Wireless signal intensity intelligent multi-step prediction method based on WOA-CNN-GRU
By using the WOA-CNN-GRU model in wireless signal strength prediction and using whale optimization algorithm to optimize hyperparameters, the problem of low prediction accuracy in wireless signal strength in the existing technology is solved, and efficient and multi-efficient under different wireless network environments is achieved Step prediction.
Patent Information
- Application Number
- CN202411556214.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-04
- Publication Date
- 2025-02-11
AI Technical Summary
When predicting wireless signal strength, the prediction accuracy is not high, and the model structure is simple, so it cannot adapt to different wireless network environments, resulting in the inability to directly apply to signal strength prediction between different wireless networks.
Using the intelligent multi-step prediction method of wireless signal strength based on WOA-CNN-GRU, the multi-step prediction model of wireless signal strength of CNN-GRU is optimized through the whale optimization algorithm WOA, and a multi-step prediction model of wireless signal strength of WOA-CNN-GRU was established, and the multi-step prediction model of wireless signal strength of WOA-CNN-GRU was established through The time-sequential timing sequence data is trained and verified.
It realizes rapid convergence and verification of the rationality of the best hyperparameters in different application scenarios such as 4G, WiFi, and industrial networks, and improves the multi-step prediction performance of wireless signal strength, and shows better prediction effects and performance indicators.
Smart Images
Figure CN119420440A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of wireless communication technology, and in particular to a wireless signal strength intelligent multi-step prediction method based on WOA-CNN-GRU. Background Art
[0002] In recent years, communication and information technology have made remarkable achievements, especially in mobile communication services, multimedia information services and Internet applications. With the advancement of science and technology and the increasing demand for communication speed, quality and security, wireless communication networks are facing many challenges. Especially in complex environments, the strength of wireless signals is affected by many factors, such as building shielding, electromagnetic interference, etc., which leads to large signal fluctuations and affects the communication quality. Therefore, how to accurately predict the strength of wireless signals, optimize network layout and improve communication efficiency has become an urgent problem to be solved in the field of wireless communication.
[0003] Most existing research methods apply Markov models to the field of communications. Among them, the application of Markov models in the field of communications was pioneered by simulating binary channels based on two states; a wireless channel prediction model was established based on a finite-state Markov model to predict changes in signal strength; and the impact of wireless performance characteristics such as multipath fading and path loss on signal strength was captured based on Markov models. Although these models provide useful insights, the model structure is relatively simple and the prediction accuracy is poor. In addition, they are usually targeted at specific network environments and rely on factors such as sampling rate, mobility, and location, so they cannot be directly applied to signal strength prediction between different wireless networks.
[0004] With the deepening of research and the improvement of automation level, machine learning has been applied to various problems in the telecommunications field, such as network management and physical layer optimization. Deep learning has become one of the latest trends in the current field of machine learning due to its unique network architecture and powerful computing power, and has brought revolutionary progress in many fields such as engineering. With the increasing requirements for prediction accuracy, the research focus has gradually shifted to prediction methods based on deep learning. Among them, the method of automatically extracting features from wireless signals based on deep learning can achieve wireless positioning; deep learning can also effectively solve key problems in wireless channel estimation and improve both performance and efficiency; and the wireless channel estimation model based on learned denoising-based approximate message passing (LDAMP) network and generative adversarial network (GAN) can improve signal strength to varying degrees; the multi-dimensional GNN framework is used to learn wireless strategies, which can effectively avoid information loss; the application of neural network architecture to signal detection in MIMO systems and OFDM systems has demonstrated its superior performance in complex channel changes. Although the above models have improved the signal strength to varying degrees, they still have certain defects in predicting the variability of future signal strength. In addition, the above models can only be applied to fixed channels, and the generalization ability of the models has certain limitations. Summary of the invention
[0005] The technical problem to be solved by the present invention is to provide an intelligent multi-step prediction method for wireless signal strength based on WOA-CNN-GRU in view of the shortcomings of the above-mentioned prior art, aiming to predict future change trends based on historical wireless signal strength, so as to adjust relevant parameters in time and intervene in the weakening of wireless signals.
