A future-year multi-target water demand prediction method and device based on LSTM

Abstract

The application discloses a future year multi-target water demand prediction method and device based on LSTM, and relates to the field of water demand prediction.The method comprises the following steps: acquiring key influence factors of water demand of each industry; inputting the known key influence factors of water demand of each industry into corresponding LSTM models after training respectively, to obtain key influence factor prediction results of a future preset period corresponding to water demand of each industry; and inputting the key influence factor prediction results of water demand of each industry into corresponding machine learning models after training respectively, to obtain water demand prediction results of the future preset period corresponding to water demand of each industry.In the application, for the prediction of water demand of each industry, the key influence factors of the future period are predicted by using the LSTM model, the predicted key influence factors of the future period are taken as input, and the water demand of the future period is obtained by inputting into the machine learning model, so that the accuracy of the prediction of water demand of each industry in the future period is improved.

Description

Technical Field

[0001] This application relates to the field of water demand forecasting, and in particular to a method and apparatus for forecasting multi-objective water demand for future years based on LSTM. Background Technology

[0002] For water demand forecasting, the existing technology generally follows these steps: 1. Determine water demand influencing factors (input to the forecasting model): Assess the correlation between influencing factors and water demand, and use those with strong correlations as input to the forecasting model. 2. Determine the forecasting model: Current water demand forecasting typically involves establishing a forecasting model, training it based on the determined influencing factors and the actual output (actual water demand) to achieve a good fit, thereby enabling forecasting. However, the accuracy of existing forecasting models still has certain limitations. Summary of the Invention

[0003] The purpose of this application is to provide a method and apparatus for predicting multi-objective water demand in the future based on LSTM, which can improve the accuracy of water demand prediction for multiple industries in the future.

[0004] To achieve the above objectives, this application provides the following solution:

[0005] Firstly, this application provides a multi-objective water demand prediction method for future years based on LSTM, including:

[0006] Identify the key influencing factors of water demand in various industries;

[0007] The known key influencing factors of water demand for each industry are input into the corresponding trained LSTM model to obtain the prediction results of key influencing factors for water demand in the future preset period for each industry.

[0008] The prediction results of the key influencing factors corresponding to the water demand of each industry are input into the corresponding trained machine learning model to obtain the water demand prediction results for the future preset period corresponding to the water demand of each industry.

[0009] Secondly, this application provides a multi-objective water demand prediction device for future years based on LSTM, including:

[0010] The key impact factor acquisition module is used to acquire the key impact factors of water demand in various industries;

[0011] The key impact factor prediction module is used to input the known key impact factors of water demand in each industry into the corresponding trained LSTM model to obtain the prediction results of key impact factors for the future preset period of water demand in each industry.

[0012] The water demand prediction module is used to input the prediction results of the key influencing factors corresponding to the water demand of each industry into the corresponding trained machine learning model to obtain the water demand prediction results for the future preset period corresponding to the water demand of each industry.

[0013] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described LSTM-based multi-objective water demand forecasting method for future years.

[0014] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described LSTM-based multi-objective water demand forecasting method for future years.

[0015] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described LSTM-based multi-objective water demand forecasting method for future years.

[0016] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0017] This application provides a method and apparatus for predicting multi-objective water demand in future years based on LSTM. The method obtains key influencing factors of water demand in various industries; inputs the known key influencing factors of water demand for each industry into the corresponding trained LSTM model to obtain the prediction results of key influencing factors for the future preset time period corresponding to each industry's water demand; and inputs the prediction results of key influencing factors for water demand for each industry into the corresponding trained machine learning model to obtain the prediction results of water demand for the future preset time period corresponding to each industry's water demand. In this application, for predicting water demand in various industries, historical key influencing factors are first used to predict key influencing factors for the future time period using an LSTM model. Then, the predicted key influencing factors for the future time period are used as input into the machine learning model to predict the water demand for the future time period. Compared with existing methods that directly use historical key influencing factors to predict industry water demand in the future time period, this application uses key influencing factors for the future time period to predict water demand in the future, improving the accuracy of water demand prediction for various industries in the future time period. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is an application environment diagram of a multi-objective water demand prediction method based on LSTM for future years, as described in one embodiment of this application.

