Ecological corridor center migration prediction method based on MVMD-LSTM

Abstract

The present invention discloses a method for predicting the migration of the center of gravity of an ecological corridor based on MVMD-LSTM, which belongs to the field of geographic information technology, and includes: collecting various historical ecological data of the ecological corridor and preprocessing them, using clustering algorithms, decision algorithms and ArcGIS geographic spatial platform for analysis and calculation, and obtaining the migration process of the spatial center of gravity of the ecological corridor in different time periods; using the migration process of the spatial center of gravity of the ecological corridor and various historical ecological data of the ecological corridor, training and predicting through MVMD multivariate variational mode decomposition and LSTM long short-term memory network model, and obtaining the prediction result of the migration of the center of gravity of the ecological corridor in the future. By combining MVMD multivariate variational mode decomposition and LSTM long short-term memory network model, the present invention can effectively integrate and process the multimodal data of the ecological corridor and accurately predict the migration process of the center of gravity of the ecological corridor.

Description

Technical Field

[0001] The present invention belongs to the technical field of ecological corridors, and in particular relates to an ecological corridor gravity center migration prediction method based on MVMD-LSTM. Background Art

[0002] With the acceleration of urbanization and the increasing frequency of human activities, natural ecosystems are facing unprecedented pressure. As a key way to connect scattered natural landscape patches, ecological corridors can promote biodiversity conservation, enhance ecosystem service functions and improve human living environment. Ecological corridors can connect different ecosystems and play a vital role in biodiversity conservation and ecological balance maintenance. With the increase of human activities and the impact of climate change, the morphology and function of ecological corridors are undergoing rapid changes, including the migration of the center of gravity of ecological corridors. The prediction of the migration of the center of gravity of ecological corridors is crucial for the formulation of effective ecological protection strategies and land management planning. Summary of the invention

[0003] In order to meet the forward-looking needs of ecological protection and land planning, the present invention provides an ecological corridor center of gravity migration prediction method based on MVMD-LSTM, which can effectively process and integrate multimodal data by combining variational modes and LSTM, and solves the problem that it is difficult to integrate and accurately analyze multimodal ecological data when predicting the center of gravity migration of ecological corridors, and at the same time provides a reference for future reasonable land planning.

[0004] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: an ecological corridor center of gravity migration prediction method based on MVMD-LSTM, comprising the following steps:

[0005] S1: Collect historical ecological data of ecological corridors and use ecosystem service assessment models to estimate ecological service functions in different time periods;

[0006] S2: Based on the ecological service functions in different time periods, clustering algorithm and decision algorithm are used to process and obtain the comprehensive resistance surface in different time periods;

[0007] S3: Analyze the comprehensive resistance surface in different time periods, and calculate the migration process of the ecological corridor space center in different time periods;

[0008] S4: Preprocess the migration process of the spatial center of gravity of the ecological corridor in different time periods and various historical ecological data of the ecological corridor to obtain the characteristic data of the ecological corridor;

[0009] S5: According to the characteristic data of the ecological corridor and the migration process of the spatial center of gravity of the ecological corridor in different time periods, the MVMD multivariate variational mode decomposition and LSTM long short-term memory network model are used to obtain the prediction results of the future migration of the center of gravity of the ecological corridor.

[0010] The beneficial effects of the present invention are as follows: the present invention utilizes MVMD multivariate variational mode decomposition to process high-dimensional data, which can reduce complexity and facilitate capturing the dynamic characteristics of the migration of the center of gravity of the ecological corridor. The LSTM long short-term memory network model is used to deeply analyze the time series, and regularization technology and cross-validation are used to prevent overfitting, thereby enhancing the generalization ability of the LSTM long short-term memory network model, and being able to accurately predict the migration process of the center of gravity of the ecological corridor.

[0011] Further: the specific steps of S2 are as follows:

[0012] S201: According to the ecological service functions in different time periods, the k-means clustering algorithm is used to classify and obtain several classification areas in different time periods;

[0013] S202: Select high-value cluster areas from several classification areas as ecological sources, and select natural and socio-economic impact factors as resistance factors;

[0014] S203: Taking the ecological source as the center, the resistance value of the resistance factor is calculated using the expert scoring method, and the weight of the resistance factor is calculated using the AHP hierarchical analysis method;

[0015] S204: constructing a comprehensive resistance surface according to the resistance value of the resistance factor and the weight of the resistance factor;

[0016] S205: Determine whether a comprehensive resistance surface is constructed for a plurality of classification areas in different time periods. If so, obtain the comprehensive resistance surface for different time periods. Otherwise, return to S202.

