Preplan fuzzy search method and device based on pipe network node feature vectorization

Abstract

The invention relates to the technical field of plan searching, and discloses a plan fuzzy searching method and device based on pipe network node feature vectorization. When a new rainfall condition occurs, a first original vector corresponding to a new first rainfall condition is generated, and the first original vector is adjusted through a plan vector model; and the information processing difficulty and the dependence on professionals are reduced. Furthermore, the pre-constructed preset pipe network rainfall plan vector library can efficiently express various plan conditions as a vector with a fixed dimension, and the vector has the characteristic of visual display result, so that the use threshold of the preset pipe network rainfall plan library is greatly reduced. Therefore, searching and matching are carried out in the preset pipe network rainfall plan vector library according to the generated plan vector, the plan most suitable for the current rainfall condition can be screened out from a large number of plans in a short time, and the response speed for dealing with the sudden rainfall event is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of emergency plan search, and in particular to an emergency plan fuzzy search method and device based on pipeline network node feature vector quantization. Background Art

[0002] In current hydraulic calculations, while traditional numerical methods can produce relatively accurate results, they require stringent data input and involve complex computational processes, resulting in lengthy calculation times. This makes it difficult to respond quickly to actual rainfall events. To more effectively address these emergencies, establishing a contingency plan library is crucial. This involves screening similar plans from the library based on the initial conditions of the drainage network calculations, and rapidly formulating emergency measures based on the calculated results.

[0003] At present, the general process of generating a plan library includes three steps: constructing a drainage network simulation model, generating drainage plans under different rainfall scenarios, and storing the plans to form a plan library. However, the presentation of the plan library generated in this way is too professional. In terms of plan selection, it mainly relies on manual experience. First, manually formulate plan selection rules or matching algorithms, and locate problems by comprehensively analyzing information such as the drainage network structure, historical operation data, and potential correlations. In actual applications, all feasible plans are compared one by one according to the priority of the rules to determine the most matching plan. Second, the system obtains the attribute information of each plan and stores it in the plan library together with the calculation results. The attribute information also needs to be manually set and matched based on professional knowledge. In actual use, the attribute information of the on-site situation is searched in the plan library to find a matching plan.

[0004] While existing technologies have achieved some optimization results in emergency plan selection, they still face numerous limitations in practical application. For example, urban drainage networks are vast and feature-rich, requiring the plan selection process to process vast amounts of information, which is extremely complex. This not only results in a significant time and resource consumption in locating the problem, severely impacting emergency response efficiency and making it difficult to meet timeliness requirements, but also requires a high level of professional knowledge and a strict skill set for staff, further increasing labor costs.

[0005] Therefore, constrained by the above factors, existing methods are unable to efficiently cope with dynamic, changing, and complex rainfall scenarios, and cannot meet the real-time and intelligent requirements in complex scenarios. Summary of the Invention

[0006] In view of this, the present invention provides a fuzzy plan search method and device based on the quantization of pipeline node feature vectors to solve the problems of existing plan search methods that rely too much on manual experience and have a low response speed to sudden rainfall events, which makes it difficult to efficiently respond to dynamic, changeable and complex rainfall scenes, and cannot meet the real-time and intelligence requirements in complex scenes.

[0007] In a first aspect, the present invention provides a fuzzy search method for emergency plans based on pipeline network node feature vectorization, the method comprising:

[0008] Obtain a real-time drainage network node attribute feature set and an embedding vector of a target node in a target area to be searched under a first rainfall condition, wherein the target node is a drainage network node corresponding to the first rainfall condition in the target area to be searched, and one drainage network node corresponds to one embedding vector; generate a first original vector corresponding to the first rainfall condition based on the embedding vector and the real-time drainage network node attribute feature set; input the first original vector into a plan vector model to obtain a first plan vector corresponding to the first rainfall condition; determine a target network rainfall plan vector in a preset network rainfall plan vector library based on the first plan vector, the preset network rainfall plan vector library including multiple plan vectors of multiple rainfall plans under different rainfall conditions; determine a target network rainfall plan in the preset network rainfall plan library based on the target network rainfall plan vector, the preset network rainfall plan library including multiple rainfall plans under different rainfall conditions.

[0009] The present invention provides a fuzzy search method for emergency plans based on pipeline network node feature vectorization. When a new rainfall condition occurs, the method combines the real-time drainage pipeline network node feature set and the corresponding embedded vector of the target node under the new first rainfall condition to generate a first original vector corresponding to the new first rainfall condition. Furthermore, the first original vector is adjusted using a plan vector model, so that the generated first plan vector more accurately reflects the emergency plan direction to be taken under the new first rainfall condition, reducing the difficulty of information processing and the reliance on professional personnel. Furthermore, the pre-built preset pipeline network rainfall plan vector library can efficiently express various emergency plan situations as a fixed-dimensional vector. At the same time, the intuitive display of vectors greatly reduces the threshold for using the preset pipeline network rainfall plan library. Therefore, by searching and matching the generated plan vector within the preset pipeline network rainfall plan vector library, the plan vector that best suits the current rainfall conditions can be quickly selected from a large number of plans. The corresponding target pipeline network rainfall plan can then be determined in the corresponding preset pipeline network rainfall plan library, significantly reducing calculation time and greatly improving the response speed to sudden rainfall events. Therefore, by implementing the present invention, it is possible to efficiently cope with dynamic, changeable and complex rainfall scenes, thereby meeting the requirements for real-time and intelligence in complex scenes.

[0010] In an optional embodiment, generating a first original vector corresponding to the first rainfall condition based on the embedded vector and the real-time drainage network node attribute feature set includes:

[0011] The attribute feature of each node in the real-time drainage network node attribute feature set is vectorized to obtain the attribute feature vector of the target node; the embedded vector and the attribute feature vector are concatenated to generate the first original vector corresponding to the first rainfall condition.

[0012] The fuzzy plan search method based on the quantization of pipeline node feature vectors provided by the present invention comprehensively integrates the physical, environmental and other characteristics of the node itself, avoids the limitations of single feature expression, and enables the generated plan vector to more richly and accurately describe the status of the corresponding plan, providing support for subsequent improvement of the quality of plan generation and search.