[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:
[0007] A wireless signal strength intelligent multi-step prediction method based on WOA-CNN-GRU, the method comprising the following steps:
[0008] Step 1: Perform time series processing on the collected wireless signal strength data to form time series data;
[0009] Step 2: Construct a CNN-GRU wireless signal strength multi-step prediction model for different application scenarios, use the whale optimization algorithm WOA to optimize the CNN-GRU wireless signal strength multi-step prediction model, establish a WOA-CNN-GRU wireless signal strength multi-step prediction model, and train and verify the WOA-CNN-GRU wireless signal strength multi-step prediction model based on the time series data after time series;
[0010] Step 2.1: Construct a CNN-GRU wireless signal strength multi-step prediction model;
[0011] Step 2.2: Use the whale optimization algorithm WOA to globally optimize multiple hyperparameters of the CNN-GRU wireless signal strength multi-step prediction model to obtain the optimal hyperparameters, and establish the WOA-CNN-GRU wireless signal strength multi-step prediction model based on the optimal hyperparameters;
[0012] Step 2.3: Train and verify the WOA-CNN-GRU wireless signal strength multi-step prediction model;
[0013] Step 3: The proposed WOA-CNN-GRU wireless signal strength multi-step prediction model is analyzed and verified from three different perspectives: prediction effect, performance indicators, and loss function and prediction performance superiority in different application scenarios.
[0014] Furthermore, each time series sequence after time series in step 1 is composed of data within a period of time and is used to predict the wireless signal strength of the next step.
[0015] Furthermore, the application scenarios in step 2 include 4G, industrial networks and WiFi. The CNN-GRU wireless signal strength multi-step prediction model structure in different application scenarios is the same, and the difference lies in the convolution kernel dimension, the number of convolution kernels, the number of iterations, and the number of neurons.
[0016] Furthermore, the whale optimization algorithm WOA in step 2 is an intelligent optimization algorithm that simulates the behavior of humpback whales, which is divided into three stages: encircling prey, bubble net predation and searching for prey.
[0017] Furthermore, the CNN-GRU wireless signal strength multi-step prediction model in step 2.1 is composed of five parts: an input layer, a CNN layer, a GRU1 layer, a GRU2 layer and an output layer, wherein the input layer is used to receive one-dimensional wireless signal strength data, the CNN layer is used to extract features of the one-dimensional wireless signal strength data received by the input layer, the GRU1 layer and the GRU2 layer are used to process time series and capture the time dependency in the time series data, and the output layer is used to output the predicted value of the wireless signal strength at the next moment.
[0018] Furthermore, the hyperparameters in step 2.2 include the learning rate, number of iterations, number of convolution kernels, dimension of convolution kernels, number of neurons in GRU1 layer, and number of neurons in GRU2 layer of the CNN-GRU wireless signal strength multi-step prediction model.
[0019] Furthermore, the specific training process in step 2.3 is as follows:
[0020] (1) Processing the collected wireless signal strength data in a time series manner;
[0021] (2) Divide the time-series data into training set and test set in a ratio of 7:3;
[0022] (3) Define the mean square error as the fitness function to evaluate the optimal solution found during the entire optimization process; calculate the fitness value and find the optimal position of the current whale individual;
[0023] (4) Determine whether the predation probability p is less than the set threshold. If so, directly proceed to (5). Otherwise, update the position information based on the position update formula of the bubble net predation stage and enter the bubble net predation stage.
[0024] (5) Determine the coefficient vector in the calculation formula for the position vector of the best individual Is it less than 1? If so, the system enters the stage of swarming prey and uses the position update formula of the prey stage to update the position information. Otherwise, the system updates the position information based on the position update formula of the prey search stage and enters the stage of searching prey.
[0025] (6) After the position information is updated, the fitness value is calculated and compared with the fitness value of the initial optimal position information, and the optimal individual position with the smaller fitness value is selected;
[0026] (7) Determine whether the current number of iterations has reached the maximum number of iterations. If so, obtain the best individual position and the optimization ends. Otherwise, continue to the next iteration and return to execute (2), (3) and (4) in sequence, input the six hyperparameters of the CNN-GRU wireless signal strength multi-step prediction model, and obtain the optimal hyperparameter value.
[0027] (8) Based on the training set, the WOA-CNN-GRU wireless signal strength multi-step prediction model is trained by continuously updating weights and biases;
[0028] (9) The training of the WOA-CNN-GRU wireless signal strength multi-step prediction model is completed;
[0029] (10) The trained WOA-CNN-GRU wireless signal strength multi-step prediction model is verified based on the test set to obtain the wireless signal strength for the next step.