[0017] The beneficial effects of the above further scheme are: by processing the ecological service functions in different time periods through the expert scoring method and the AHP hierarchical analysis method, the resistance value and weight of the resistance factor can be comprehensively evaluated, and the importance of the resistance factor in the evaluation system can be fully and accurately reflected.

[0018] Further: the specific steps of S3 are as follows:

[0019] S301: Analyze and calculate the minimum cost path for each time period based on the comprehensive resistance surface in different time periods to obtain potential ecological corridors;

[0020] S302: Use the gravity model to quantify the interaction forces between ecological sources in different time periods, and define the potential ecological corridors between ecological sources with interaction force values ​​greater than the set threshold as important corridors, and define the potential ecological corridors that are not defined as important corridors as general corridors;

[0021] S303: According to the important corridors and general corridors in different time periods, the spatial center of gravity migration process of the ecological corridors in different time periods is calculated;

[0022] The beneficial effect of the above further scheme is that the spatial center of gravity migration process of the ecological corridor is calculated through the ArcGIS geospatial platform, which can ensure the accuracy and scientificity of the calculation results.

[0023] Further: The gravity model is used to quantify the interaction force between ecological sources, and its calculation expression is as follows:

[0024]

[0025] Among them, G ab is the interaction force between ecological source a and ecological source b, N a is the weight value of ecological source a, N b is the weight value of ecological source b, D ab is the resistance value between ecological source a and ecological source b, P a is the resistance value of ecological source a, P b is the resistance value of ecological source b, S a is the area of ​​ecological source a, S b is the area of ​​ecological source b, L ab is the cost path value between ecological source a and ecological source b, L max The maximum cumulative resistance value for all corridors.

[0026] The beneficial effect of the above further scheme is: by calculating the interaction force between ecological sources by the above formula, it is possible to scientifically determine the important corridors between ecological sources and improve reliability.

[0027] Furthermore, the ArcGIS geospatial platform is used to calculate and obtain the migration process of the spatial center of gravity of the ecological corridor in different time periods. The calculation expression is as follows:

[0028]

[0029] Among them, x is the longitude of the center of gravity of the ecological corridor space, y is the latitude of the center of gravity of the ecological corridor space, and W i is the weight of the ith region, X i is the longitude of the ith region, Y i is the latitude of the ith region.

[0030] The beneficial effect of the above further scheme is: through the calculation expression of the spatial center of gravity migration process of the ecological corridor, the spatial center of gravity migration process of the ecological corridor can be accurately calculated based on the historical ecological data of the ecological corridor, providing data for the subsequent training of the MVMD-LSTM mixed variational modal long-term and short-term neural network model.

[0031] Further: the specific steps of S5 are as follows:

[0032] S501: According to the characteristic data of the ecological corridor, the principal component analysis method is used to reduce the dimension and obtain the characteristic factors of the ecological corridor;

[0033] S502: Decomposing the characteristic factors of the ecological corridor using MVMD multivariate variational mode to obtain modal components of several ecological corridors;

[0034] S503: adding and reconstructing the modal components of several ecological corridors to obtain fused modal components;

[0035] S504: According to the fused modal components, the LSTM long short-term memory network model is trained in combination with regularization technology and cross-validation to obtain the MVMD-LSTM hybrid variational modal long short-term neural network model;

[0036] S505: Based on the historical ecological data of the ecological corridor and the migration process of the spatial center of gravity of the ecological corridor in different time periods, the MVMD-LSTM hybrid variational mode long-term and short-term neural network model is used to predict and obtain the prediction results of the future migration of the center of gravity of the ecological corridor.

[0037] The beneficial effects of the above further scheme are: according to the characteristic data of the ecological corridor, based on the MVMD multivariate variational modal decomposition and the LSTM long short-term memory network model, a MVMD-LSTM hybrid variational modal long-term and short-term neural network model is constructed, which can effectively process complex multimodal ecological corridor data and accurately predict the future ecological corridor center of gravity migration process.

[0038] Further: the specific steps of S504 are as follows:

[0039] A1: According to the fused modal components, the dataset is divided into training set, validation set and test set using cross-validation.

[0040] A2: According to the regularization technology, a Dropout layer is added after the LSTM layer of the LSTM long short-term memory network model, and a regularization term is added to the loss function of the LSTM long short-term memory network model to obtain an improved LSTM long short-term memory network model;

[0041] A3: Use the training set, validation set, and test set to train the LSTM long short-term memory network model improved by regularization technology, and monitor the loss of the validation set;

[0042] A4: Determine whether the loss of the validation set stops improving. If so, use the callback function Callbacks to end the training early to obtain the trained LSTM long short-term memory network model, and combine it with the MVMD multivariate mode decomposition to obtain the MVMD-LSTM hybrid variational mode long short-term neural network model.