[0013] In an optional embodiment, the method further includes:

[0014] A historical pipeline rainfall plan library, an undirected drainage pipeline network topology map, and multiple historical drainage pipeline network node attribute feature sets for the target area to be searched under multiple second rainfall conditions are obtained. The multiple historical drainage pipeline network node attribute feature sets are used to characterize the attribute characteristics of multiple drainage pipeline network nodes in the target area to be searched under multiple second rainfall conditions. Each drainage pipeline network node in the undirected drainage pipeline network topology map is vectorized to obtain multiple embedded vectors of the multiple drainage pipeline network nodes. Based on the multiple embedded vectors and the multiple historical drainage pipeline network node attribute feature sets, multiple second original vectors corresponding to the multiple second rainfall conditions are generated. Based on the multiple second original vectors, a plan vector model is constructed.

[0015] In an optional embodiment, constructing a plan vector model based on the plurality of second original vectors includes:

[0016] Based on the preset weights, a weighted calculation is performed on multiple second original vectors to obtain multiple second plan vectors corresponding to multiple second rainfall conditions; a loss function is constructed based on the historical pipeline network rainfall plan library and multiple second plan vectors; the preset weights are optimized using the gradient descent method until the loss function value of the loss function meets the requirements, thereby obtaining the target weights; based on the target weights, a weighted calculation is performed on multiple second original vectors to obtain multiple target pipeline network rainfall plan vectors corresponding to multiple second rainfall conditions; a plan vector model is constructed using multiple second original vectors as input data and multiple target pipeline network rainfall plan vectors as output data.

[0017] The proposed fuzzy plan search method, based on the quantization of network node feature vectors, constructs a loss function and optimizes weights using gradient descent. This allows the weights of the various features in the second original vector corresponding to the second rainfall condition to be optimized and adjusted based on a library of historical drainage network plans. This allows the generated target plan vector to more accurately reflect actual conditions, improving the consistency between the plan vector and the actual drainage network operation. Furthermore, the plan vector model, constructed using the original vector as input and the target plan vector as output, can better learn the patterns in historical data. Consequently, when faced with new rainfall conditions, the model can more accurately generate corresponding plan vectors, providing strong support for efficient target plan search.

[0018] In an optional embodiment, a loss function is constructed based on a historical pipeline network rainfall plan library and a plurality of second plan vectors, including:

[0019] According to the historical pipeline network rainfall plan library, a plan target distance function is constructed; according to multiple second plan vectors, a plan vector actual distance function is constructed; according to the plan target distance function and the plan vector actual distance function, a loss function is constructed.

[0020] The fuzzy plan search method based on the quantization of pipeline node feature vectors provided by the present invention accurately measures the difference between the second plan vector and the actual situation by separately constructing a plan target distance function and a plan vector actual distance function, and combining the two to construct a loss function. This provides accurate guidance for subsequent weight optimization using the gradient descent method, allowing the adjustment of weights to more specifically reduce the deviation between the second plan vector and the actual situation, thereby improving the quality and accuracy of the generated target plan vector. Furthermore, because the loss function is the difference between the initial plan vector and the actual situation, after a new rainfall situation occurs, even if there are no similar rainfall conditions in the preset pipeline network rainfall plan library, an ideal plan can be found in the preset pipeline network rainfall plan library.

[0021] In an optional embodiment, the method further includes: generating a preset pipeline network rainfall plan vector library based on the plan vector model.

[0022] The fuzzy search method for plans based on the quantization of pipeline node feature vectors provided by the present invention generates a preset pipeline network rainfall plan vector library according to the plan vector model, and can convert the plan into a fixed-dimensional vector, providing support for the subsequent rapid search for plan vectors similar to the current situation in new rainfall situations.

[0023] In an optional embodiment, the preset pipeline network rainfall plan library includes:

[0024] Obtain the drainage network topology structure; generate an undirected topology graph of the drainage network based on the drainage network topology structure.

[0025] The fuzzy search method for emergency plans based on the vectorization of network node features provided by the present invention effectively eliminates the interference caused by direction by converting the acquired drainage network topology structure into an undirected topology graph, thereby enabling nodes that originally had large differences in the directed structure due to different directions to show similarities in the vector space of the undirected graph, which helps to vectorize the nodes more accurately.

[0026] In a second aspect, the present invention provides a fuzzy search device for a plan based on the vectorization of pipeline network node features, the device comprising:

[0027] An acquisition module is used to obtain a real-time drainage network node attribute feature set and an embedding vector of a target node in a target area to be searched under a first rainfall condition, wherein the target node is a drainage network node corresponding to the first rainfall condition in the target area to be searched, and one drainage network node corresponds to one embedding vector; a generation module is used to generate a first original vector corresponding to the first rainfall condition based on the embedding vector and the real-time drainage network node attribute feature set; an input module is used to input the first original vector into a plan vector model to obtain a first plan vector corresponding to the first rainfall condition; a first determination module is used to determine a target network rainfall plan vector in a preset network rainfall plan vector library based on the first plan vector, the preset network rainfall plan vector library including multiple plan vectors of multiple rainfall plans under different rainfall conditions; a second determination module is used to determine a target network rainfall plan in a preset network rainfall plan library based on the target network rainfall plan vector, the preset network rainfall plan library including multiple rainfall plans under different rainfall conditions.

[0028] In a third aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, computer instructions being stored in the memory, and the processor executing the computer instructions to thereby execute the fuzzy search method for plans based on the quantization of pipeline network node features according to the first aspect or any corresponding embodiment thereof.

[0029] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the fuzzy search method for plans based on the quantization of pipeline node feature vectors according to the above-mentioned first aspect or any corresponding embodiment thereof.

[0030] In a fifth aspect, the present invention provides a computer program product comprising computer instructions for enabling a computer to execute the fuzzy search method for a plan based on the quantization of pipeline network node features according to the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0032] Figure 1 1 is a flow chart of a method for fuzzy search of emergency plans based on quantization of pipeline network node features according to an embodiment of the present invention;

[0033] Figure 2 1 is a flow chart of another method for fuzzy search of emergency plans based on quantization of pipeline network node features according to an embodiment of the present invention;

[0034] Figure 3 1 is a flow chart of another method for fuzzy search of emergency plans based on quantization of pipeline network node features according to an embodiment of the present invention;

[0035] Figure 4 is a topological diagram of a rainwater pipe network according to an embodiment of the present invention;

[0036] Figure 5 2. It is a structural block diagram of a fuzzy search device for emergency plans based on pipeline network node feature vectorization according to an embodiment of the present invention;

[0037] Figure 6 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are 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 those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0039] Existing emergency plan search methods rely too much on manual experience and have a low response speed to sudden rainfall events. In addition, in urban drainage networks, the same calculation results of monitoring nodes may come from different initial conditions. For example, rainfall at point A and rainfall at point B may have the same impact on monitoring node C. However, existing emergency plan library technologies mostly match the initial calculation conditions of actual conditions with the calculation results one by one. This means that when the spatial structures of points A and B are significantly different, existing technologies cannot accurately match the plan at point A with the initial calculation conditions at point B, greatly reducing the coverage of the emergency plan library. Therefore, constrained by the above factors, existing methods are difficult to efficiently respond to dynamic, changing, and complex rainfall scenarios, and cannot meet the real-time and intelligent requirements in complex scenarios.