[0030] The beneficial effects of adopting the above technical scheme are: a multi-step prediction method of wireless signal strength based on WOA-CNN-GRU provided by the present invention searches for the optimal hyperparameters of the CNN-GRU wireless signal strength multi-step prediction model in various application scenarios through the whale optimization algorithm WOA, constructs the WOA-CNN-GRU wireless signal strength multi-step prediction model, and observes the loss function and performance indicators. The results show that the WOA-CNN-GRU wireless signal strength multi-step prediction model can converge quickly during the training process, verifying the rationality of the optimal hyperparameters, and intuitively evaluating the prediction performance of the model; secondly, the present invention simulates the application of this method in three scenarios of 4G, WiFi, and industrial network. The results show that the WOA-CNN-GRU wireless signal strength multi-step prediction model has good prediction performance in different scenarios; finally, by performing ablation experiments, the WOA-CNN-GRU wireless signal strength multi-step prediction model is compared with MLP, GRU, and CNN-GRU in terms of prediction effect and performance indicators. The results show that the prediction effect and performance indicators of the WOA-CNN-GRU wireless signal strength multi-step prediction model have better superiority. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 A flowchart of a multi-step prediction method for wireless signal strength based on WOA-CNN-GRU provided in an embodiment of the present invention;
[0032] Figure 2 A schematic diagram of the timing results of an embodiment of the present invention;
[0033] Figure 3 It is a structural diagram of a CNN-GRU wireless signal strength multi-step prediction model according to an embodiment of the present invention;
[0034] Figure 4 It is an optimization flow chart of the whale optimization algorithm WOA according to an embodiment of the present invention;
[0035] Figure 5 The present invention is a training flowchart of the WOA-CNN-GRU wireless signal strength multi-step prediction model. DETAILED DESCRIPTION
[0036] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.
[0037] The WOA-CNN-GRU wireless signal strength multi-step prediction model can predict future change trends based on historical wireless signal strength, thereby adjusting relevant parameters in a timely manner and intervening in the weakening of wireless signals.
[0038] In this embodiment, a wireless signal strength intelligent multi-step prediction method based on WOA-CNN-GRU is provided. Figure 1 As shown, the specific steps include:
[0039] Step 1: Perform time series processing on the collected wireless signal strength data to form new time series data;
[0040] The Gated Recurrent Unit (GRU) model is widely used in various sequence modeling tasks, and has high requirements for input features. Since the time series is consistent with the recursive structure of the GRU model, the collected wireless signal strength data is processed in a time series manner so that the GRU model can better capture the time dependency in the data. Based on this, this embodiment processes the collected wireless signal strength data in a time series manner and converts it into the input form required by the GRU model, that is, the time series after time series processing. The time series result is as follows: Figure 2 As shown;
[0041] Each time series sequence after time series is composed of data within a period of time, and is used to predict the wireless signal strength of the next step; this embodiment is described by taking the first three time series sequences as examples, time series sequence 1 is composed of wireless signal strengths at time 1-11, and is used to predict the wireless signal strength at time 12; time series sequence 2 is composed of wireless signal strengths at time 2-12, and is used to predict the wireless signal strength at time 13; time series sequence 3 is composed of wireless signal strengths at time 3-13, and is used to predict the wireless signal strength at time 14; based on this rolling method, time series sequences are established in sequence until all training data are established, and the formula for the prediction process is as follows:
[0042] Y(k)=f(Y(k-11), Y(k-10), Y(k-9),…, Y(k-1))(1)
[0043] Wherein, Y(k) is the predicted value of wireless signal strength at time k, Y(k-11), Y(k-10), Y(k-9), …, Y(k-1) are the historical wireless signal strength values in the time series data, and f() represents the CNN-GRU wireless signal strength multi-step prediction model;
[0044] Step 2: Construct a CNN-GRU wireless signal strength multi-step prediction model for different application scenarios, and use the Whale Optimization Algorithm (WOA) to optimize the CNN-GRU wireless signal strength multi-step prediction model;
[0045] In this embodiment, the whale optimization algorithm WOA is used to optimize the six hyperparameters of the CNN-GRU wireless signal strength multi-step prediction model, namely, the learning rate, the number of iterations, the number of convolution kernels, the convolution kernel dimension, the number of neurons in the GRU1 layer, and the number of neurons in the GRU2 layer, and the optimal hyperparameters are selected. Based on the above six optimal hyperparameters, a WOA-CNN-GRU wireless signal strength multi-step prediction model is established to better capture the feature relationship between the data, so as to maximize the prediction performance of the WOA-CNN-GRU wireless signal strength multi-step prediction model in various application scenarios;
