A collaborative optimization method and system for river comprehensive management and control data based on the Internet of Things

Abstract

The present invention relates to the field of data processing technology, specifically a collaborative optimization method and system for river channel comprehensive management and control data based on the Internet of Things. First, multi-source heterogeneous data, analysis demand reports and river channel status assessment reports collected by Internet of Things sensors are obtained, and multi-source preprocessed data, analysis demand vectors and river channel status assessment vectors are obtained after preprocessing, which are input into a dynamic multi-source heterogeneous data fusion module based on an attention mechanism to obtain dynamic fusion data; then, the dynamic fusion data is input into a cross-domain associated river channel status collaborative prediction module to obtain a collaborative prediction result of the river channel status; finally, the dynamic fusion data and the collaborative prediction result are input into a multi-objective collaborative optimization module, and the target weights and decision priorities are dynamically adjusted according to the module input, and the multi-objective optimization algorithm is used to solve the problem, and the optimized decision parameters are obtained and optimized according to the optimized decision parameters; the present invention can effectively improve the management decision-making level of river channel data.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and specifically to a method and system for collaborative optimization of river comprehensive management and control data based on the Internet of Things. Background Art

[0002] With the development of the Internet of Things (IoT), it has become possible to deploy a large number of sensors in river management, providing a data foundation for obtaining comprehensive, real-time information on river status. However, how to effectively process and utilize this massive amount of multi-source, heterogeneous data to support scientific and efficient integrated river management decisions remains a major challenge facing current data processing technologies.

[0003] In the existing technology, the processing of river data often has the following deficiencies: on the one hand, for IoT river data with diverse sources, different formats and uneven quality, it is difficult for existing technology to dynamically adjust the importance of each data source according to the real-time changing river conditions or management priorities, resulting in insufficient relevance and effectiveness of the fusion results, affecting the accuracy of subsequent analysis and decision-making; on the other hand, the existing technology adopts a multi-objective optimization method based on fixed weights or rules, which is difficult to adaptively adjust the optimization objectives, constraints and priorities according to the dynamic risks and prediction results revealed by real-time data, and lacks accurate prediction of cross-domain impacts; therefore, it is urgent to propose a new data processing method that can dynamically fuse multi-source heterogeneous IoT river data, realize cross-domain collaborative prediction, and make adaptive multi-objective collaborative optimization decisions based on this. Summary of the Invention

[0004] The purpose of the present invention is to provide a collaborative optimization method and system for river channel comprehensive management and control data based on the Internet of Things. First, the multi-source heterogeneous data, analysis demand report and river channel status assessment report collected by the Internet of Things sensor are obtained, and after preprocessing, the multi-source preprocessed data, analysis demand vector and river channel status assessment vector are obtained, which are input into the dynamic multi-source heterogeneous data fusion module based on the attention mechanism to obtain dynamic fusion data; then, the dynamic fusion data is input into the cross-domain associated river channel status collaborative prediction module to obtain the collaborative prediction result of the river channel status; finally, the dynamic fusion data and the collaborative prediction result are input into the multi-objective collaborative optimization module, the target weight and decision priority are dynamically adjusted according to the module input, and the multi-objective optimization algorithm is used to solve the problem, the optimized decision parameters are obtained and the optimization is performed according to the optimized decision parameters; the present invention can effectively improve the management decision-making level of river channel data.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A collaborative optimization method for river integrated management and control data based on the Internet of Things, comprising:

[0007] Obtain multi-source heterogeneous data collected by IoT sensors, analyze demand reports and river status assessment reports;

[0008] Preprocessing the multi-source heterogeneous data, the analysis requirement report, and the river channel state assessment report to obtain multi-source preprocessed data, an analysis requirement vector, and a river channel state assessment vector;

[0009] Inputting the multi-source pre-processed data, the analysis requirement vector, and the river state assessment vector into a dynamic multi-source heterogeneous data fusion module based on an attention mechanism to generate dynamic fused data;

[0010] Inputting the dynamic fusion data into a cross-domain associated river channel state collaborative prediction module to obtain a collaborative prediction result of the river channel state;

[0011] Construct a multi-objective collaborative optimization module, input the dynamic fusion data and the collaborative prediction results into the multi-objective collaborative optimization module, dynamically adjust the target weights and decision priorities according to the dynamic fusion data and the collaborative prediction results, use the multi-objective optimization algorithm to solve, obtain the optimized decision parameters and optimize according to the optimized decision parameters.

[0012] Furthermore, the multi-source heterogeneous data includes: river hydrological status data, river water quality status data and river engineering status data.