[0040] According to an embodiment of the present invention, an embodiment of a fuzzy search method for a plan based on the quantization of pipeline network node feature vectors is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0041] In this embodiment, a fuzzy search method for emergency plans based on the quantization of pipeline network node features is provided, which can be used in electronic devices such as computers, mobile phones, tablet computers, etc. Figure 1 is a flow chart of a fuzzy search method for a plan based on the quantization of pipeline network node features according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0042] Step S101: obtaining a real-time drainage network node attribute feature set and an embedding vector of a target node in a target area to be searched under a first rainfall condition.

[0043] The real-time drainage network node attribute feature set may include rainfall information and pipe segment and node attributes that affect the drainage speed of the network, such as slope, width, pipe segment length, etc.

[0044] Furthermore, the target node is a drainage network node corresponding to the first rainfall condition in the target area to be searched, and there may be one or more nodes. In this embodiment, one target node is used as an example for description.

[0045] Furthermore, the embedding vector represents the representation of the drainage network node topology structure mapped to the vector space, that is, one drainage network node corresponds to one embedding vector.

[0046] Step S102: Generate a first original vector corresponding to a first rainfall condition based on the embedded vector and the real-time drainage network node attribute feature set.

[0047] Specifically, by combining the embedded vector and the real-time drainage network node attribute feature set, the current first rainfall condition can be vectorized, and then the first original vector can be generated that integrates the node's topological structure information and various attribute feature information, comprehensively describing the situation of the target node under the first rainfall condition.

[0048] Step S103: input the first original vector into the plan vector model to obtain a first plan vector corresponding to the first rainfall condition.

[0049] The plan vector model represents a trained model that can convert the input original vector into a meaningful plan vector.

[0050] Specifically, the first original vector of the first rainfall condition can be input into the plan vector model. Further, the plan vector model can process and calculate the first original vector corresponding to the first rainfall condition based on its learned patterns and rules and output a first plan vector corresponding to the first rainfall condition.

[0051] Furthermore, in some optional embodiments, when the new rainfall condition involves multiple target nodes, each target node and the first rainfall condition will generate a corresponding plan vector. Furthermore, the multiple plan vectors can be processed into a final first plan vector corresponding to the first rainfall condition by vector averaging.

[0052] Step S104: determining a target pipe network rainfall plan vector in a preset pipe network rainfall plan vector library according to the first plan vector.

[0053] The preset pipeline network rainfall plan vector library is pre-built and may include multiple plan vectors for multiple rainfall plans under different rainfall conditions. A drainage pipeline network node corresponds to different rainfall plans under different rainfall conditions, and each rainfall plan corresponds to one plan vector.

[0054] Specifically, each plan vector is taken out from the preset pipeline network rainfall plan vector library in turn, and its similarity with the first plan vector corresponding to the first rainfall condition is calculated. Then, based on the similarity calculation result, the plan vector with the highest similarity is selected as the target pipeline network rainfall plan vector.

[0055] Step S105 : determining the target pipe network rainfall plan in a preset pipe network rainfall plan library according to the target pipe network rainfall plan vector.

[0056] The preset pipeline network rainfall plan library may include multiple rainfall plans under different rainfall conditions, and each rainfall plan corresponds one-to-one to each plan vector in the preset pipeline network rainfall plan vector library.

[0057] Specifically, the corresponding target pipe network rainfall plan can be found in a preset pipe network rainfall plan library according to the determined target pipe network rainfall plan vector.

[0058] The fuzzy search method for emergency plans based on pipeline network node feature vectorization provided in this embodiment generates a first original vector corresponding to the new first rainfall condition when a new rainfall condition occurs, combining the real-time drainage pipeline network node feature set and the corresponding embedded vector of the target node under the new first rainfall condition. Furthermore, the first original vector is adjusted using the emergency plan vector model, so that the generated first emergency plan vector more accurately reflects the emergency plan direction to be taken under the new first rainfall condition, reducing the difficulty of information processing and the reliance on professional personnel. Furthermore, the pre-built preset pipeline network rainfall emergency plan vector library can efficiently express various emergency plan situations as a fixed-dimensional vector. At the same time, the intuitive nature of vectors greatly reduces the threshold for using the preset pipeline network rainfall emergency plan library. Therefore, by searching and matching the preset pipeline network rainfall emergency plan vector library based on the generated emergency plan vector, the emergency plan vector that best suits the current rainfall conditions can be quickly selected from a large number of emergency plans. The corresponding target pipeline network rainfall emergency plan can then be determined in the corresponding preset pipeline network rainfall emergency plan library, significantly reducing calculation time and greatly improving the response speed to sudden rainfall events. Therefore, by implementing the present invention, it is possible to efficiently cope with dynamic, changeable and complex rainfall scenes, thereby meeting the requirements for real-time and intelligence in complex scenes.

[0059] In this embodiment, a fuzzy search method for emergency plans based on the quantization of pipeline network node features is provided, which can be used in electronic devices such as computers, mobile phones, tablet computers, etc. Figure 2 is a flow chart of a fuzzy search method for a plan based on the quantization of pipeline network node features according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0060] Step S201: Obtain the real-time drainage network node attribute feature set and embedding vector of the target node in the target area to be searched under the first rainfall condition. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0061] Step S202: Generate a first original vector corresponding to a first rainfall condition based on the embedded vector and the real-time drainage network node attribute feature set.

[0062] Specifically, the above step S202 includes:

[0063] Step S2021 : performing vectorization processing on the attribute feature of each node in the real-time drainage network node attribute feature set to obtain an attribute feature vector of the target node.

[0064] The number and selection of attribute feature vectors can be adjusted as needed, and more detailed attribute feature vectors can help improve model accuracy.

