methods. It encapsulates integrated collaborative instructions into structured data packets containing task mode codes and transmission strategy parameters, and sends integrated collaborative instructions to the unmanned vessel. The entire generation process ensures the accuracy and timeliness of the instructions through conditional branching logic, thereby achieving precise control over the behavior of the unmanned vessel.

[0036] Among them, the wireless communication protocol can be selected from 4GLTE protocol (3GPPTS36.211), LoRaWAN protocol (LoRaAlliance), or NB-IoT protocol (3GPPTS36.201). Data compression methods can include GZIP compression algorithm (RFC1952), LZ77 compression algorithm, and differential coding compression; Furthermore, the steps for defining state boundaries for each cluster center and determining state affiliation by obtaining the Euclidean distance from each data point to the cluster center include: Historical hydrological big data is divided into K mutually exclusive subsets, and stratified sampling is used to ensure that the proportion of each state category in each mutually exclusive subset is consistent with the original dataset. Each subset is used as a validation subset in turn, and the remaining K-1 subsets are used as training subsets to train the state transition contribution evaluation model. For each trained state transition contribution evaluation model, its accuracy index on the validation subset is calculated. The accuracy index uses a weighted F1 score, and its calculation formula is as follows: ; In the formula, Let be the proportion of samples in the i-th state. Let be the accuracy for the i-th class of states. Let be the recall rate of the i-th class of states; Accuracy The calculation formula is: ; In the formula, Let be the number of true positive samples in the i-th class, that is, the number of samples that the model predicts are positive and whose true labels are positive. denoted as the number of false positive samples in the i-th class, i.e., the number of samples that the model predicts to be positive but whose true label is negative; Recall rate The calculation formula is: ; In the formula, is the number of false negative samples in the i-th class, that is, the number of samples that the model predicts as negative but the true label is positive; The arithmetic mean of the F1 scores obtained from K validations is used as the final accuracy index of the model. This final accuracy index is used to evaluate the contribution of state transitions to assess the overall performance of the model.

[0037] Furthermore, the steps to obtain the model's accuracy metrics on the validation subset include: Transmission priority is determined based on the ratio of contribution to the transition threshold, and the formula for calculating the transmission priority is as follows: ; In the formula, For transmission priority, As for contribution level, The transition threshold, As a contribution weighting factor, For information value, As an information value weighting factor, Channel quality factor, Channel quality impact coefficient; Among them, the channel quality impact coefficient The calculation method is as follows: ; SNR stands for Signal-to-Noise Ratio, which is measured at the physical layer and is measured in dB.

[0038] Information value The calculation formula is: ; In the formula, The information entropy of the system state before transmission, The information entropy of the system state after transmission. This represents the maximum possible information entropy of the system. Channel quality factor The calculation formula is: ; In the formula, For bit error rate, Packet loss rate; According to transmission priority The numerical range is used to select the corresponding wireless communication protocol and data compression method. When the data exceeds the first priority threshold, it is determined to be urgent data, and the first communication protocol is used in conjunction with lossless compression. when Data that is not greater than the first priority threshold but greater than the second priority threshold is considered important and is compressed using the second communication protocol with lossy compression. when If the data is not greater than the second priority threshold, it is considered regular data and is transmitted using the third communication protocol in conjunction with data aggregation.

[0039] like Figure 2 As shown, this application provides an unmanned vessel hydrological data early warning system based on big data analysis to implement the unmanned vessel hydrological data early warning method based on big data analysis, including: a model building module, a data acquisition module, a contribution acquisition module, an instruction generation module, and an instruction execution module; The model building module is used to acquire historical hydrological big data, establish a business status model, preset the triggering conditions for transitions between states, and construct a state transition contribution evaluation model. The data acquisition module is used to acquire real-time hydrological data collected by the unmanned vessel in the target water area and to preprocess the real-time hydrological data. The contribution acquisition module is used to input the preprocessed real-time hydrological data into the state transition contribution evaluation model based on the business state model, and obtain its contribution to driving the current business state to a higher level business state. The instruction generation module is used to compare the contribution level with a transition threshold that is dynamically adjusted according to the current business status, and generate an integrated collaborative instruction based on the comparison result. The instruction execution module is used for the unmanned vessel to execute the integrated collaborative instructions and submit key decision-making data to the command center.

[0040] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0041] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0042] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0043] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0044] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0045] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for early warning of hydrological data from unmanned vessels based on big data analysis, characterized in that: Includes the following steps: Acquire historical hydrological big data, establish a business status model, preset the triggering conditions for transitions between each status, and construct a status transition contribution evaluation model. Acquire real-time hydrological data collected by the unmanned vessel in the target water area, and preprocess the real-time hydrological data; Based on the business state model, the preprocessed real-time hydrological data is substituted into the state transition contribution evaluation model to obtain its contribution to driving the current business state to a higher level business state. The contribution level is compared with a transition threshold that is dynamically adjusted according to the current business status, and an integrated collaborative instruction is generated based on the comparison result. The unmanned vessel executes the integrated collaborative commands and submits key decision-making data to the command center.

2. The method for early warning of hydrological data from unmanned vessels based on big data analysis as described in claim 1, characterized in that, The steps to acquire historical hydrological big data and establish a business status model include: Historical hydrological big data is extracted from the hydrological database, including water level data, flow velocity data, turbidity data, and temperature data. Historical hydrological big data is preprocessed to obtain preprocessed historical hydrological big data. The preprocessed historical hydrological big data was classified using the K-means clustering algorithm, with four cluster centers set to correspond to stable state, alert state, warning state and emergency state. Define state boundaries for each cluster center and determine state affiliation by obtaining the Euclidean distance from each data point to the cluster center; Preset the trigger conditions for transitions between states; A business state model is constructed by mapping historical hydrological big data to these states and recording the transition paths, forming a state transition matrix.