[0013] Furthermore, the process of preprocessing the multi-source heterogeneous data, the analysis requirement report, and the river channel state assessment report to obtain the multi-source preprocessed data, the analysis requirement vector, and the river channel state assessment vector includes:

[0014] Performing data cleaning, format standardization, and spatiotemporal alignment on the multi-source heterogeneous data to obtain the multi-source preprocessed data;

[0015] The analysis requirement report and the river channel status assessment report are respectively subjected to text cleaning, text analysis and vectorization to obtain the analysis requirement vector and the river channel status assessment vector.

[0016] Furthermore, the multi-source pre-processed data, the analysis requirement vector, and the river state assessment vector are input into a dynamic multi-source heterogeneous data fusion module based on an attention mechanism to generate dynamic fused data, including:

[0017] Inputting the multi-source pre-processed data, the analysis requirement vector and the river channel state assessment vector into the data embedding layer of the dynamic multi-source heterogeneous data fusion module to obtain a multi-source embedding vector, an analysis requirement embedding vector and a river channel state assessment embedding vector;

[0018] Inputting the multi-source embedding vector, the analysis requirement embedding vector, and the river state assessment embedding vector into the attention weighting layer of the dynamic multi-source heterogeneous data fusion module to obtain a multi-source weighted feature vector;

[0019] Inputting the multi-source weighted feature vector, the analysis requirement embedding vector, and the river state assessment embedding vector into the comprehensive feature fusion layer of the dynamic multi-source heterogeneous data fusion module to obtain a fused feature vector;

[0020] The fused feature vector is input into the output layer of the dynamic multi-source heterogeneous data fusion module to obtain the dynamic fused data.

[0021] Furthermore, the process of inputting the dynamic fusion data into a cross-domain associated river channel state collaborative prediction module to obtain a collaborative prediction result of the river channel state includes:

[0022] Defining nodes and edges in a graph structure, and constructing node feature vectors based on the dynamic fusion data;

[0023] The river channel state collaborative prediction module is trained using historical dynamic fusion data, and graph data processing and river channel state prediction are performed according to the spatiotemporal graph convolution layer and prediction output layer in the river channel state collaborative prediction module, thereby obtaining a pre-trained river channel state collaborative prediction module;

[0024] The dynamic fusion data is input into the pre-trained river channel state collaborative prediction module to obtain a collaborative prediction result of the river channel state.

[0025] Furthermore, the dynamic fusion data and the collaborative prediction results are input into the multi-objective collaborative optimization module, the objective weights and decision priorities are dynamically adjusted according to the dynamic fusion data and the collaborative prediction results, and the multi-objective optimization algorithm is used to solve the problem to obtain the optimized decision parameters, which includes:

[0026] Defining the objective function and constraints of the multi-objective collaborative optimization module and determining the decision variables;

[0027] Inputting the dynamic fusion data and the collaborative prediction results into the multi-objective collaborative optimization module, and dynamically adjusting the objective weights and decision priorities according to the dynamic fusion data and the collaborative prediction results;

[0028] The adjusted target weights and decision priorities are combined, and a multi-objective optimization algorithm is used to solve the optimization problem to obtain the optimized decision parameters.

[0029] A river channel integrated management and control data collaborative optimization system based on the Internet of Things, comprising: a data acquisition unit, a data processing unit, a data dynamic fusion unit, a river channel state collaborative prediction unit, and a multi-objective collaborative optimization unit;

[0030] The data acquisition unit is used to obtain multi-source heterogeneous data collected by IoT sensors, analyze demand reports and river status assessment reports;

[0031] The data processing unit is used to preprocess the multi-source heterogeneous data, the analysis requirement report and the river state assessment report to obtain multi-source preprocessed data, an analysis requirement vector and a river state assessment vector;

[0032] The data dynamic fusion unit is used to input the multi-source pre-processed data, the analysis requirement vector and the river state assessment vector into a dynamic multi-source heterogeneous data fusion module based on an attention mechanism to generate dynamic fusion data;

[0033] The river channel state collaborative prediction unit is used to input the dynamic fusion data into the cross-domain associated river channel state collaborative prediction module to obtain a collaborative prediction result of the river channel state;

[0034] The multi-objective collaborative optimization unit is used to input the dynamic fusion data and the collaborative prediction results into the multi-objective collaborative optimization module, dynamically adjust the target weight and decision priority according to the dynamic fusion data and the collaborative prediction results, use the multi-objective optimization algorithm to solve, obtain the optimized decision parameters and optimize according to the optimized decision parameters.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1. The present invention proposes a dynamic fusion method for multi-source data for obtaining dynamic fusion data; this method inputs multi-source preprocessed data, analysis demand vectors and river status assessment vectors into a dynamic multi-source heterogeneous data fusion module, uses an attention weighting layer to dynamically assign feature weights, highlights key features related to analysis demand and status assessment, and uses a comprehensive feature fusion layer to comprehensively fuse the weighted multi-source data, analysis demand and river status assessment information; this method can adaptively adjust the data fusion method according to different analysis requirements and the ever-changing river status to meet actual needs, providing data support for subsequent river status prediction and decision optimization.