[0065] Specifically, the rainfall information of each node in the drainage network node feature set is time series data, which is redundant if used directly. Therefore, the time series principal component analysis method is first used to reduce the dimension of the rainfall information.

[0066] Furthermore, different rainfall feature vectors can be generated according to different rainfall scenarios.

[0067] Furthermore, for the time it takes for water flow from nodes in the drainage network node feature set to reach the monitoring node, since the drainage network relies on gravity flow for drainage, the water flow time is first calculated using a formula based on the slope and length of the pipe section, and then the depth-first search (DFS) algorithm is used to calculate the time from each node to the monitoring node. Finally, a set of one-dimensional vectors is obtained after normalization.

[0068] Finally, the vectors obtained above are integrated to form the attribute feature vector of the target node.

[0069] Step S2022: Concatenate the embedded vector and the attribute feature vector to generate a first original vector corresponding to the first rainfall condition.

[0070] Specifically, the embedding vector and attribute feature vector of the target node are spliced ​​in a certain order. Usually, the elements of the embedding vector and attribute feature vector are connected in sequence to form a new vector, that is, the original vector V corresponding to the first rainfall condition. j ={v j1 ,v j2 …v jm}. Where m represents the number of splicing vectors.

[0071] Step S203: Input the first original vector into the plan vector model to obtain the first plan vector corresponding to the first rainfall condition. Figure 1 Step S103 of the illustrated embodiment will not be described in detail here.

[0072] Step S204: Determine the target network rainfall plan vector from the preset network rainfall plan vector library based on the first plan vector. Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.

[0073] Step S205: Determine the target pipe network rainfall plan in the preset pipe network rainfall plan library according to the target pipe network rainfall plan vector. Figure 1 Step S105 of the illustrated embodiment will not be described in detail here.

[0074] The fuzzy plan search method based on the quantization of pipeline node feature vectors provided in this embodiment avoids the limitations of single feature expression by comprehensively integrating the physical, environmental and other characteristics of the node itself, so that the generated plan vector can more richly and accurately describe the status of the corresponding plan, providing support for subsequent improvement of the quality of plan generation and search.

[0075] In some optional implementations, before the above step S203, the plan vector model may be constructed by the following steps:

[0076] Step a1: Obtain a historical drainage network rainfall plan library, an undirected drainage network topology graph, and multiple historical drainage network node attribute feature sets for the target area under multiple second rainfall conditions.

[0077] Among them, the historical pipe network rainfall plan library can include multiple historical rainfall plans for multiple second rainfall conditions; multiple historical drainage pipe network node attribute feature sets are used to characterize the attribute characteristics of multiple drainage pipe network nodes in the target area to be searched under multiple second rainfall conditions.

[0078] Furthermore, the drainage network undirected topological graph G(N, E) is used to represent the topological structure of the drainage network. Where N represents the node and E represents the edge between the nodes. It can be obtained by the following steps:

[0079] Step a11: Obtain the drainage network topology.

[0080] Step a12: Generate an undirected topology graph of the drainage network according to the topological structure of the drainage network.

[0081] Specifically, the drainage network is generally regarded as a directed tree structure, but the direction is an interference factor in the vectorization process of the network nodes. Therefore, in this embodiment, the obtained drainage network topology structure is converted into an undirected graph so that nodes in different directions can also show similarity in the vector space.

[0082] Step a2: performing vectorization processing on each drainage network node in the undirected topological graph of the drainage network to obtain multiple embedding vectors of the multiple drainage network nodes.

[0083] Specifically, the node2vec algorithm can be used to vectorize each drainage network node in the undirected topological graph G(N, E). The node2vec algorithm is a graph embedding algorithm based on random walks that can generate vector representations of nodes in the graph structure, capturing both local and global structural information of the nodes.

[0084] First, set the parameters for the node2vec algorithm. For example, define the length of the random walk, which is the number of nodes traversed during each random walk; set the return parameter and the in-out parameter. The return parameter determines the probability of returning to the previous node, and the in-out parameter determines whether the random walk tends to explore a local neighborhood or a broader graph structure.

[0085] Secondly, starting from each drainage network in the undirected topological graph of the drainage network, a random walk is performed according to the set parameters. During each random walk, the node sequence passed is recorded.

[0086] Finally, the random walk sequences of all drainage network nodes are processed and mapped into fixed-dimensional vectors using machine learning or deep learning methods (such as neural networks). These vectors are then used to map each drainage network node into an embedding vector. Furthermore, these embedding vectors contain information about the node's location within the overall network topology and its connections to other drainage network nodes, reflecting the topological characteristics of the drainage network node.

[0087] Step a3: generating a plurality of second original vectors corresponding to the second rainfall condition based on the plurality of embedded vectors and the plurality of drainage network node attribute feature sets.

[0088] Among them, multiple second original vectors can be expressed as: V1, V2, ... V n .

[0089] The specific process can be referred to the description of step S202 above, which will not be repeated here.

[0090] Step a4: constructing a plan vector model based on the multiple second original vectors.

[0091] Specifically, the above step a4 includes:

[0092] Step a41 : performing weighted calculation on the plurality of second original vectors based on preset weights to obtain a plurality of second plan vectors corresponding to the plurality of second rainfall conditions.

[0093] The distance of the second original vectors obtained by simple splicing in the vector space cannot represent the relationship between different plans. In order to better characterize the relationship between plans, the second original vectors need to be weighted.

[0094] Specifically, a set of weights W = {w1, w2, ... w m}, then, for each second original vector V1, V2, ... V, multiply the values ​​of each dimension by the corresponding preset weight value, and then add the multiplication results to obtain a new value. Furthermore, the vector composed of the obtained new values ​​is the multiple second plan vectors V′1, V′2, ... V′ corresponding to the multiple second rainfall conditions. n .in,

[0095] Step a42: constructing a loss function based on the historical pipe network rainfall plan library and multiple second plan vectors.

[0096] Among them, the loss function is used to measure the difference between the planned target distance and the actual distance.

[0097] In some optional implementations, the above step a42 includes:

[0098] Step a421: construct a plan target distance function based on the historical drainage network plan library.

[0099] Step a422: constructing a plan vector actual distance function based on the plurality of second plan vectors.

[0100] Step a423: construct a loss function based on the plan target distance function and the plan vector actual distance function.

[0101] Among them, the plan target distance function is used to reflect the distance measurement corresponding to the similarity of monitoring results of different plans under ideal conditions; the plan vector actual distance function is used to measure the actual distance between different plan vectors in the vector space.