[0046] The application scenarios include 4G, industrial networks and WiFi. The CNN-GRU wireless signal strength multi-step prediction models in different application scenarios are represented by WOA-CNN-GRU1, WOA-CNN-GRU2 and WOA-CNN-GRU3 respectively. The CNN-GRU wireless signal strength multi-step prediction models in different application scenarios have the same structure, except that the convolution kernel dimension, the number of convolution kernels, the number of iterations and the number of neurons are different, so as to better complete the prediction tasks in different application scenarios;
[0047] Step 2.1: Construct a CNN-GRU wireless signal strength multi-step prediction model; the CNN-GRU wireless signal strength multi-step prediction model consists of five parts: input layer, CNN layer, GRU1 layer, GRU2 layer and output layer. The specific network structure is as follows: Figure 3 As shown, the input layer is used to receive one-dimensional wireless signal strength data, the CNN layer is used to extract features from the one-dimensional wireless signal strength data received by the input layer, the GRU1 layer and the GRU2 layer are used to process time series and capture the time dependency in the time series data, and the output layer is used to output the predicted value of the wireless signal strength at the next moment;
[0048] Step 2.2: Use the whale optimization algorithm WOA to globally optimize the six hyperparameters of the CNN-GRU wireless signal strength multi-step prediction model, including the learning rate, number of iterations, number of convolution kernels, convolution kernel dimension, number of neurons in the GRU1 layer, and number of neurons in the GRU2 layer, to obtain the optimal hyperparameters, and build the WOA-CNN-GRU wireless signal strength multi-step prediction model based on the optimal hyperparameters;
[0049] The Whale Optimization Algorithm (WOA) is an intelligent optimization algorithm that simulates the behavior of humpback whales. It is mainly divided into three stages: encircling prey, bubble net predation, and searching for prey.
[0050] Step 2.2.1: The prey-catching phase, that is, updating the position of the individual based on the current position relationship between the individual and the prey, and obtaining the position vector of the best individual in the next iteration, the formula is as follows:
[0051]
[0052] in, is the distance between the current individual and the best individual, is the position vector of the current individual, i is the number of iterations, is the position vector of the current best individual, is the position vector of the best individual at the next iteration, represents the coefficient vector, Represents two random numbers in the range 0-1, It is a value that decreases linearly from 2 to 0;
[0053] Step 2.2.2: Bubble net predation stage, that is, the process of humpback whale spiraling, simulate its spiral motion state by establishing the spiral equation between the whale and the prey, and update the position vector of the best individual in the next iteration. The formula is as follows:
[0054]
[0055] in, Represents the distance between the current search individual and the current best individual, represents the position vector of the best individual in the next iteration after the update, l is a random number uniformly distributed in the interval [-1,1], b is a constant, generally set to 1, p is the predation probability, which is a random number in the interval [0,1], and e bl is a parameter that changes with the number of iterations and is used to adjust the search range and convergence. Represents the position vector of the best individual in the next iteration after updating again;
[0056] Step 2.2.3: Prey search phase, that is, updating the position vector of the best individual in the next iteration by randomly searching the positions of other individuals. The formula is as follows:
[0057]
[0058] in, Represents the position vector of the best individual at the next iteration after the third update, is the relationship between the current search individual and the random individual, is the position vector of a random individual;
[0059] In this embodiment, based on the whale optimization algorithm WOA, the number of convolution kernels, the dimension of the convolution kernel, the number of neurons in the GRU1 layer, the number of neurons in the GRU2 layer, the learning rate, and the number of iterations of the CNN-GRU wireless signal strength multi-step prediction model in different application scenarios are globally optimized. The search ranges are: [1,100], [1,10], [1,100], [1,100], [0.001,0.01], [10,100], and the hyperparameter optimization process is as follows: Figure 4 As shown;
[0060] In the optimization process, the mean squared error (MSE) is defined as the fitness function to evaluate the individuals searched during the entire optimization process. First, a group of individuals are randomly initialized. In each iteration, the position is updated according to the random search or the best individual obtained so far. Secondly, it is determined whether the predation probability p is less than 0.5. If it is not less than 0.5, the bubble net predation stage is carried out, and the individual position is updated using formulas (6) and (7). If it is less than 0.5, the other two stages are directly determined. When , the prey hunting phase is carried out, and the individual positions are updated using formulas (2) and (3); When , the prey search phase is carried out, and the individual position is updated using formulas (9) and (10); finally, it is determined whether the maximum number of iterations has been reached. If it has been reached, the optimal hyperparameter with the smallest fitness value is output, otherwise the iteration continues. The optimal hyperparameters of the CNN-GRU wireless signal strength multi-step prediction model under various application scenarios are shown in Table 1, where WOA-CNN-GRU1, WOA-CNN-GRU2 and WOA-CNN-GRU3 represent the application scenarios of 4G, industrial network and WiFi respectively;