[0037] 2. The present invention proposes a collaborative prediction method for river channel status for obtaining river channel status prediction results; this method utilizes the spatiotemporal graph convolution layer in the collaborative prediction module of river channel status to simultaneously capture the spatial dependencies in the graph structure and the temporal dependencies in the time series, which enables the module to learn richer and more comprehensive river channel status feature representations; this method obtains more accurate river channel status prediction results by capturing and utilizing the inherent correlation between hydrological, water quality and engineering field data, which is conducive to the subsequent collaborative optimization of comprehensive management and control data.

[0038] 3. The present invention proposes a multi-objective collaborative optimization method for effectively adjusting river data decisions; this method combines dynamic fusion data, collaborative prediction results and multi-objective collaborative optimization modules, and uses dynamic fusion data and collaborative prediction results to dynamically adjust target weights and decision priorities to achieve more flexible optimization strategy adjustments; using a multi-objective optimization algorithm for solution, it can achieve balance and coordination among various objectives; this method realizes the intelligence and refinement of river comprehensive data decision-making, effectively improving the level of comprehensive river management. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Schematic diagram of a process of collaborative optimization method for river comprehensive management and control data based on the Internet of Things of the present invention;

[0040] Figure 2 Schematic diagram of the structure of the dynamic multi-source heterogeneous data fusion module based on the attention mechanism of the present invention;

[0041] Figure 3 This is a structural diagram of a river comprehensive management and control data collaborative optimization system based on the Internet of Things of the present invention. DETAILED DESCRIPTION

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0043] See also Figures 1 to 3 The present invention provides a method and system for collaborative optimization of river comprehensive management and control data based on the Internet of Things. The technical solution is as follows:

[0044] Example 1:

[0045] In order to improve the level of river comprehensive management, a certain enterprise used a river comprehensive management and control data collaborative optimization method based on the Internet of Things proposed in this invention. The process diagram of this method can be referred to Figure 1 , specifically including:

[0046] Obtain multi-source heterogeneous data collected by IoT sensors, analyze demand reports and river status assessment reports;

[0047] Furthermore, the multi-source heterogeneous data include: river hydrological status data, river water quality status data and river engineering status data;

[0048] Furthermore, river hydrological status data includes: water level, flow, flow velocity, rainfall, water temperature, etc.; river water quality status data includes: dissolved oxygen, biochemical oxygen demand, ammonia nitrogen, pH value, turbidity, etc.; river engineering status data mainly refers to the status data of engineering facilities such as dams, sluices, and reservoirs;

[0049] Furthermore, the analysis needs report includes: analysis topic, analysis objectives, analysis scope, analysis indicators, etc.; the river status assessment report includes: report title, assessment scope, assessment indicators, etc.

[0050] By using multi-source data from three different fields, namely hydrology, water quality and engineering, solid data support is provided for subsequent data fusion, state prediction and multi-objective collaborative optimization; at the same time, there is correlation between data from different fields and they influence each other, which provides an objective factual basis for interactive processing and collaborative analysis of data.

[0051] Preprocess multi-source heterogeneous data, analysis demand reports, and river status assessment reports to obtain multi-source preprocessed data, analysis demand vectors, and river status assessment vectors;

[0052] Furthermore, the multi-source heterogeneous data, analysis requirement report, and river status assessment report are preprocessed to obtain the multi-source preprocessed data, analysis requirement vector, and river status assessment vector, including the following steps:

[0053] Perform data cleaning, format standardization, and spatiotemporal alignment on multi-source heterogeneous data to obtain multi-source pre-processed data;

[0054] Perform text cleaning, text analysis, and vectorization on the analysis demand report and river status assessment report respectively to obtain the analysis demand vector and river status assessment vector;

[0055] Furthermore, the data cleaning process includes: missing value processing, outlier processing, and noise processing; format standardization includes unifying data types, units, and date formats; spatiotemporal alignment uses timestamp alignment to ensure that the timestamps of multi-source data are at the same moment, and spatial alignment is performed using geographic coordinate conversion and spatial matching operations;

[0056] Furthermore, the text cleaning process includes: removing useless characters, case conversion, and spelling correction; the text analysis process includes: word segmentation, part-of-speech tagging, named entity recognition, and keyword extraction; the vectorization process is to use the bag-of-words model to convert text data into vector form.