[0102] Specifically, the calculation results of the monitoring nodes (i.e., the plans corresponding to the drainage network nodes) can be obtained from the historical drainage network plan library. Based on these results, the cosine similarity is used to calculate the similarity between plans. For example, if there are n plans, an n×n dimensional similarity matrix can be generated. And since the similarity between the plan itself and itself is 1, the similarity matrix The elements on the diagonal of are 1.

[0103] Furthermore, the similarity matrix It represents the similarity of the monitoring results of plan i and plan j, that is, the matrix actually reflects the target distance relationship between plans, and the corresponding plan target distance function can be further constructed.

[0104] Furthermore, each weighted second plan vector V can be calculated by using distance measurement methods such as Euclidean distance and Manhattan distance. i ′ and V j′ The actual distance between

[0105] For example, when the Euclidean distance calculation is selected,

[0106] Finally, construct the loss function

[0107] Furthermore, since the loss function is the difference between the second plan vector and the actual situation, even if there is no similar rainfall condition in the preset pipe network rainfall plan library in the subsequent plan search, the ideal plan can be found in the preset pipe network rainfall plan library.

[0108] By constructing the plan target distance function and the plan vector actual distance function respectively, and combining the two to construct the loss function, the difference between the second plan vector and the actual situation can be accurately measured, which provides accurate guidance for the subsequent optimization of weights using the gradient descent method, so that the adjustment of weights can more specifically reduce the deviation between the second plan vector and the actual situation, thereby improving the quality and accuracy of the generated target plan vector.

[0109] Step a43, using the gradient descent method to optimize the preset weights until the loss function value of the loss function meets the requirements and obtains the target weights.

[0110] Among them, the gradient descent method represents an iterative optimization algorithm used to solve unconstrained optimization problems. Its core purpose is to find the minimum value (or approximate minimum value) of the objective function by continuously adjusting parameters.

[0111] Specifically, set the relevant parameters of the gradient descent method, such as the learning rate and the maximum number of iterations. Then, use the gradient descent method to set the preset weights W = {w1, w2, ... w m}Perform iterative optimization.

[0112] Furthermore, in each iteration, the gradient of the loss function L with respect to the preset weight W can be calculated, and then the preset weight can be updated according to the direction and magnitude of the gradient, so that the weight is adjusted in a direction that can reduce the value of the loss function.

[0113] Finally, after each iteration, the loss function is checked to see if it is the smallest, that is, if it meets the preset requirements. If so, the iteration stops, and the weight obtained at this time is the target weight; otherwise, the next round of iteration continues until the termination condition is met.

[0114] By constructing a loss function and optimizing the weights using the gradient descent method, the weights of each feature in each second original vector can be optimized and adjusted according to the historical drainage network plan library, so that the generated target plan vector can more accurately reflect the actual situation and improve the fit between the plan vector and the actual operation status of the drainage network.

[0115] Step a44: performing weighted calculation on the plurality of second original vectors based on the target weights to obtain a plurality of target pipe network rainfall plan vectors corresponding to the plurality of second rainfall conditions.

[0116] Specifically, based on the optimized target weights, the weighted calculation process in step a41 is repeated to obtain the final target plan vector for each target network rainfall plan. Through optimization, the target plan vector can more accurately reflect the status of the drainage network plan.

[0117] Step a45 : constructing a plan vector model using the plurality of second original vectors as input data and the plurality of target pipe network rainfall plan vectors as output data.

[0118] Specifically, the obtained multiple second original vectors can be used as input data, and the corresponding multiple target pipeline network rainfall plan vectors as output data, and then input into a pre-set model architecture for training. During the training process, the model learns the mapping relationship between the original vectors and the target plan vectors and adjusts its internal parameters so that the model can accurately output the corresponding target pipeline network rainfall plan vector based on the input original vectors.

[0119] Among them, the pre-set model architecture can be a machine learning or deep learning model architecture, such as a multi-layer perceptron (MLP), a recurrent neural network (RNN) and its variants (LSTM, GRU), etc.

[0120] Furthermore, the plan vector model constructed through the training process can better learn the patterns in historical data, so that when faced with new rainfall conditions in the future, the model can more accurately generate corresponding plan vectors, providing strong support for efficient search for target plans.

[0121] The fuzzy plan search method based on the quantization of network node features provided in this embodiment accurately measures the difference between the second plan vector and the actual situation by separately constructing a plan target distance function and a plan vector actual distance function, and combining the two to form a loss function. Furthermore, by optimizing weights using a gradient descent method, the weights of each feature component in the original vector of the drainage network node can be optimized and adjusted based on the historical drainage network plan library. This allows the generated target plan vector to more accurately reflect the actual situation, improving the consistency between the plan vector and the actual drainage network operation status. Furthermore, the plan vector model constructed with the original vector as input and the target plan vector as output can better learn the patterns in historical data. Consequently, when faced with new rainfall conditions, the model can more accurately generate corresponding plan vectors, providing strong support for efficient target plan search. Furthermore, because the loss function is the difference between the initial plan vector and the actual situation, even if similar rainfall conditions are not available in the preset pipeline network rainfall plan library, an ideal plan can be found in the preset pipeline network rainfall plan library after a new rainfall condition occurs.

[0122] In some optional implementations, after step S203 and before step S204, the following steps may be further included:

[0123] Step a7: Generate a preset pipeline network rainfall plan vector library based on the plan vector model.

[0124] Specifically, by continuously collecting original vector data of different rainfall conditions and inputting them into the plan vector model, a corresponding preset pipe network rainfall plan vector library can be continuously generated.

[0125] The fuzzy search method for plans based on the quantization of pipeline node feature vectors provided in this embodiment generates a preset pipeline network rainfall plan vector library according to the plan vector model, and can convert the plan into a fixed-dimensional vector, providing support for quickly finding plan vectors similar to the current situation in new rainfall situations.

[0126] In one example, Figure 3 As shown in the figure, a fuzzy search method for emergency plans based on the vectorization of pipeline network node features is provided. By vectorizing the characteristics of drainage network nodes, a pipeline network rainfall emergency plan vector library is constructed. The node characteristics include the pipeline network topology and node attributes such as rainfall information and length information. The core of the method is to introduce vectorization technology to vectorize the initial calculation conditions in the emergency plan. The size and direction of the vector are continuously adjusted according to the calculation results of the emergency plan, thus transforming complex physical problem solving into simple vector operations.