[0061] Table 1 Optimal hyperparameters of the CNN-GRU wireless signal strength multi-step prediction model in various application scenarios
[0062]
[0063] Step 2.3: Train and verify the WOA-CNN-GRU wireless signal strength multi-step prediction model;
[0064] After determining the optimal hyperparameters, we started training the WOA-CNN-GRU wireless signal strength multi-step prediction model. The training process is as follows: Figure 5As shown, first, the time series data after time series is divided into a training set and a test set in a ratio of 7:3, and the optimal hyperparameters of the CNN-GRU wireless signal strength multi-step prediction model under various application scenarios are searched globally based on the whale optimization algorithm WOA, and the weights and biases are continuously updated on the training set for iteration, thereby completing the training process of the WOA-CNN-GRU wireless signal strength multi-step prediction model under different application scenarios; after the training is completed, the trained WOA-CNN-GRU wireless signal strength multi-step prediction model is verified based on the divided test set, and the prediction performance of the WOA-CNN-GRU wireless signal strength multi-step prediction model is verified by observing and comparing the degree of fit between the actual curve and the predicted curve, thereby realizing the multi-step prediction of the wireless signal strength under various application scenarios;
[0065] The specific training process is as follows:
[0066] (1) Processing the collected wireless signal strength data in a time series manner;
[0067] (2) Divide the time-series data into training set and test set in a ratio of 7:3;
[0068] (3) Calculate the fitness value and find the best position of the current whale individual;
[0069] (4) Determine whether the predation probability p is less than 0.5. If so, directly proceed to (5); otherwise, update the position information based on formula (8) and enter the bubble net predation stage;
[0070] (5)Judgment Is it less than 1? If yes, enter the stage of hunting prey and update the position information based on formula (3); otherwise, update the position information based on formula (10) and enter the stage of searching prey;
[0071] (6) After the position information is updated, the fitness value is calculated and compared with the fitness value of the initial optimal position information, and the optimal individual position with the smaller fitness value is selected;
[0072] (7) Determine whether the current number of iterations has reached the maximum number of iterations. If so, obtain the best individual position and end the optimization search. Otherwise, continue with the next iteration and return to execute (2), (3) and (4) in sequence, input the six hyperparameters of the CNN-GRU wireless signal strength multi-step prediction model, and obtain the optimal hyperparameter value.
[0073] (8) Based on the training set, the WOA-CNN-GRU wireless signal strength multi-step prediction model is trained by continuously updating weights and biases;
[0074] (9) The training of the WOA-CNN-GRU wireless signal strength multi-step prediction model is completed;
[0075] (10) The trained WOA-CNN-GRU wireless signal strength multi-step prediction model is verified based on the test set to obtain the wireless signal strength for the next step;
[0076] Step 3: Analyze and verify the proposed WOA-CNN-GRU wireless signal strength multi-step prediction model from three different perspectives: prediction effect, performance indicators, and loss function and prediction performance superiority in different application scenarios;
[0077] This embodiment is experimented in three application scenarios: 4G network, WiFi network and industrial network. The experimental results show that the WOA-CNN-GRU wireless signal strength multi-step prediction model proposed in the present invention has good prediction performance in different scenarios. At the same time, through ablation experiments, the WOA-CNN-GRU wireless signal strength multi-step prediction model is compared with MLP, GRU and CNN-GRU in terms of prediction effect and performance indicators. The results show that the prediction effect and performance indicators of the WOA-CNN-GRU wireless signal strength multi-step prediction model are superior.
[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.
Claims
1. A wireless signal strength intelligent multi-step prediction method based on WOA-CNN-GRU, characterized in that: The following steps are involved: Step 1: Perform time series processing on the collected wireless signal strength data to form time series data; Step 2: Construct a CNN-GRU wireless signal strength multi-step prediction model for different application scenarios, use the whale optimization algorithm WOA to optimize the CNN-GRU wireless signal strength multi-step prediction model, establish a CNN-GRU wireless signal strength multi-step prediction model based on WOA, and train and verify the CNN-GRU wireless signal strength multi-step prediction model based on WOA according to the time series data after time series; Step 3: The proposed WOA-CNN-GRU wireless signal strength multi-step prediction model is analyzed and verified from three different perspectives: prediction effect, performance indicators, and loss function and prediction performance superiority in different application scenarios.