3. The method for early warning of hydrological data from unmanned vessels based on big data analysis as described in claim 1, characterized in that, The steps to construct a state transition contribution evaluation model include: Extract the state transition matrix from the business state model. This matrix contains the transition probabilities between stable state, alert state, warning state and emergency state. Define a contribution calculation formula for each state; The state transition contribution evaluation model is trained by using logistic regression algorithm, with historical hydrological big data as the training set, multidimensional parameters as input features, and transition occurrence as output feature. To verify the accuracy of the model, historical hydrological big data was divided into training and validation subsets using cross-validation, and the accuracy index of the model on the validation subset was obtained. Adjust the model parameters until the accuracy index is not less than the preset accuracy threshold.

4. The method for early warning of hydrological data from unmanned vessels based on big data analysis as described in claim 1, characterized in that, The steps to determine its contribution to driving the transition from the current business state to a higher-level business state include: Receive preprocessed real-time hydrological data and convert the preprocessed real-time hydrological data into a real-time state vector; Load the business status model and determine whether the current business status is one of stable, alert, warning, or emergency. Input the real-time state vector into the state transition contribution evaluation model to obtain the contribution of each component; The formula for calculating the contribution of a component is: ； In the formula, As for contribution level, The weight of the i-th parameter, Let i be the standardized value of the i-th parameter. Let be the threshold of the i-th parameter in the current state.

5. The method for early warning of hydrological data from unmanned vessels based on big data analysis as described in claim 1, characterized in that, The steps for generating integrated collaborative instructions based on the comparison results include: Based on contribution and current business status, the transition threshold is dynamically adjusted according to the current business status. The contribution is compared with the adjusted transition threshold. If the contribution is less than the transition threshold, an instruction to maintain the current task mode is generated, and the data transmission strategy is set to low-frequency transmission of real-time state vector. If the contribution is not less than the transition threshold, an instruction to switch to a higher-level mission mode is generated, and the data transmission strategy is set to high-frequency transmission of real-time status vectors and additional alarm data, while simultaneously integrating the subsequent mission modes of the unmanned vessel. The data transmission strategy integrates real-time state vectors, encapsulates integrated collaborative instructions into structured data packets, and sends integrated collaborative instructions to the unmanned vessel.

6. The method for early warning of hydrological data from unmanned vessels based on big data analysis as described in claim 2, characterized in that, The steps for defining state boundaries for each cluster center and determining state affiliation by obtaining the Euclidean distance from each data point to the cluster center include: Divide historical hydrological big data into K mutually exclusive subsets; Each subset is used as a validation subset in turn, and the remaining K-1 subsets are used as training subsets to train the state transition contribution evaluation model. For each trained state transition contribution evaluation model, its accuracy index on the validation subset is calculated. The accuracy index uses a weighted F1 score, and its calculation formula is as follows: ； In the formula, Let be the proportion of samples in the i-th state. Let be the accuracy for the i-th class of states. Let be the recall rate of the i-th class of states; Accuracy The calculation formula is: ； In the formula, Let be the number of true positive samples in the i-th state. Let be the number of false positive samples in the i-th class. Recall rate The calculation formula is: ； In the formula, Let be the number of false negative samples in the i-th class. The arithmetic mean of the F1 scores obtained from K validations is used as the final accuracy metric for the model.

7. The method for early warning of hydrological data from unmanned vessels based on big data analysis as described in claim 1, characterized in that, The steps involved in dividing historical hydrological big data into training and validation subsets using cross-validation and obtaining the model's accuracy metrics on the validation subset include: Transmission priority is determined based on the ratio of contribution to the transition threshold, and the formula for calculating the transmission priority is as follows: ； In the formula, For transmission priority, As for contribution level, The transition threshold, As a contribution weighting factor, For information value, As an information value weighting factor, Channel quality factor, Channel quality impact coefficient; Information value The calculation formula is: ； In the formula, The information entropy of the system state before transmission, The information entropy of the system state after transmission. This represents the maximum possible information entropy of the system. Channel quality factor The calculation formula is: ； In the formula, For bit error rate, Packet loss rate; According to transmission priority The numerical range is used to select the corresponding wireless communication protocol and data compression method. When the data exceeds the first priority threshold, it is determined to be urgent data, and the first communication protocol is used in conjunction with lossless compression. when Data that is not greater than the first priority threshold but greater than the second priority threshold is considered important and is compressed using the second communication protocol with lossy compression. when If the data is not greater than the second priority threshold, it is considered regular data and is transmitted using the third communication protocol in conjunction with data aggregation.

8. A big data-based unmanned vessel hydrological data early warning system, used to implement the big data-based unmanned vessel hydrological data early warning method according to any one of claims 1-7, characterized in that, include: Model building module, data acquisition module, contribution acquisition module, instruction generation module, instruction execution module; The model building module is used to acquire historical hydrological big data, establish a business status model, preset the triggering conditions for transitions between states, and construct a state transition contribution evaluation model. The data acquisition module is used to acquire real-time hydrological data collected by the unmanned vessel in the target water area and to preprocess the real-time hydrological data. The contribution acquisition module is used to input the preprocessed real-time hydrological data into the state transition contribution evaluation model based on the business state model, and obtain its contribution to driving the current business state to a higher level business state. The instruction generation module is used to compare the contribution level with a transition threshold that is dynamically adjusted according to the current business status, and generate an integrated collaborative instruction based on the comparison result. The instruction execution module is used for the unmanned vessel to execute the integrated collaborative instructions and submit key decision-making data to the command center.

Images