[0127] Furthermore, when encountering new rainfall inputs, the system no longer relies on traditional segmented hydraulic solutions or manual rule matching. Instead, it directly vectorizes the initial calculation conditions based on the trained vectorized model. Using vector matching technology, it quickly finds plan vectors similar to the current conditions, thereby obtaining the estimated output of the monitoring node, significantly reducing calculation time. Furthermore, because the vectorized model is trained based on the calculation results of the plan, the distances between different initial calculation conditions in vector space can be close due to similar calculation results, enhancing the system's flexibility in responding to different rainfall events.

[0128] like Figure 4 The figure shows the actual topology of the rainwater pipe network in a certain area, in which the circle represents the inspection well node of the pipe network, the triangle represents the outlet of the pipe network, and the direction of the arrow indicates the upstream and downstream relationship of the drainage pipe network. The drainage pipe network mainly relies on gravity flow for drainage, and its topological structure is a typical tree structure feature. In the technical solution of the present invention, the main feature is to vectorize the plan to realize the fuzzy search function. This requires preparing a certain amount of plans as a training data set. It is assumed that the plan library contains multiple calculated plan models. Each plan represents the impact of a manhole node on the monitoring node when encountering rainfall. The present invention constructs a pipe network rainfall plan vector library by vectorizing these plans. It mainly includes the following steps:

[0129] S1. Generate embedding vectors of pipeline network nodes.

[0130] S101. Obtain the pipe network topology.

[0131] First, the network topology is obtained, forming an undirected graph G(N, E), where N represents a node and E represents an edge between nodes. Although drainage networks are generally viewed as directed trees, the direction of the network nodes is a factor in the vectorization process. In this example, it is converted into an undirected graph, ensuring that nodes with different orientations can be similar in the vector space.

[0132] S102. Node vectorization.

[0133] Node2vec is used to vectorize each node in the undirected graph G(N, E). A node sequence is generated through random walks, and the parameters are adjusted to make it more biased towards depth-first search, thereby capturing the local and global information of the node in the topological structure, and finally obtaining the embedding vector of each node.

[0134] S2. Generate attribute feature vectors for each node.

[0135] Attribute feature vectors are divided into two main categories: one is the pipe segment and node attributes that affect the drainage rate of the pipeline network, such as slope, width, and pipe segment length; the other is rainfall information. Each model in the plan library records the impact of each node on the monitoring results under different rainfall conditions. These rainfall conditions need to be vectorized as part of the node attribute feature vector. The number and selection of attribute feature vectors can be adjusted according to needs. Detailed attribute feature vectors can improve model accuracy.

[0136] S3. Concatenate the embedding vector and the attribute feature vector to form a plan vector.

[0137] S301. Splicing vectors.

[0138] The embedding vector of each plan node is spliced ​​with the attribute feature vector to form the original vector of the plan. Assume that there are n plans in the plan library, and each plan calculates the data of the monitoring node under different rainfall conditions for a node. The n plans in the plan library are spliced ​​into n original vectors according to the embedding vector and attribute feature vector of the node. The original vector of plan i is V i ={v i1 ,v i2 …v im}, n plans in the plan library form a set of original vectors V1, V2, ... V n .

[0139] S302. Define weight W = {w1, w2, ... w m}.

[0140] The distance of the original vectors obtained by simple concatenation in the vector space cannot represent the relationship between different plans. In order to better represent the relationship between plans, it is necessary to weight the original vectors. Define a set of weights W = {w1, w2, ... w m}, through weighted calculation, adjust the position of each plan original vector in the vector space to obtain the final plan vector V′1, V′2,…V′ n ,in

[0141] S303. Define a distance metric.

[0142] Define the target distance of the plan vector and actual distance

[0143] The distance between the plan vectors in the vector space should correspond to the similarity of the calculation results of the monitoring nodes. Plan vectors with similar calculation results are closer in the vector space. Define the target distance Indicates the similarity of monitoring results between plan i and plan j. Define the actual distance Represents the plan vector Vi ′ and the plan vector V j ′ The distance between them.

[0144] S304. Define the loss function and optimize the weights.

[0145] Define the loss function This function is used to measure the difference between the target distance and the actual distance. By minimizing this loss function, the weight W is adjusted to ensure that plans with similar calculation results are closer to each other in the vector space, thus optimizing the accuracy and efficiency of plan matching.

[0146] Furthermore, based on the above example, a specific embodiment is provided.

[0147] According to Figure 4 The stormwater pipe network structure shown in this example contains 32 nodes. Three different rainfall scenarios are set for each node and run in the SWMM model to record the calculation results of the monitoring nodes to form a plan library. At this time, the plan library contains 32×3=96 plans. Next, we will build a plan vector library based on the plan library. The specific implementation steps are as follows:

[0148] S1. Obtain an undirected graph and vectorize the nodes.

[0149] according to Figure 4 The stormwater network topology shown in Figure 1 is first represented as an undirected graph G(N, E), where N represents a node and E represents a connection between nodes. The node2vec algorithm is then used to vectorize each node in the undirected graph G(N, E), setting the embedding vector for each node to 5-dimensional. This results in a set of 5-dimensional node embedding vectors. The vector values ​​for "J119" and "J120" are shown below, as shown in Table 1:

[0150] Table 1. Node embedding vector values

[0151] Node number Node embedding vector J119 (-4.2967，-8.9209，-7.3558，-0.4359，0.4373) J120 (-4.5254，-9.0370，-6.3816，0.9147，0.8478)

[0152] S2. Obtain attribute feature vector.

[0153] Attribute feature vector 1 represents rainfall information for each node in the plan library. Rainfall information is a time series data. Directly treating rainfall data as a vector would result in redundancy. Therefore, time-series principal component analysis (PCA) is used to reduce the dimensionality of the rainfall data. The number of principal components is determined to be two, based on the criterion that the cumulative variance contribution of the principal components is greater than 90%. Three rainfall vectors are generated for three different rainfall scenarios. Attribute feature vector 2 represents the time it takes for water flow from each node to reach the monitoring node. Because the drainage network relies on gravity flow, the flow time is calculated using a formula based on the slope and length of the pipe segment. A depth-first search (DFS) algorithm is then used to calculate the flow time from each node to the monitoring node. After normalization, a set of one-dimensional vectors is obtained. The values ​​of attribute feature vector 1 are based on rainfall. The selected rainfall sequences are "2-yr," "10-yr," and "100-yr." The values ​​of attribute feature vector 2 are shown in Table 2 below, using "J119" and "J120" as examples.