[0043] The beneficial effects of the above further scheme are: the LSTM long short-term memory network model is improved and trained through regularization technology, the trained LSTM long short-term memory network model is combined with MVMD multivariate variational modal decomposition, and the multimodal ecological data is integrated and processed by MVMD multivariate variational modal decomposition, so that the trained LSTM long short-term memory network model can comprehensively analyze the data and accurately predict the center of gravity migration of the ecological corridor. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a flow chart of the ecological corridor center of gravity migration prediction method based on MVMD-LSTM. DETAILED DESCRIPTION

[0045] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0046] like Figure 1 As shown in the figure, a method for predicting the migration of the center of gravity of an ecological corridor based on MVMD-LSTM includes the following steps:

[0047] S1: Collect historical ecological data of ecological corridors and use ecosystem service assessment models to estimate ecological service functions in different time periods;

[0048] S2: Based on the ecological service functions in different time periods, clustering algorithm and decision algorithm are used to process and obtain the comprehensive resistance surface in different time periods;

[0049] S3: Analyze the comprehensive resistance surface in different time periods, and calculate the migration process of the ecological corridor space center in different time periods;

[0050] S4: Preprocess the migration process of the spatial center of gravity of the ecological corridor in different time periods and various historical ecological data of the ecological corridor to obtain the characteristic data of the ecological corridor;

[0051] S5: According to the characteristic data of the ecological corridor and the migration process of the spatial center of gravity of the ecological corridor in different time periods, the MVMD multivariate variational mode decomposition and LSTM long short-term memory network model are used to obtain the prediction results of the future migration of the center of gravity of the ecological corridor.

[0052] In one embodiment of the present invention, the ecosystem service assessment model in S1 includes the InVEST model, the RUSLE model, the RWEQ model and the CASA model. The ecological data input and ecological service function output of each model are shown in Table 1 below.

[0053] Table 1

[0054]

[0055]

[0056] Among them, the ecological data input of the InVEST model includes land use data, topography data, soil data, vegetation data and evapotranspiration data, which can estimate the water conservation service function; the ecological data input of the RUSLE model includes rainfall data, soil data, vegetation data and topography data, which can estimate the soil conservation service function; the ecological data input of the RWEQ model includes meteorological data, soil data, vegetation cover data and wind speed data, which can estimate the water conservation service function; the ecological data input of the CASA model includes meteorological data, vegetation data and solar radiation data, which can estimate the carbon storage service function.

[0057] In one embodiment of the present invention, the specific steps of S2 are as follows:

[0058] S201: According to the ecological service functions in different time periods, the k-means clustering algorithm is used to classify and obtain several classification areas in different time periods;

[0059] S202: Select high-value cluster areas from several classification areas as ecological sources, and select natural and socio-economic impact factors as resistance factors;

[0060] S203: Taking the ecological source as the center, the resistance value of the resistance factor is calculated using the expert scoring method, and the weight of the resistance factor is calculated using the AHP hierarchical analysis method;

[0061] S204: constructing a comprehensive resistance surface according to the resistance value of the resistance factor and the weight of the resistance factor;

[0062] S205: Determine whether a comprehensive resistance surface is constructed for a plurality of classification areas in different time periods. If so, obtain the comprehensive resistance surface for different time periods. Otherwise, return to S202.

[0063] In one embodiment of the present invention, the specific steps of S3 are as follows:

[0064] S301: Analyze and calculate the minimum cost path of each time period according to the comprehensive resistance surface of different time periods to obtain potential ecological corridors; wherein, the cost-distance path planning module and the cost-path cost path module of the ArcGIS geospatial platform are used for calculation, so that the potential ecological corridors can be obtained simply and quickly;

[0065] S302: The gravity model is used to quantify the interaction force between ecological sources in the comprehensive resistance surface in different time periods, and the potential ecological corridors with interaction force values ​​greater than the set threshold are defined as important corridors, and the potential ecological corridors that are not defined as important corridors are defined as general corridors; wherein, the gravity model is used to quantify the interaction force between ecological sources, and its calculation expression is as follows:

[0066]

[0067] Among them, G ab is the interaction force between ecological source a and ecological source b, N a is the weight value of ecological source a, N b is the weight value of ecological source b, D ab is the resistance value between ecological source a and ecological source b, P a is the resistance value of ecological source a, P b is the resistance value of ecological source b, S a is the area of ​​ecological source a, S b is the area of ​​ecological source b, L ab is the cost path value between ecological source a and ecological source b, L max The maximum value of the accumulated resistance of all corridors;