[0020] Figure 2 A flowchart illustrating a multi-objective water demand forecasting method based on LSTM for future years, provided as an embodiment of this application;

[0021] Figure 3 A schematic diagram illustrating the technical concept of a multi-objective water demand forecasting method based on LSTM for future years provided in an embodiment of this application;

[0022] Figure 4 This is a schematic diagram of the structure of an LSTM model provided in one embodiment of this application;

[0023] Figure 5 This is a schematic diagram of the structure of a BP neural network model provided in one embodiment of this application;

[0024] Figure 6 A schematic diagram of the training loss of an LSTM model provided in an embodiment of this application;

[0025] Figure 7 A schematic diagram of the training loss of an LSTM model provided in an embodiment of this application;

[0026] Figure 8 A schematic diagram of the training loss of a BP neural network model provided in an embodiment of this application;

[0027] Figure 9 A schematic diagram of prediction results under the LSTM-BP prediction framework provided in an embodiment of this application;

[0028] Figure 10 A schematic diagram of the functional modules of a multi-objective water demand prediction device based on LSTM for future years provided in an embodiment of this application;

[0029] Figure 11 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0030] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0031] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0032] The LSTM-based multi-objective water demand prediction method for future years provided in this application can be applied to, for example... Figure 1 In the application environment shown, the terminal communicates with the server via a network. The data storage system stores the data the server needs to process. The data storage system can be set up independently, integrated into the server, or placed in the cloud or on another server. The terminal can send key influencing factors of water demand for each industry to the server. After receiving these factors, the server inputs the known key influencing factors for each industry's water demand into the corresponding trained LSTM model to obtain the prediction results of key influencing factors for the future preset time period corresponding to each industry's water demand. The server then inputs these prediction results into the corresponding trained machine learning model to obtain the prediction results of water demand for the future preset time period corresponding to each industry's water demand. The server can then feed back the obtained prediction results of water demand for the future preset time period corresponding to each industry to the terminal. Furthermore, in some embodiments, the LSTM-based multi-objective water demand prediction method for future years can also be implemented independently by the server or the terminal. For example, the terminal can directly perform LSTM-based multi-objective water demand prediction for future years based on the key influencing factors of water demand for each industry, or the server can obtain the key influencing factors of water demand for each industry from the data storage system and perform LSTM-based multi-objective water demand prediction for future years.

[0033] The terminal can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. The server can be a standalone server, a server cluster consisting of multiple servers, or a cloud server.

[0034] In one exemplary embodiment, such as Figure 2 and Figure 3 As shown, a multi-objective water demand forecasting method based on LSTM is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 The following steps, 101 to 103, are used as an example to illustrate the process of using a server in the example.

[0035] Step 101: Obtain the key influencing factors of water demand in various industries. Specifically, the water demand of each industry includes: agricultural water demand, forestry, animal husbandry, fishery and livestock water demand, industrial water demand, urban public water demand, domestic water demand, and ecological environment water demand.

[0036] Step 102: Input the known key influencing factors of water demand for each industry into the corresponding trained LSTM model to obtain the prediction results of key influencing factors for the future preset period of water demand for each industry.

[0037] Step 103: Input the predicted results of the key influencing factors corresponding to the water demand of each industry into the corresponding trained machine learning model to obtain the predicted water demand for the future preset period for each industry. As an example, the machine learning model can use a BP neural network.

[0038] By implementing steps 101 to 103 above, for the prediction of water demand in various industries, firstly, key influencing factors from historical data are used to predict key influencing factors for future periods using an LSTM model. Then, the predicted key influencing factors for future periods are used as input to a machine learning model to predict water demand for future periods. Existing prediction methods often focus on constructing a model of the relationship between influencing factors (key influencing factors) and water demand, neglecting the need to calculate key influencing factors for future years when predicting water demand for future years. Although some researchers in the water conservancy industry use traditional methods such as the GM grey prediction model and ARIMA (differential autoregressive moving average model) to calculate input for future years, these methods suffer from problems such as lack of time correlation and limited input data, leading to simplistic trends and less than ideal results. Compared to existing methods that directly use historical key influencing factors to predict industry water demand for future periods, this application uses key influencing factors for future periods to predict water demand for future periods, improving the accuracy of water demand prediction for various industries in the future. In addition, compared with existing predictions of key influencing factors for future years, this application uses an LSTM model for prediction, which takes into account the temporal correlation of key influencing factors, improves the accuracy of key influencing factor prediction, and further improves the accuracy of water demand prediction for various industries in the future period.