[0057] By cleaning, formatting, and spatiotemporally aligning multi-source heterogeneous data to ensure data consistency and comparability, the pre-processing vector results of the analysis demand report and river status assessment report can provide a demand basis for subsequent dynamic data fusion, thereby ensuring the effectiveness of river status prediction and data decision optimization.

[0058] Input multi-source pre-processed data, analysis demand vectors, and river status assessment vectors into a dynamic multi-source heterogeneous data fusion module based on an attention mechanism to generate dynamic fused data;

[0059] Furthermore, the structure of the dynamic multi-source heterogeneous data fusion module based on the attention mechanism is as follows: Figure 2 As shown, it includes: data embedding layer, attention weighting layer, comprehensive feature fusion layer and output layer;

[0060] Furthermore, the multi-source pre-processed data, analysis requirement vector, and river status assessment vector are input into the dynamic multi-source heterogeneous data fusion module based on the attention mechanism. The process of generating dynamic fused data includes:

[0061] Input the multi-source pre-processed data, analysis requirement vector and river channel status assessment vector into the data embedding layer of the dynamic multi-source heterogeneous data fusion module to obtain the multi-source embedding vector, analysis requirement embedding vector and river channel status assessment embedding vector;

[0062] The multi-source embedding vector, analysis requirement embedding vector, and river status assessment embedding vector are input into the attention weighting layer of the dynamic multi-source heterogeneous data fusion module to obtain the multi-source weighted feature vector;

[0063] Input the multi-source weighted feature vector, analysis requirement embedding vector and river status assessment embedding vector into the comprehensive feature fusion layer of the dynamic multi-source heterogeneous data fusion module to obtain the fused feature vector;

[0064] Input the fused feature vector into the output layer of the dynamic multi-source heterogeneous data fusion module to obtain dynamic fused data;

[0065] Furthermore, the data embedding layer processing process is as follows: using a recurrent neural network to process the multi-source pre-processed data to obtain a multi-source embedding vector; using a linear layer to process the analysis demand vector and the river state assessment vector respectively to obtain the analysis demand embedding vector and the river state assessment embedding vector;

[0066] Furthermore, the process of obtaining the multi-source weighted feature vector in the attention weighting layer is as follows: the channel-merged embedding vector of the analysis requirement embedding vector and the river state assessment embedding vector is used as the query vector, and the multi-source embedding vector is used as the key vector and value vector; the attention weight is obtained by performing a dot product operation on the query vector and the key vector and normalizing it with the softmax activation function; the attention weight is applied to the value vector to obtain the multi-source weighted feature vector;

[0067] Furthermore, the comprehensive feature fusion layer uses a weighted summation method to perform feature fusion; among them, the weights of the multi-source weighted feature vector come from the attention weights in the attention weighted layer; the weights of the analysis demand embedding vector and the river state assessment embedding vector are gated weights obtained through the linear layer and Sigmoid activation function.

[0068] By using a dynamic multi-source heterogeneous data fusion module based on the attention mechanism, it is possible to adaptively adjust data fusion to meet actual needs based on different analysis requirements and the ever-changing river status, combining attention and gating mechanisms, and provide data support for subsequent river status prediction and decision optimization.

[0069] The dynamic fusion data is input into the cross-domain associated river state collaborative prediction module to obtain the collaborative prediction results of the river state;

[0070] Furthermore, the dynamic fusion data is input into the cross-domain associated river state collaborative prediction module to obtain the collaborative prediction results of the river state. The process includes:

[0071] Define nodes and edges in the graph structure and construct node feature vectors based on dynamic fusion data;

[0072] The river state collaborative prediction module is trained using historical dynamic fusion data. The spatiotemporal graph convolution layer and prediction output layer in the river state collaborative prediction module perform graph data processing and river state prediction respectively, thus obtaining a pre-trained river state collaborative prediction module.

[0073] The dynamic fusion data is input into the pre-trained river state collaborative prediction module to obtain the collaborative prediction results of the river state;

[0074] Furthermore, nodes include hydrological monitoring stations, key water quality monitoring points, and engineering facilities. Edges represent the connections between nodes, including flow path / spatial adjacency, control relationships, and strong correlations. Edges representing flow path / spatial adjacency are directional, i.e., the direction of the flow. Edges representing strong correlations are derived based on domain prior knowledge and data analysis and are used to connect spatially non-adjacent nodes.

[0075] Furthermore, at each time step, for each node in the graph structure, its corresponding dynamic fusion data record is used as the node feature vector of the node at that moment;

[0076] Furthermore, the spatiotemporal graph convolutional layer in the river state collaborative prediction module includes a graph convolutional network (GCN) and a long short-term memory network (LSTM). The spatiotemporal graph convolutional layer first uses the GCN to aggregate the spatial information of neighboring nodes and update the node representations to reflect spatial dependencies. The updated node representation sequence is then processed by the LSTM to capture temporal dynamics.