[0068] S303: Based on the ecological corridors in different time periods, the ArcGIS geographic spatial platform is used to calculate and obtain the spatial center of gravity migration process of the ecological corridors in different time periods, including: the spatial center of gravity migration process of important corridors and the spatial center of gravity migration process of general corridors. The expression is as follows:

[0069]

[0070] Among them, x is the longitude of the center of gravity of the ecological corridor space, y is the latitude of the center of gravity of the ecological corridor space, and W i is the weight of the ith region, Xi is the longitude of the ith region, Y i is the latitude of the ith region.

[0071] In one embodiment of the present invention, the specific steps of S5 are as follows:

[0072] S501: According to the characteristic data of the ecological corridor, the principal component analysis method is used to reduce the dimension and obtain the characteristic factors of the ecological corridor;

[0073] S502: Decompose the characteristic factors of the ecological corridor using MVMD multivariate modal decomposition to obtain modal components of several ecological corridors;

[0074] S503: adding and reconstructing the modal components of several ecological corridors to obtain fused modal components;

[0075] S504: According to the fused modal components, the LSTM long short-term memory network model is trained in combination with regularization technology and cross-validation to obtain the MVMD-LSTM hybrid variational modal long short-term neural network model;

[0076] S505: Based on the historical ecological data of the ecological corridor and the migration process of the spatial center of gravity of the ecological corridor in different time periods, the MVMD-LSTM hybrid variational mode long-term and short-term neural network model is used to predict and obtain the prediction results of the future migration of the center of gravity of the ecological corridor.

[0077] According to the fused modal components, the LSTM long short-term memory network model is trained by combining regularization technology and cross-validation to obtain the MVMD-LSTM hybrid variational modal long short-term neural network model. The specific implementation method is as follows:

[0078] A1: According to the fused modal components, the dataset is divided into training set, validation set and test set using cross-validation.

[0079] A2: Based on the regularization technology, a Dropout layer is added after the LSTM layer of the LSTM long short-term memory network model. The Dropout layer is set to the optimal parameters. The Dropout layer can randomly discard a certain proportion of neurons during the training process to reduce the risk of overfitting.

[0080] A regularization term is added to the loss function of the LSTM long short-term memory network model to obtain an improved LSTM long short-term memory network model; wherein the regularization term includes an L1 regularization term or an L2 regularization term, L1 regularization tends to produce a sparse weight matrix, in which many weight parameters are zero, and L2 regularization tends to make the weight parameters close to zero but not zero;

[0081] A3: Use the training set, validation set, and test set to train the LSTM long short-term memory network model improved by regularization technology, and monitor the loss of the validation set;

[0082] A4: Determine whether the loss of the validation set stops improving. If so, use the callback function Callbacks to end the training early to avoid overfitting, obtain the trained LSTM long short-term memory network model, and combine it with the MVMD multivariate mode decomposition to obtain the MVMD-LSTM hybrid variational mode long short-term neural network model.

[0083] The beneficial effects of the present invention are as follows: by combining MVMD multivariate variational mode decomposition and LSTM long short-term memory network model, the present invention can effectively integrate and process the multimodal data of ecological corridors, thereby improving the prediction accuracy; using MVMD multivariate variational mode decomposition to process high-dimensional data can reduce complexity and facilitate capturing the dynamic characteristics of the migration of the center of gravity of the ecological corridor; using the LSTM long short-term memory network model to deeply analyze the time series, and preventing overfitting through regularization technology and cross-validation, the generalization ability of the LSTM long short-term memory network model is enhanced, and the migration process of the center of gravity of the ecological corridor can be accurately predicted.