[0154] Table 2. Attribute feature vector values

[0155]

[0156]

[0157] S3. Generate a plan vector.

[0158] S301. Splicing vectors.

[0159] The node embedding vector and the attribute feature vector are concatenated to form the original vector. The 96 plans in the plan library are thus converted into 96 original vectors V = {V1, V2, ... V q6}, each original vector contains the embedding vector and feature vector of a node, such as V i ={v i1 ,v i2 ,v i3 Taking "J119" and "J120" as examples, the original vectors are displayed. After adding rainfall to each node, 2×3=6 original vectors are spliced ​​together, as shown in Table 3 below:

[0160] Table 3. Original vectors

[0161] Nodes and rainfall Original vector (eigenvector 1, eigenvector 2, embedding vector) J119+2-yr (-4.8492，0.1382，0.5684，-4.2967，-8.9209，-7.3558，-0.4359，0.4373) J119+10-yr (-1.8575，-0.1865，0.5684,-4.2967，-8.9209，-7.3558，-0.4359，0.4373) J119+100-yr (6.7066，0.0483，0.5684，-4.2967，-8.9209，-7.3558，-0.4359，0.4373) J120+2-yr (-4.8492，0.1382，0.3679，-4.5254，-9.0370，-6.3816，0.9147，0.8478) J120+10-yr (-1.8575，-0.1865，0.3679，-4.5254，-9.0370，-6.3816，0.9147，0.8478) J120+100-yr (6.7066，0.0483，0.3679，-4.5254，-9.0370，-6.3816，0.9147，0.8478)

[0162] S302. Define weights.

[0163] In this embodiment, the original vector is composed of an embedding vector and two types of attribute feature vectors, so a three-dimensional weight W = {w1, w2, w3} is defined to be used for weighted calculation of different parts of each original vector.

[0164] S303. Calculate the similarity matrix and the weighted vector distance.

[0165] According to the calculation results of the monitoring nodes in the plan library, the cosine similarity is used to calculate the similarity between the 96 plans to generate a 96×96 dimensional similarity matrix Next, calculate the weighted plan vector and The actual distance is calculated using the Euclidean distance || V i ′-V j ′||. is a 96×96 matrix with all diagonal elements set to 1, as shown in the following relationship:

[0166] tensor([[1.0000, 0.9902, 0.8566,..., -0.2986, –0.2488, 0.9528],

[0167] [0.9902, 1.0000, 0.9128, ..., -0.3165, -0.2757, 0.9029],

[0168] [0.8566, 0.9128, 1.0000, ..., -0.3201, -0.3130, 0.6759],

[0169] ...,

[0170] [-0.2986, -0.3165, -0.3201, ..., 1.0000, 0.9831, -0.1836],

[0171] [-0.2488, -0.2757, -0.3130, ..., 0.9831, 1.0000, -0.1088], |

[0172] [0.9528, 0.9029, 0.6759, ..., -0.1836, -0.1088, 1.0000]])

[0173] S304. Optimize weights and minimize the loss function.

[0174] Define the loss function The gradient descent method is used to continuously optimize the weights and minimize the loss function. The goal of the optimization process is to make the distance between the weighted different plan vectors in the vector space as close as possible to the similarity of the monitoring node calculation results. After training, the final plan vector is obtained. The final plan vector is shown in Table 4 below, taking "J119" and "J120" as examples:

[0175] Table 4. Plan vectors

[0176]

[0177] S305. Plan retrieval and matching.

[0178] Under a new rainfall scenario, the embedding vector and attribute feature vector of the node in the initial conditions can be directly calculated, and weighted according to the pre-trained model to obtain a weighted vector. Then, the vector matching algorithm is used to quickly retrieve the plan vector closest to the current input conditions in the plan vector library to achieve fuzzy search of the plan.

[0179] For example, for the "J119" node, a 50-year rainfall sequence different from the training set was implemented. The results were compared using a vectorized approach to find the first three similar plans in the plan library. These were then compared with similar plans that were hydraulically calculated using the cosine similarity of the monitoring node calculation results. The results, shown in Table 5 below, show that the plans selected by the two methods are nearly identical, demonstrating that the plans found using the vectorized approach avoid hydraulic calculations while still maintaining a certain degree of credibility.

[0180] Table 5

[0181]

[0182] This example provides a fuzzy search method for emergency plans based on the vectorization of pipeline node features. By building a rainfall emergency plan vector library in advance, multiple emergency plan situations can be efficiently expressed as a vector of fixed dimensions. When a new rainfall situation occurs, the initial calculation conditions are vectorized through a vector model, and the best matching solution is quickly found in the emergency plan library based on the similarity calculation technology of the vector space. Even if there are no similar initial calculation conditions in the emergency plan library, since the loss function of the vector model is the difference in the calculation results, different initial calculation conditions may have the same calculation results, so it is still possible to find an ideal emergency plan to support subsequent work. This significantly solves the problems of high computational complexity and response cost of existing technologies, and greatly improves the coverage of the emergency plan library.

[0183] Meanwhile, traditional emergency plan database matching methods require high levels of expertise and staff. This solution, through vectorized expression, significantly reduces the barrier to entry for using the emergency plan database due to the intuitive presentation of vector calculations. Furthermore, through appropriate training, multiple features can be incorporated into the emergency plan vector, fully leveraging the expressive power of historical data and multidimensional features. This avoids matching errors caused by insufficient features in traditional methods, allowing the emergency plan database to be flexibly expanded to accommodate more complex rainfall scenarios, thus providing strong support for accurate prediction and efficient scheduling of drainage pipe networks.

[0184] In this embodiment, a fuzzy search device for a plan based on the vectorization of pipeline node features is also provided. The device is used to implement the above-mentioned embodiments and preferred implementation methods. The details that have been described will not be repeated here. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.

[0185] This embodiment provides a fuzzy search device for emergency plans based on the quantization of pipeline network node features. Figure 5 As shown, the device includes:

[0186] Acquisition module 501 is used to obtain the real-time drainage network node attribute feature set and embedding vector of the target node in the target area to be found under the first rainfall condition, wherein the target node is the drainage network node corresponding to the first rainfall condition in the target area to be found, and one drainage network node corresponds to one embedding vector.

[0187] The generating module 502 is configured to generate a first original vector corresponding to the first rainfall condition according to the embedded vector and the real-time drainage network node attribute feature set.