[0077] Furthermore, after multi-layer spatiotemporal graph convolution processing, the module generates a final hidden representation for each node that is rich in spatiotemporal and cross-domain information; the prediction output layer is used to perform prediction processing on the final hidden representation to obtain the state prediction result; the prediction output layer uses a fully connected layer.

[0078] In order to verify the effectiveness of the collaborative prediction method for river channel status proposed in this invention, 500 sets of historical data were input into three different prediction schemes to test the effectiveness of collaborative prediction of river channel status. Scheme 1 is the prediction method proposed in this invention, i.e., dynamic fusion data + collaborative prediction module for river channel status. Scheme 2 replaces the dynamic fusion data with the data that can only be collected by each node. Scheme 3 replaces the spatiotemporal graph convolution layer in the collaborative prediction module for river channel status with a graph convolution network. The modules of each scheme are pre-trained, and the prediction results obtained by each scheme are manually proofread with the actual results. The test results of the effectiveness of collaborative prediction of river channel status are shown in Table 1.

[0079] Table 1. Test results of the effectiveness of collaborative prediction of river status

[0080]

[0081] As can be seen from Table 1, compared with other schemes, the state prediction scheme (Scheme 1) proposed in the present invention can obtain more accurate and reliable prediction results, which helps to improve the effectiveness and implementation efficiency of subsequent river data decision optimization, thereby improving the level of river data management and control.

[0082] By utilizing the spatiotemporal graph convolution layer in the river state collaborative prediction module, it is possible to simultaneously capture the spatial dependencies in the graph structure and the temporal dependencies in the time series, which enables the module to learn a richer and more comprehensive representation of river state features. By capturing and utilizing the inherent correlations between hydrological, water quality and engineering field data, the module can obtain more accurate river state prediction results, which is conducive to the subsequent collaborative optimization of comprehensive management and control data.

[0083] Construct a multi-objective collaborative optimization module, input the dynamic fusion data and collaborative prediction results into the multi-objective collaborative optimization module, dynamically adjust the target weights and decision priorities according to the dynamic fusion data and collaborative prediction results, use the multi-objective optimization algorithm to solve, obtain the optimized decision parameters and optimize according to the optimized decision parameters.

[0084] Furthermore, the dynamic fusion data and collaborative prediction results are input into the multi-objective collaborative optimization module. The objective weights and decision priorities are dynamically adjusted according to the dynamic fusion data and collaborative prediction results. The multi-objective optimization algorithm is used to solve the problem. The process of obtaining the optimized decision parameters includes:

[0085] Define the objective function and constraints of the multi-objective collaborative optimization module and determine the decision variables;

[0086] The dynamic fusion data and collaborative prediction results are input into the multi-objective collaborative optimization module, and the objective weights and decision priorities are dynamically adjusted according to the dynamic fusion data and collaborative prediction results;

[0087] Combining the adjusted target weights with decision priorities, and using a multi-objective optimization algorithm to solve the optimization problem, we can obtain the optimized decision parameters.

[0088] Furthermore, the objective function is divided into flood control safety objectives, ecological protection objectives, and economic benefit objectives. Flood control safety objectives include minimizing flood risk and reducing flooded areas; ecological protection objectives include maximizing dissolved oxygen indicators, minimizing ammonia nitrogen indicators, and maximizing species richness; and economic benefit objectives include maximizing water resource utilization efficiency, minimizing operating costs, and minimizing energy consumption of engineering facilities.

[0089] Furthermore, the constraints include: flood control safety threshold, minimum ecological flow requirements, water quality standards, etc.; the decision variables include: flood discharge volume, water storage capacity, desilting area, embankment reinforcement strength, sewage discharge volume, etc.;

[0090] The different objective functions in the multi-objective collaborative optimization module proposed in this embodiment affect each other. For example, optimizing the ecological protection goal will increase the operating cost, which will affect the optimization of the economic benefit goal. For this reason, a dynamic weight adjustment strategy is adopted to balance the influence between goals in the multi-objective collaborative optimization process.

[0091] Furthermore, the process of dynamically adjusting target weights and decision priorities based on dynamic fusion data and collaborative prediction results is as follows: inputting the dynamic fusion data and collaborative prediction results into a regression model based on weight prediction to obtain dynamic weights of flood control safety goals, ecological protection goals, and economic benefit goals; inputting the dynamic weights into a regression model based on priority prediction to obtain the priorities of different decision variables;

[0092] Furthermore, the regression model can adopt support vector machine, decision tree, random forest, gradient boosting decision tree, neural network, etc.