2. According to claim 1, a wireless signal strength intelligent multi-step prediction method based on WOA-CNN-GRU is characterized in that: Each of the time-series sequences in step 1 is composed of data within a period of time and is used to predict the wireless signal strength of the next step.
3. According to the WOA-CNN-GRU-based wireless signal strength intelligent multi-step prediction method according to claim 1, it is characterized in that: The application scenarios in step 2 include 4G, industrial networks and WiFi. The CNN-GRU wireless signal strength multi-step prediction model structure in different application scenarios is the same, and the difference lies in the convolution kernel dimension, the number of convolution kernels, the number of iterations, and the number of neurons.
4. According to claim 1, a wireless signal strength intelligent multi-step prediction method based on WOA-CNN-GRU is characterized in that: The whale optimization algorithm WOA in step 2 is an intelligent optimization algorithm that simulates the behavior of humpback whales, which is divided into three stages: encircling prey, bubble net predation and searching for prey.
5. According to the WOA-CNN-GRU-based wireless signal strength intelligent multi-step prediction method according to claim 1, it is characterized in that: The step 2 specifically includes the following steps: Step 2.1: Construct a CNN-GRU wireless signal strength multi-step prediction model; Step 2.2: Use the whale optimization algorithm WOA to globally optimize multiple hyperparameters of the CNN-GRU wireless signal strength multi-step prediction model to obtain the optimal hyperparameters, and establish the WOA-CNN-GRU wireless signal strength multi-step prediction model based on the optimal hyperparameters; Step 2.3: Train and verify the WOA-CNN-GRU wireless signal strength multi-step prediction model.
6. According to claim 4, a wireless signal strength intelligent multi-step prediction method based on WOA-CNN-GRU is characterized in that: The CNN-GRU wireless signal strength multi-step prediction model in step 2.1 consists of five parts: an input layer, a CNN layer, a GRU1 layer, a GRU2 layer and an output layer. The input layer is used to receive one-dimensional wireless signal strength data, the CNN layer is used to extract features from the one-dimensional wireless signal strength data received by the input layer, the GRU1 layer and the GRU2 layer are used to process time series and capture the time dependency in the time series data, and the output layer is used to output the predicted value of the wireless signal strength at the next moment.
7. According to claim 4, a wireless signal strength intelligent multi-step prediction method based on WOA-CNN-GRU is characterized in that: The hyperparameters in step 2.2 include the learning rate, number of iterations, number of convolution kernels, dimension of convolution kernels, number of neurons in GRU1 layer, and number of neurons in GRU2 layer of the CNN-GRU wireless signal strength multi-step prediction model.
8. According to claim 4, a wireless signal strength intelligent multi-step prediction method based on WOA-CNN-GRU is characterized in that: The specific training process in step 2.3 is as follows: (1) Processing the collected wireless signal strength data in a time series manner; (2) Divide the time-series data into training set and test set in a ratio of 7:3; (3) Define the mean square error as the fitness function to evaluate the optimal solution found during the entire optimization process; calculate the fitness value and find the optimal position of the current whale individual; (4) Determine whether the predation probability p is less than the set threshold. If so, directly proceed to (5). Otherwise, update the position information based on the position update formula of the bubble net predation stage and enter the bubble net predation stage. (5) Determine the coefficient vector in the calculation formula for the position vector of the best individual Is it less than 1? If so, the system enters the stage of swarming prey and uses the position update formula of the prey stage to update the position information. Otherwise, the system updates the position information based on the position update formula of the prey search stage and enters the stage of searching prey. (6) After the position information is updated, the fitness value is calculated and compared with the fitness value of the initial optimal position information, and the optimal individual position with the smaller fitness value is selected; (7) Determine whether the current number of iterations has reached the maximum number of iterations. If so, obtain the best individual position and the optimization ends. Otherwise, continue to the next iteration and return to execute (2), (3) and (4) in sequence, input the six hyperparameters of the CNN-GRU wireless signal strength multi-step prediction model, and obtain the optimal hyperparameter value. (8) Based on the training set, the WOA-CNN-GRU wireless signal strength multi-step prediction model is trained by continuously updating weights and biases; (9) The training of the WOA-CNN-GRU wireless signal strength multi-step prediction model is completed; (10) The trained WOA-CNN-GRU wireless signal strength multi-step prediction model is verified based on the test set to obtain the wireless signal strength for the next step.