[0039] In another exemplary embodiment of this application, in step 101, regarding the screening of key influencing factors, the initially selected potentially relevant influencing factors are Gross Domestic Product, total national population, total urban population, total rural population, cultivated land area, effective irrigated area, number of large livestock, number of small livestock, number of poultry, annual rainfall, and water resources. The outputs to be predicted are agricultural water demand, forestry, animal husbandry, fishery and livestock demand, industrial water demand, urban public water demand, domestic water demand, and ecological environment water demand.

[0040] Based on the above, step 101 involves obtaining the key influencing factors of water demand in various industries, specifically including:

[0041] (1) Obtain preliminary influencing factors of water demand in various industries; the preliminary influencing factors include gross national product, total national population, total urban population, total rural population, cultivated land area, effective irrigated area, number of large livestock, number of small livestock, number of poultry, annual rainfall and water resources.

[0042] (2) Use the Spearman rank correlation test to screen out the key influencing factors corresponding to the water demand of each industry from the initial selection of influencing factors.

[0043] In another exemplary embodiment of this application, the Spearman rank correlation test is used to determine the correlation between input and output. The calculation method involves arranging any set of inputs and outputs in ascending order and assigning them ranks. The corresponding rank difference is then calculated as: D = {d1, d2, d3, ... d...} n Then the correlation of each input-output pair is calculated as follows:

[0044]

[0045] Therefore, the correlation coefficient matrix for each input-output pair can be obtained as follows:

[0046]

[0047] Where n represents the number of samples; i refers to the water demand of each industry, such as the water demand of the 6 industries mentioned above; j refers to each of the initial selection factors, such as the 11 initial selection factors mentioned above.

[0048] Finally, by setting the correlation threshold ε for water demand in each industry, the most relevant influencing factors for water demand in each industry are determined, namely the key influencing factors.

[0049] Therefore, in step (2) above, the Spearman rank correlation test is used to screen out the key influencing factors corresponding to the water demand of each industry from the initially selected influencing factors, specifically including:

[0050] (2-1) For each variable pair of water demand in each industry, sort all historical observations of the variable pair according to their magnitude; a preliminary selection of influencing factors and the corresponding industry water demand constitute a variable pair.

[0051] (2-2) For each pair of variables, calculate the rank difference of each set of historical observations of the pair of variables based on the sorted position of the pair.

[0052] (2-3) For each pair of variables, calculate the correlation of the pair of variables based on the rank difference corresponding to each set of historical observations of the pair of variables.

[0053] (2-4) Based on the comparison results of the correlation of each variable pair and the corresponding correlation threshold, the key influencing factors corresponding to the water demand of each industry are selected from the preliminary influencing factors.

[0054] In another exemplary embodiment of this application, in step 102, the LSTM model is a Long Short-Term Memory network, which is an optimization based on RNN networks and has good performance in handling short time series prediction problems. This application constructs LSTM models with different parameters for each type of output (water demand of each industry) corresponding to the input. Taking four inputs (four key influencing factors) corresponding to one type of output (water demand of a certain industry) as an example, 10 years of historical data are selected for model training to calculate the water demand value for the next k years. The known historical data for the past 10 years is shown below.

[0055]

[0056] Where a, b, c, and d represent four key influencing factors, and o represents the historical actual water demand.

[0057] The training dataset for an LSTM model is constructed using key impact factor data from the past 10 years. A sliding window is set; for example, a sliding window of w=4 is chosen (the optimal value can be determined through adjustments). The training dataset used for training is then represented as follows:

[0058]

[0059] The data in the above dataset that predicts the next year is merely an example; the number of future years to be predicted can be adjusted according to needs. After constructing the training dataset, due to the limited amount of data (only 6 years of dataset can be created from 10 years of data), 5 years of the dataset were used as the training dataset to train the LSTM model, with the data randomly shuffled during training. The last year was used as the prediction data. After training, all data was used as the training set to redetermine the model parameters. Experimental testing showed that the most suitable network structure for this dataset is as follows: Figure 4 As shown, since key impact factors include multiple types, each type of key impact factor corresponds to an LSTM model. Therefore, the four key impact factors a, b, c, and d mentioned above have four LSTM models. The four LSTM models have the same model structure, but the model parameters, such as weights, biases, and learning rates, are different.