[0093] Furthermore, the multi-objective optimization algorithm adopts a multi-objective evolutionary algorithm.

[0094] Given a flood control analysis requirements report and a river status assessment report, Table 2 below compares the changes in decision parameters before and after using the multi-objective collaborative optimization module. The decision parameters include reservoir discharge, downstream gate storage capacity, sewage treatment plant discharge capacity limit, and embankment reinforcement frequency.

[0095] Table 2. Comparison of decision parameter changes

[0096]

[0097] By combining dynamic fusion data, collaborative prediction results and multi-objective collaborative optimization modules, dynamic fusion data and collaborative prediction results are used to dynamically adjust target weights and decision priorities to achieve more flexible optimization strategy adjustments; using multi-objective optimization algorithms for solving problems can achieve balance and coordination among various targets; and the intelligent and refined decision-making of comprehensive river data is realized, effectively improving the level of comprehensive river management.

[0098] This embodiment proposes a collaborative optimization method for river channel comprehensive management data based on the Internet of Things. First, the multi-source heterogeneous data, analysis demand report and river channel status assessment report collected by the Internet of Things sensors are obtained. After preprocessing, the multi-source preprocessed data, analysis demand vector and river channel status assessment vector are obtained, which are input into the dynamic multi-source heterogeneous data fusion module based on the attention mechanism to obtain dynamic fusion data; then, the dynamic fusion data is input into the cross-domain associated river channel status collaborative prediction module to obtain the collaborative prediction result of the river channel status; finally, the dynamic fusion data and the collaborative prediction result are input into the multi-objective collaborative optimization module, the target weight and decision priority are dynamically adjusted according to the module input, and the multi-objective optimization algorithm is used to solve the problem, the optimized decision parameters are obtained and the optimization is performed according to the optimized decision parameters; the present invention can effectively improve the management decision-making level of river channel data.

[0099] Example 2:

[0100] The present invention also proposes a river comprehensive management and control data collaborative optimization system based on the Internet of Things. The structure of the system is as follows: Figure 3 As shown, it includes: data acquisition unit, data processing unit, data dynamic fusion unit, river state collaborative prediction unit and multi-objective collaborative optimization unit;

[0101] The data acquisition unit is used to obtain multi-source heterogeneous data collected by IoT sensors, analyze demand reports and river status assessment reports;

[0102] The data processing unit is used to preprocess multi-source heterogeneous data, analysis demand reports and river status assessment reports to obtain multi-source preprocessed data, analysis demand vectors and river status assessment vectors;

[0103] The data dynamic fusion unit is used to input multi-source pre-processed data, analysis demand vectors and river status assessment vectors into the dynamic multi-source heterogeneous data fusion module based on the attention mechanism to generate dynamic fusion data;

[0104] Furthermore, the process of generating dynamic fusion data by the data dynamic fusion unit includes:

[0105] Input the multi-source pre-processed data, analysis requirement vector and river channel status assessment vector into the data embedding layer of the dynamic multi-source heterogeneous data fusion module to obtain the multi-source embedding vector, analysis requirement embedding vector and river channel status assessment embedding vector;

[0106] The multi-source embedding vector, analysis requirement embedding vector, and river status assessment embedding vector are input into the attention weighting layer of the dynamic multi-source heterogeneous data fusion module to obtain the multi-source weighted feature vector;

[0107] Input the multi-source weighted feature vector, analysis requirement embedding vector and river status assessment embedding vector into the comprehensive feature fusion layer of the dynamic multi-source heterogeneous data fusion module to obtain the fused feature vector;

[0108] The fused feature vector is input into the output layer of the dynamic multi-source heterogeneous data fusion module to obtain dynamic fused data.

[0109] To verify the effectiveness of the dynamic multi-source heterogeneous data fusion module, 400 sets of historical test data were randomly selected and tested using three test modules. The historical test data included historical multi-source pre-processed data, historical analysis demand vectors, and historical river state assessment vectors. Since dynamic fusion data cannot intuitively reflect differences, the pre-trained river state collaborative prediction module of the river state collaborative prediction unit was used for verification. The historical dynamic fusion data output by each module was input into the pre-trained river state collaborative prediction module to obtain the historical state prediction results and compare them with the true values ​​to determine the proportion of prediction results within a reasonable range.