Claims

1. A method for predicting the migration of the center of gravity of an ecological corridor based on MVMD-LSTM, characterized in that: The following steps are involved: S1: Collect historical ecological data of ecological corridors and use ecosystem service assessment models to estimate ecological service functions in different time periods; S2: Based on the ecological service functions in different time periods, clustering algorithm and decision algorithm are used to process and obtain the comprehensive resistance surface in different time periods; S3: Analyze the comprehensive resistance surface in different time periods, and calculate the migration process of the ecological corridor space center in different time periods, which is as follows: S301: Analyze and calculate the minimum cost path for each time period based on the comprehensive resistance surface in different time periods to obtain potential ecological corridors; S302: Use the gravity model to quantify the interaction forces between ecological sources in different time periods, and define the potential ecological corridors between ecological sources with interaction force values ​​greater than the set threshold as important corridors, and define the potential ecological corridors that are not defined as important corridors as general corridors; S303: According to the important corridors and general corridors in different time periods, the spatial center of gravity migration process of the ecological corridors in different time periods is calculated; S4: Preprocess the migration process of the spatial center of gravity of the ecological corridor in different time periods and various historical ecological data of the ecological corridor to obtain the characteristic data of the ecological corridor; S5: According to the characteristic data of the ecological corridor and the migration process of the spatial center of gravity of the ecological corridor in different time periods, the MVMD multivariate variational mode decomposition and LSTM long short-term memory network model are used to obtain the prediction results of the future migration of the center of gravity of the ecological corridor.

2. The method for predicting the migration of the center of gravity of an ecological corridor based on MVMD-LSTM according to claim 1 is characterized in that: The specific steps of S2 are as follows: S201: According to the ecological service functions in different time periods, the k-means clustering algorithm is used to classify and obtain several classification areas in different time periods; S202: Select high-value cluster areas from several classification areas as ecological sources, and select natural and socio-economic impact factors as resistance factors; S203: Taking the ecological source as the center, the resistance value of the resistance factor is calculated using the expert scoring method, and the weight of the resistance factor is calculated using the AHP hierarchical analysis method; S204: constructing a comprehensive resistance surface according to the resistance value of the resistance factor and the weight of the resistance factor; S205: Determine whether a comprehensive resistance surface is constructed for a plurality of classification areas in different time periods. If so, obtain the comprehensive resistance surface for different time periods. Otherwise, return to S202.

3. The method for predicting the migration of the center of gravity of an ecological corridor based on MVMD-LSTM according to claim 1 is characterized in that: The gravity model is used to quantify the interaction force between ecological sources, and its calculation expression is as follows: in, As ecological source and ecological source The interaction force between As ecological source The weight value of As ecological source The weight value of As ecological source and ecological source The resistance value between As ecological source The resistance value, As ecological source The resistance value, As ecological source The area of As ecological source The area of As ecological source and ecological source The cost path value between The maximum cumulative resistance of all corridors.

4. The method for predicting the migration of the center of gravity of an ecological corridor based on MVMD-LSTM according to claim 1 is characterized in that: According to the ecological corridors in different time periods, the migration process of the spatial center of gravity of the ecological corridors in different time periods is calculated, and the calculation expression is as follows: in, is the longitude of the spatial center of gravity of the ecological corridor, is the latitude of the spatial center of gravity of the ecological corridor, For the The weight of the region, For the The longitude of the region, For the The latitude of the region.

5. The method for predicting the migration of the center of gravity of an ecological corridor based on MVMD-LSTM according to claim 1 is characterized in that: The specific steps of S5 are as follows: S501: According to the characteristic data of the ecological corridor, the principal component analysis method is used to reduce the dimension and obtain the characteristic factors of the ecological corridor; S502: Decomposing the characteristic factors of the ecological corridor using MVMD multivariate variational mode to obtain modal components of several ecological corridors; S503: adding and reconstructing the modal components of several ecological corridors to obtain fused modal components; S504: According to the fused modal components, the LSTM long short-term memory network model is trained in combination with regularization technology and cross-validation to obtain the MVMD-LSTM hybrid variational modal long short-term neural network model; S505: Based on the historical ecological data of the ecological corridor and the migration process of the spatial center of gravity of the ecological corridor in different time periods, the MVMD-LSTM hybrid variational mode long-term and short-term neural network model is used to predict and obtain the prediction results of the future migration of the center of gravity of the ecological corridor.

6. The method for predicting the migration of the center of gravity of an ecological corridor based on MVMD-LSTM according to claim 5 is characterized in that: The specific steps of S504 are as follows: A1: According to the fused modal components, the dataset is divided into training set, validation set and test set using cross-validation. A2: According to the regularization technology, a Dropout layer is added after the LSTM layer of the LSTM long short-term memory network model, and a regularization term is added to the loss function of the LSTM long short-term memory network model to obtain an improved LSTM long short-term memory network model; A3: Use the training set, validation set, and test set to train the LSTM long short-term memory network model improved by regularization technology, and monitor the loss of the validation set; A4: Determine whether the loss of the validation set stops improving. If so, use the callback function Callbacks to end the training early to obtain the trained LSTM long short-term memory network model, and combine it with the MVMD multivariate mode decomposition to obtain the MVMD-LSTM hybrid variational mode long short-term neural network model.

Images