[0060] Based on the above, in step 102, the known key influencing factors of water demand for each industry are input into the corresponding trained LSTM model to obtain the prediction results of key influencing factors for the future preset period corresponding to water demand for each industry, specifically including:

[0061] (A1) Obtain historical water demand data for each industry; the historical water demand data includes key historical influencing factors and corresponding historical actual water demand over several years.

[0062] (A2) Divide the historical key impact factors over several years into multiple arrays (a sliding window method can be used for array division); each array includes the historical key impact factors for the first historical period and the corresponding historical key impact factors for the second historical period; the time points of the first historical period are shorter than the time points of the second historical period. For example, the first historical period is from year 1 to year 4, and the second historical period is year 5; the first historical period is from year 2 to year 5, and the second historical period is year 6; the first historical period is from year 3 to year 6, and the second historical period is year 6.

[0063] (A3) Using the historical key impact factors of the first historical period in each array as input, and the corresponding historical key impact factors of the second historical period as labels, train the corresponding LSTM model to obtain the trained LSTM model.

[0064] Figure 4 In the basic network structure of the LSTM model shown, the number of input layers is consistent with the number of sliding windows, and the numbering is the computation timing sequence, containing 4 LSTM gate units.

[0065] (A4) Input the known key influencing factors of water demand for each industry into the corresponding trained LSTM model to obtain the prediction results of key influencing factors for the future preset period of water demand for each industry.

[0066] After training, using data from recent years as input, and setting a sliding window, the predicted value for each future year is calculated sequentially for each category of output. The data calculated using ten years of data (e.g., 2012-2021) is then represented as follows:

[0067]

[0068] The data (a, b, c, d) from 2021 to 2035 represent the predicted key influencing factors.

[0069] In another exemplary embodiment of this application, in step 103, after acquiring the key influencing factor data for future years, a BP neural network is trained using historical key influencing factors and corresponding historical actual water demand. The trained BP neural network is then used to predict the water demand forecast for future years based on the key influencing factors. Again, using four key influencing factors as input, after extensive debugging and parameter tuning, the constructed BP network structure is as follows: Figure 5 As shown. Figure 5It contains 3 hidden layers and 3 drop layers, each layer containing 9 neurons. The drop layers require a probability p (0.5 in this case), which randomly sets the neuron output to 0. The input and output correspondence for each layer is as follows:

[0070] y t =relu(Y t-1 *W t-1 +b t-1 )

[0071] In the formula, y t Y represents the output value of a single neuron in the next layer, where ReLU is the activation function. t-1 W is the input vector of the previous layer. t-1 Let b be the weight matrix. t-1 This is the bias vector. The final output value is calculated through forward propagation, and the loss is then calculated using the loss function.

[0072]

[0073] In the formula This represents the true value of the calculated result. The gradient is updated in the reverse direction using the gradient descent principle. Here, the learning rate is set to 0.03, the number of iterations is 1500, and data from the past 10 years is used as the training and testing sets. The model is adjusted to minimize its loss, thereby determining the parameters and completing the multi-objective water demand prediction for future years.

[0074] Based on the above, in step 103, the prediction results of the key influencing factors corresponding to the water demand of each industry are input into the corresponding trained machine learning model to obtain the water demand prediction results for the future preset period corresponding to the water demand of each industry, specifically including:

[0075] (B1) Obtain historical water demand data for each industry; the historical water demand data includes key historical influencing factors and corresponding historical actual water demand over several years.

[0076] (B2) Using the historical key influencing factors of each year as input and the corresponding historical actual water demand as label, train the corresponding machine learning model to obtain the trained machine learning model.

[0077] (B3) Input the prediction results of the key influencing factors corresponding to the water demand of each industry into the corresponding trained machine learning model to obtain the prediction results of the water demand for the future preset period corresponding to the water demand of each industry.