[0110] The test modules are: the dynamic multi-source heterogeneous data fusion module proposed in this invention, denoted as test module one; removing the attention weighting layer in the dynamic multi-source heterogeneous data fusion module and directly feeding the output vector of the data embedding layer into the comprehensive feature fusion layer for fusion, denoted as test module two; removing the comprehensive feature fusion layer in the dynamic multi-source heterogeneous data fusion module and directly feeding the multi-source weighted feature vector output by the attention weighting layer into the output layer, denoted as test module three;

[0111] The effectiveness test results of the dynamic multi-source heterogeneous data fusion module can be found in Table 3.

[0112] Table 3. Effectiveness test results of dynamic multi-source heterogeneous data fusion module

[0113]

[0114] As can be seen from Table 3, the module validity test results obtained by using test module 1, that is, the dynamic multi-source heterogeneous data fusion module proposed in the present invention, are better than those obtained by using other modules, indicating that the module output data proposed in the present invention helps to improve the accuracy of river status prediction, thereby further improving the level of river data management and control.

[0115] The river channel state collaborative prediction unit is used to input the dynamic fusion data into the cross-domain associated river channel state collaborative prediction module to obtain the collaborative prediction results of the river channel state;

[0116] Furthermore, the process of the river channel state collaborative prediction unit obtaining the collaborative prediction result of the river channel state is as follows:

[0117] Define nodes and edges in the graph structure and construct node feature vectors based on dynamic fusion data;

[0118] The river state collaborative prediction module is trained using historical dynamic fusion data. The spatiotemporal graph convolution layer and prediction output layer in the river state collaborative prediction module perform graph data processing and river state prediction respectively, thus obtaining a pre-trained river state collaborative prediction module.

[0119] The dynamic fusion data is input into the pre-trained river state collaborative prediction module to obtain the collaborative prediction results of the river state.

[0120] The multi-objective collaborative optimization unit is used to input dynamic fusion data and collaborative prediction results into the multi-objective collaborative optimization module, dynamically adjust the target weights and decision priorities according to the dynamic fusion data and collaborative prediction results, use the multi-objective optimization algorithm to solve, obtain the optimized decision parameters and optimize according to the optimized decision parameters.

[0121] Furthermore, the process of obtaining the optimization decision parameters by the multi-objective collaborative optimization unit includes:

[0122] Define the objective function and constraints of the multi-objective collaborative optimization module and determine the decision variables;

[0123] The dynamic fusion data and collaborative prediction results are input into the multi-objective collaborative optimization module, and the objective weights and decision priorities are dynamically adjusted according to the dynamic fusion data and collaborative prediction results;

[0124] The adjusted objective weights and decision priorities are combined, and the multi-objective optimization algorithm is used to solve the optimization problem to obtain the optimized decision parameters.

[0125] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A collaborative optimization method for river comprehensive management and control data based on the Internet of Things, characterized by: include: Obtain multi-source heterogeneous data collected by IoT sensors, analyze demand reports and river status assessment reports; Preprocess multi-source heterogeneous data, analysis demand reports, and river status assessment reports to obtain multi-source preprocessed data, analysis demand vectors, and river status assessment vectors; The multi-source pre-processed data, analysis requirement vector, and river channel status assessment vector are input into a dynamic multi-source heterogeneous data fusion module based on the attention mechanism to generate dynamic fused data. The module obtains the multi-source weighted feature vector by: using the channel-merged embedding vector of the analysis requirement embedding vector and the river channel status assessment embedding vector as the query vector, and the multi-source embedding vector as the key vector and value vector; using the dot product operation of the query vector and the key vector and the normalization processing of the softmax activation function to obtain the attention weight; applying the attention weight to the value vector to obtain the multi-source weighted feature vector; The dynamic fusion data is input into the cross-domain associated river state collaborative prediction module to obtain the collaborative prediction results of the river state; Construct a multi-objective collaborative optimization module, input the dynamic fusion data and collaborative prediction results into the multi-objective collaborative optimization module, dynamically adjust the target weights and decision priorities according to the dynamic fusion data and collaborative prediction results, use the multi-objective optimization algorithm to solve, obtain the optimized decision parameters and optimize according to the optimized decision parameters; the process is: define the objective function and constraints of the multi-objective collaborative optimization module, determine the decision variables; input the dynamic fusion data and collaborative prediction results into the multi-objective collaborative optimization module, dynamically adjust the target weights and decision priorities according to the dynamic fusion data and collaborative prediction results; combine the adjusted target weights and decision priorities, and use the multi-objective optimization algorithm to solve the optimization problem to obtain the optimized decision parameters.

2. The collaborative optimization method for river comprehensive management and control data based on the Internet of Things according to claim 1 is characterized in that: The multi-source heterogeneous data includes: river hydrological status data, river water quality status data and river engineering status data.