[0188] The input module 503 is configured to input the first original vector into the plan vector model to obtain a first plan vector corresponding to the first rainfall condition.

[0189] The first determining module 504 is configured to determine a target pipeline network rainfall plan vector from a preset pipeline network rainfall plan vector library according to the first plan vector. The preset pipeline network rainfall plan vector library includes multiple plan vectors for multiple rainfall plans under different rainfall conditions.

[0190] The second determining module 505 determines the target pipeline network rainfall plan in a preset pipeline network rainfall plan library according to the target pipeline network rainfall plan vector. The preset pipeline network rainfall plan library includes multiple rainfall plans under different rainfall conditions.

[0191] The further functional description of each of the above modules is the same as that of the above corresponding embodiments and will not be repeated here.

[0192] The plan fuzzy search device based on the quantization of pipeline node feature vectors in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0193] The embodiment of the present invention also provides a computer device having the above Figure 5The fuzzy search device for emergency plans based on the vectorization of pipeline network node features is shown.

[0194] See also Figure 6 , Figure 6 is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 6 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of a GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 6 A processor 10 is taken as an example.

[0195] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0196] The memory 20 stores instructions that can be executed by at least one processor 10, so as to enable at least one processor 10 to execute the method shown in the above embodiment.

[0197] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0198] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0199] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0200] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0201] A portion of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium that can be accessed by the computer.

[0202] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A fuzzy search method for emergency plans based on the quantization of pipeline network node feature vectors, characterized in that: The method comprises: Obtaining a real-time drainage network node attribute feature set and an embedding vector of a target node in a target area to be searched under a first rainfall condition, wherein the target node is a drainage network node corresponding to the first rainfall condition in the target area to be searched, and each drainage network node corresponds to one embedding vector; generating a first original vector corresponding to the first rainfall condition according to the embedded vector and the real-time drainage network node attribute feature set; Inputting the first original vector into a plan vector model to obtain a first plan vector corresponding to the first rainfall condition; Determining a target pipeline network rainfall plan vector from a preset pipeline network rainfall plan vector library according to the first plan vector, wherein the preset pipeline network rainfall plan vector library includes a plurality of plan vectors for a plurality of rainfall plans under different rainfall conditions; A target pipeline network rainfall plan is determined in a preset pipeline network rainfall plan library according to the target pipeline network rainfall plan vector, wherein the preset pipeline network rainfall plan library includes multiple rainfall plans under different rainfall conditions.

2. The method according to claim 1, characterized in that Generating a first original vector corresponding to the first rainfall condition according to the embedded vector and the real-time drainage network node attribute feature set includes: Vectorizing the attribute features of each node in the real-time drainage network node attribute feature set to obtain an attribute feature vector of the target node; The embedded vector and the attribute feature vector are concatenated to generate the first original vector corresponding to the first rainfall condition.

3. The method according to claim 1, characterized in that The method further comprises: Obtaining a historical pipeline network rainfall plan library, a drainage pipeline network undirected topology graph, and multiple historical drainage pipeline network node attribute feature sets for the target area to be searched under multiple second rainfall conditions, wherein the multiple historical drainage pipeline network node attribute feature sets are used to characterize attribute characteristics of multiple drainage pipeline network nodes in the target area to be searched under the multiple second rainfall conditions; Performing vectorization processing on each drainage network node in the undirected topological graph of the drainage network to obtain multiple embedding vectors of the multiple drainage network nodes; generating a plurality of second original vectors corresponding to the plurality of second rainfall conditions according to the plurality of embedded vectors and the plurality of historical drainage network node attribute feature sets; The plan vector model is constructed according to the multiple second original vectors.

4. The method according to claim 3, characterized in that Constructing the plan vector model according to the plurality of second original vectors includes: Based on preset weights, weighted calculation is performed on the plurality of second original vectors to obtain a plurality of second plan vectors corresponding to the plurality of second rainfall conditions; Constructing a loss function based on the historical pipe network rainfall plan library and the plurality of second plan vectors; Optimizing the preset weights using a gradient descent method until the loss function value of the loss function meets the requirements, thereby obtaining a target weight; Based on the target weights, weighted calculation is performed on the plurality of second original vectors to obtain a plurality of target pipe network rainfall plan vectors corresponding to the plurality of second rainfall conditions; The plan vector model is constructed using the multiple second original vectors as input data and the multiple target pipe network rainfall plan vectors as output data.

5. The method according to claim 4, characterized in that Constructing a loss function based on the historical pipe network rainfall plan library and the plurality of second plan vectors, including: Constructing a plan target distance function based on the historical pipe network rainfall plan library; Constructing a plan vector actual distance function according to the plurality of second plan vectors; The loss function is constructed based on the plan target distance function and the plan vector actual distance function.

6. The method according to claim 4, characterized in that The method further comprises: The preset pipe network rainfall plan vector library is generated according to the plan vector model.

7. The method according to claim 3, characterized in that Obtain an undirected topological map of the drainage network, including: Obtain the drainage network topology; An undirected topological graph of the drainage pipe network is generated according to the topological structure of the drainage pipe network.

8. A fuzzy search device for emergency plans based on the quantization of pipeline network node features, characterized in that: The device comprises: an acquisition module, configured to acquire a real-time drainage network node attribute feature set and an embedding vector of a target node in a target area to be searched under a first rainfall condition, wherein the target node is a drainage network node corresponding to the first rainfall condition in the target area to be searched, and each drainage network node corresponds to one embedding vector; A generating module, configured to generate a first original vector corresponding to the first rainfall condition based on the embedded vector and the real-time drainage network node attribute feature set; an input module, configured to input the first original vector into a plan vector model to obtain a first plan vector corresponding to the first rainfall condition; a first determining module, configured to determine a target pipeline network rainfall plan vector from a preset pipeline network rainfall plan vector library based on the first plan vector, wherein the preset pipeline network rainfall plan vector library includes a plurality of plan vectors for a plurality of rainfall plans under different rainfall conditions; The second determining module is used to determine the target pipeline network rainfall plan in a preset pipeline network rainfall plan library according to the target pipeline network rainfall plan vector, and the preset pipeline network rainfall plan library includes multiple rainfall plans under different rainfall conditions.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the plan fuzzy search method based on pipeline network node feature vectorization according to any one of claims 1 to 7.

10. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the fuzzy search method for pre-plan based on pipeline network node feature vectorization according to any one of claims 1 to 7.

Images