[0078] This application first assesses the correlation between influencing factors and water demand in various industries, identifying key influencing factors for each industry's water demand. For example, for six industries' water demand, there should be six different key influencing factor matrices. After designing and training an LSTM network structure, the future values ​​of key influencing factors for the future time period required for water demand prediction in each industry are obtained. A BP network is trained using existing historical data, and the corresponding water demand values ​​are obtained by inputting key influencing factor data for future years. This application utilizes the LSTM network principle for key influencing factor prediction, which enhances the time dependence of the input and can more quickly and accurately find patterns of change from short-sequence sample data. Then, using a BP network for output prediction allows the calculated predicted data to have a certain complexity. The complexity of the changes can be seen from historical data, making this method more meaningful than other methods that predict single changes. This application is the first to propose an LSTM-BP network model framework for multi-objective water demand prediction in the field of hydrology and water resources.

[0079] Sample data was collected and organized to validate the above LSTM-BP model framework. Taking a city as an example, with the goal of calculating the water demand for agriculture, forestry, animal husbandry, fishery and livestock, industry, urban public utilities, domestic water, and ecological environment up to 2035, the key influencing factors of water demand for various industries in the city over the past ten years were collected and organized. Figure 6 and Figure 7 This demonstrates the training loss of the LSTM model. Figure 8 The training loss of the BP model is shown. Figure 9 The final prediction results of the LSTM-BP model are shown. Figure 9 The horizontal axis represents the water demand forecast data for each year from 2012 to 2035, where "12", "14", "16", "18", "20", "22", "24", "26", "28", "30", "32", and "34" represent 2012, 2014, 2016, 2018, 2020, 2022, 2024, 2026, 2028, 2030, 2032, and 2034, respectively.

[0080] This application proposes a neural network-based prediction method for water demand forecasting in various industries. The Spearman rank correlation coefficient is used to identify key influencing factors for the target industry's water demand. These selected key influencing factors are then used as input to train an LSTM (Long Short-Term Memory) neural network to predict the key influencing factor sequence for future years. Finally, a BP neural network is used to calculate the multi-objective future water demand. Experimental results show that the proposed method achieves excellent fitting performance on the training set and can provide a reference for water demand forecasting in various industries in future years.

[0081] This application also provides an application scenario in which the aforementioned LSTM-based multi-objective water demand forecasting method for future years is applied. Specifically, the LSTM-based multi-objective water demand forecasting method for future years provided in this embodiment can be applied to water demand forecasting scenarios in different industries. This scenario includes a data acquisition stage, a data prediction stage, and a prediction result display stage. The LSTM-based multi-objective water demand forecasting method for future years provided in this embodiment belongs to the data prediction stage. Specifically, in the video content processing chain, videos can be tagged using a collaborative method of machine tagging and manual tagging, that is, corresponding video tags can be added to the videos.

[0082] Based on the same inventive concept, this application also provides an LSTM-based multi-objective water demand forecasting device for implementing the aforementioned LSTM-based multi-objective water demand forecasting method for future years. The solution provided by this device is similar to the implementation described in the above method. Therefore, the specific limitations of one or more LSTM-based multi-objective water demand forecasting device embodiments provided below can be found in the limitations of the LSTM-based multi-objective water demand forecasting method for future years described above, and will not be repeated here.

[0083] In one exemplary embodiment, such as Figure 10 As shown, a multi-objective water demand prediction device based on LSTM for future years is provided, comprising:

[0084] The key influencing factor acquisition module M1 is used to acquire the key influencing factors of water demand in various industries.

[0085] The key impact factor prediction module M2 is used to input the known key impact factors of water demand for each industry into the corresponding trained LSTM model to obtain the prediction results of key impact factors for the future preset period of water demand for each industry.

[0086] The water demand prediction module M3 is used to input the prediction results of the key influencing factors corresponding to the water demand of each industry into the corresponding trained machine learning model to obtain the water demand prediction results for the future preset period corresponding to the water demand of each industry.

[0087] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 11As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores LSTM-based multi-objective water demand forecasting data for future years. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements an LSTM-based multi-objective water demand forecasting method for future years.