3. The collaborative optimization method for river comprehensive management and control data based on the Internet of Things according to claim 1 is characterized in that: The process of preprocessing the multi-source heterogeneous data, the analysis requirement report, and the river channel state assessment report to obtain the multi-source preprocessed data, the analysis requirement vector, and the river channel state assessment vector includes: Performing data cleaning, format standardization, and spatiotemporal alignment on the multi-source heterogeneous data to obtain the multi-source preprocessed data; The analysis requirement report and the river channel status assessment report are respectively subjected to text cleaning, text analysis and vectorization to obtain the analysis requirement vector and the river channel status assessment vector.

4. The collaborative optimization method for river comprehensive management and control data based on the Internet of Things according to claim 1 is characterized in that: The multi-source pre-processed data, the analysis requirement vector, and the river state assessment vector are input into a dynamic multi-source heterogeneous data fusion module based on an attention mechanism to generate dynamic fused data. The process includes: Inputting the multi-source pre-processed data, the analysis requirement vector and the river channel state assessment vector into the data embedding layer of the dynamic multi-source heterogeneous data fusion module to obtain a multi-source embedding vector, an analysis requirement embedding vector and a river channel state assessment embedding vector; Inputting the multi-source embedding vector, the analysis requirement embedding vector, and the river state assessment embedding vector into the attention weighting layer of the dynamic multi-source heterogeneous data fusion module to obtain a multi-source weighted feature vector; Inputting the multi-source weighted feature vector, the analysis requirement embedding vector, and the river state assessment embedding vector into the comprehensive feature fusion layer of the dynamic multi-source heterogeneous data fusion module to obtain a fused feature vector; The fused feature vector is input into the output layer of the dynamic multi-source heterogeneous data fusion module to obtain the dynamic fused data.

5. The collaborative optimization method for river comprehensive management and control data based on the Internet of Things according to claim 1 is characterized in that: The process of inputting the dynamic fusion data into the cross-domain associated river state collaborative prediction module to obtain the collaborative prediction result of the river state includes: Defining nodes and edges in a graph structure, and constructing node feature vectors based on the dynamic fusion data; The river channel state collaborative prediction module is trained using historical dynamic fusion data, and graph data processing and river channel state prediction are performed according to the spatiotemporal graph convolution layer and prediction output layer in the river channel state collaborative prediction module, thereby obtaining a pre-trained river channel state collaborative prediction module; The dynamic fusion data is input into the pre-trained river channel state collaborative prediction module to obtain a collaborative prediction result of the river channel state.

6. A river comprehensive management and control data collaborative optimization system based on the Internet of Things, characterized by: include: Data acquisition unit, data processing unit, data dynamic fusion unit, river state collaborative prediction unit and multi-objective collaborative optimization unit; The data acquisition unit is used to obtain multi-source heterogeneous data, analysis demand reports and river status assessment reports collected by Internet of Things sensors; the data processing unit is used to preprocess the multi-source heterogeneous data, the analysis demand report and the river status assessment report to obtain multi-source preprocessed data, analysis demand vector and river status assessment vector; the data dynamic fusion unit is used to input the multi-source preprocessed data, the analysis demand vector and the river status assessment vector into the dynamic multi-source heterogeneous data fusion module based on the attention mechanism to generate dynamic fusion data; the process of the dynamic multi-source heterogeneous data fusion module to obtain the multi-source weighted feature vector is as follows: the channel merging embedding vector of the analysis demand embedding vector and the river status assessment embedding vector is used as the query vector, and the multi-source embedding vector is used as the key vector and the value vector; the attention weight is obtained by using the dot product operation of the query vector and the key vector and the normalization processing of the softmax activation function; the attention weight is applied to the value vector to obtain the multi-source weighted feature vector. Weighted feature vector; the river state collaborative prediction unit is used to input the dynamic fusion data into the cross-domain associated river state collaborative prediction module to obtain the collaborative prediction result of the river state; the multi-objective collaborative optimization unit is used to input the dynamic fusion data and the collaborative prediction result into the multi-objective collaborative optimization module, dynamically adjust the target weight and decision priority according to the dynamic fusion data and the collaborative prediction result, use the multi-objective optimization algorithm to solve, obtain the optimized decision parameters and optimize according to the optimized decision parameters; the process of the multi-objective collaborative optimization unit to obtain the optimized decision parameters is: define the objective function and constraint conditions of the multi-objective collaborative optimization module, and determine the decision variables; input the dynamic fusion data and the collaborative prediction result into the multi-objective collaborative optimization module, and dynamically adjust the target weight and decision priority according to the dynamic fusion data and the collaborative prediction result; combine the adjusted target weight and decision priority, and use the multi-objective optimization algorithm to solve the optimization problem to obtain the optimized decision parameters.

Images