[0088] Those skilled in the art will understand that Figure 11 The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0089] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0090] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0091] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0092] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0093] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0094] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0095] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A multi-objective water demand forecasting method for future years based on LSTM, characterized in that, The LSTM-based multi-objective water demand forecasting method for future years includes: Identify the key influencing factors of water demand in various industries; The known key influencing factors of water demand for each industry are input into the corresponding trained LSTM model to obtain the prediction results of key influencing factors for water demand in the future preset period for each industry. The prediction results of the key influencing factors corresponding to the water demand of each industry are input into the corresponding trained machine learning model to obtain the water demand prediction results for the future preset period corresponding to the water demand of each industry; wherein, the machine learning model is a BP neural network trained using historical key influencing factors and corresponding historical actual water demand. The methods for obtaining the key influencing factors of water demand in various industries include: Obtain preliminary influencing factors for water demand in various industries; The Spearman rank correlation test was used to screen out the key influencing factors corresponding to the water demand of each industry from the initial selection of influencing factors, specifically: For each pair of variables related to water demand in each industry, all historical observations of the pair are sorted by magnitude; a preliminary influencing factor and its corresponding industry water demand constitute a pair of variables. For each pair of variables, calculate the rank difference for each set of historical observations of the pair based on the sorted position of the variable pair; For each pair of variables, the correlation between the pairs is calculated based on the rank difference of each set of historical observations for each pair of variables. Based on the comparison results of the correlation of each variable pair and the corresponding correlation threshold, the key influencing factors corresponding to the water demand of each industry are screened from the preliminary influencing factors.

2. The LSTM-based multi-objective water demand forecasting method for future years according to claim 1, characterized in that: The preliminary influencing factors include Gross National Product, total national population, total urban population, total rural population, cultivated land area, effective irrigated area, number of large livestock, number of small livestock, number of poultry, annual rainfall, and water resources.

3. The LSTM-based multi-objective water demand forecasting method for future years according to claim 1, characterized in that, The known key influencing factors of water demand for each industry are input into the corresponding trained LSTM model to obtain the prediction results of key influencing factors for water demand in the future preset period for each industry, specifically including: Obtain historical water demand data for each industry; the historical water demand data includes key historical influencing factors and corresponding historical actual water demand over several years; The historical key impact factors over several years are divided into multiple arrays; each array includes the historical key impact factors of the first historical period and the corresponding historical key impact factors of the second historical period; the time points of the first historical period are shorter than the time points of the second historical period. Using the historical key impact factors of the first historical period in each array as input, and the historical key impact factors of the corresponding second historical period as labels, train the corresponding LSTM model to obtain the trained LSTM model. The known key influencing factors of water demand for each industry are input into the corresponding trained LSTM model to obtain the prediction results of key influencing factors for the future preset period of water demand for each industry.

4. The LSTM-based multi-objective water demand forecasting method for future years according to claim 1, characterized in that, The predicted results of the key influencing factors corresponding to the water demand of each industry are input into the corresponding trained machine learning model to obtain the predicted water demand for the future preset period for each industry, specifically including: Obtain historical water demand data for each industry; the historical water demand data includes key historical influencing factors and corresponding historical actual water demand over several years; Using the historical key influencing factors of each year as input and the corresponding historical actual water demand as label, the corresponding machine learning model is trained to obtain the trained machine learning model. The prediction results of the key influencing factors corresponding to the water demand of each industry are input into the corresponding trained machine learning model to obtain the water demand prediction results for the future preset period corresponding to the water demand of each industry.

5. The LSTM-based multi-objective water demand forecasting method for future years according to claim 1, characterized in that, The specific water needs of various industries include: agricultural water needs, forestry, animal husbandry, fishery and livestock water needs, industrial water needs, urban public water needs, domestic water needs, and ecological environment water needs.

6. A multi-objective water demand prediction device for future years based on LSTM, characterized in that, This device is used to implement the LSTM-based multi-objective water demand forecasting method for future years as described in any one of claims 1-5: The key impact factor acquisition module is used to acquire the key impact factors of water demand in various industries; The key impact factor prediction module is used to input the known key impact factors of water demand in each industry into the corresponding trained LSTM model to obtain the prediction results of key impact factors for the future preset period of water demand in each industry. The water demand prediction module is used to input the prediction results of the key influencing factors corresponding to the water demand of each industry into the corresponding trained machine learning model to obtain the water demand prediction results for the future preset period corresponding to the water demand of each industry.

7. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the LSTM-based multi-objective water demand forecasting method for future years according to any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the LSTM-based multi-objective water demand forecasting method for future years as described in any one of claims 1-5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the LSTM-based multi-objective water demand forecasting method for future years as described in any one of claims 1-5.

Images