Wave energy power generation power prediction method, device, equipment and medium

Abstract

The present disclosure relates to a method, device, equipment and medium for predicting wave power generation. The method comprises: obtaining the measured power of wave power generation in the area to be predicted; using a finely tuned wave power generation prediction model to process the measured power of wave power generation in the area to be predicted, and determining the predicted power of wave power generation in the area to be predicted; wherein the wave power generation prediction model uses the wave height data of the reference area to pre-train the initial neural network, and uses the measured power of wave power generation in the reference area to fine-tune the obtained pre-trained model, and the wave height data is positively correlated with the measured power of wave power generation in the reference area, and the relationship value of the positive correlation is greater than the preset positive correlation value. Thus, publicly available wave height data and a small amount of measured power of wave power generation are used for pre-training and fine-tuning respectively, so as to realize accurate prediction of wave power generation using a small amount of measured power of wave power generation.

Description

Technical Field

[0001] The present disclosure relates to the technical field of power prediction, and in particular to a method, device, equipment and medium for predicting wave energy power generation. Background Art

[0002] Wave energy is a renewable energy source with a wide distribution range and high energy density. Wave power generation technology can convert the kinetic energy in waves into electrical energy, and is a clean and environmentally friendly way of generating electricity. Since wave energy is affected by complex factors such as ocean currents and climate, the output of wave power generation is unstable, which will have an impact on the power system. Therefore, it is necessary to predict the power of wave power generation so that wave power generation is predictable, which is convenient for wave power generation to participate in the dispatch of the power system, thereby promoting the development and utilization of wave energy.

[0003] At present, due to the small number of wave power generation devices in operation and the lack of measured power data of wave power generation, it is difficult to predict the power of wave power generation. In the absence of measured power data, it is a technical problem that urgently needs to be solved to propose a wave power prediction with high accuracy. Summary of the invention

[0004] In order to solve the above technical problems, the present disclosure provides a wave energy power prediction method, device, equipment and medium.

[0005] In a first aspect, the present disclosure provides a wave energy power prediction method, comprising:

[0006] Obtain the measured power of wave power generation in the area to be predicted;

[0007] Using the fine-tuned wave power prediction model, the measured wave power of the area to be predicted is processed to determine the predicted wave power of the area to be predicted;

[0008] The wave power prediction model uses wave height data of a reference area to pre-train an initial neural network, and uses the measured wave power of the reference area to fine-tune the obtained pre-trained model, wherein the wave height data is positively correlated with the measured wave power of the reference area, and the relationship value of the positive correlation is greater than a preset positive correlation value.

[0009] In a second aspect, the present disclosure provides a wave energy power prediction device, comprising:

[0010] An acquisition module is used to obtain the actual wave power generation power in the area to be predicted;

[0011] A prediction module, used to process the measured wave power of the area to be predicted by using the fine-tuned wave power prediction model, and determine the predicted wave power of the area to be predicted;

[0012] The wave power prediction model uses wave height data of a reference area to pre-train an initial neural network, and uses the measured wave power of the reference area to fine-tune the obtained pre-trained model, wherein the wave height data is positively correlated with the measured wave power of the reference area, and the relationship value of the positive correlation is greater than a preset positive correlation value.

[0013] In a third aspect, an embodiment of the present disclosure further provides an electronic device, including:

[0014] processor;

[0015] A memory for storing executable instructions;

[0016] The processor is used to read the executable instructions from the memory and execute the executable instructions to implement the method described in the first aspect.

[0017] In a fourth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, which implements the method provided in the first aspect when the computer program is executed by a processor.

[0018] Compared with the prior art, the technical solution provided by the embodiments of the present disclosure has the following advantages:

[0019] A wave energy power generation prediction method, device, equipment and medium of the disclosed embodiment include: obtaining the measured power of wave power generation in the area to be predicted; using a fine-tuned wave power generation prediction model to process the measured power of wave power generation in the area to be predicted, and determine the predicted power of wave power generation in the area to be predicted; wherein the wave power generation prediction model uses the wave height data of the reference area to pre-train the initial neural network, and uses the measured power of wave power generation in the reference area to fine-tune the obtained pre-trained model, and the wave height data is positively correlated with the measured power of wave power generation in the reference area, and the relationship value of the positive correlation is greater than the preset positive correlation value. Therefore, using publicly available wave height data for model pre-training can solve the current problem of lack of measured power generation, and quickly fine-tune the pre-trained model based on a small amount of measured power of wave power generation to adapt to the wave power prediction scenario. In this way, accurate prediction of wave power generation is achieved using a small amount of measured power of wave power generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0022] Figure 1 A schematic diagram of a flow chart of a wave energy power generation prediction method provided in an embodiment of the present disclosure;

[0023] Figure 2 A schematic flow chart of a pre-training method and a fine-tuning method for a wave power prediction model provided in an embodiment of the present disclosure;

[0024] Figure 3 A schematic diagram of the structure of a wave energy power generation prediction device provided by an embodiment of the present disclosure;

[0025] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0026] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features in the embodiments can be combined with each other without conflict.

[0027] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments.

[0028] Faced with the lack of measured power data, related technologies have proposed two methods for wave power generation prediction. The first method is to directly use wave parameters combined with the power matrix of wave power generation to fit the theoretical power generation, and then use deep learning models such as gated cyclic units to conduct power prediction research. Although this strategy can solve the problem of shortage of measured data, the theoretical wave power generation power is very different from the actual wave power generation power, and it cannot directly guide the use of deep learning methods for prediction research. The second method is to run the wave power generation device for a period of time, and then use deep learning methods for prediction research after accumulating more measured data. Since the deep learning model requires more data for training, this method will take a long time to accumulate data, which is not conducive to the timely grid-connected power generation of the power generation device.

[0029] In order to solve the above problems, the embodiments of the present disclosure provide a wave energy power prediction method, device, equipment and medium.

[0030] Combine the following Figure 1 The wave energy power generation prediction method provided by the embodiment of the present disclosure is described. In the embodiment of the present disclosure, the wave energy power generation prediction method can be executed by an electronic device or a server. The electronic device may include a device with a communication function such as a tablet computer, a desktop computer, a laptop computer, or a device simulated by a virtual machine or a simulator. The server may include a server cluster and a cloud server. Below, the wave energy power generation prediction method is specifically explained with the electronic device as the execution subject.

[0031] Figure 1 A schematic flow chart of a wave energy power generation prediction method provided in an embodiment of the present disclosure is shown.

[0032] like Figure 1 As shown, the wave energy power generation prediction method may include the following steps.

[0033] S110, obtaining the actually measured power of wave power generation in the area to be predicted.

[0034] In this embodiment, when it is necessary to predict the wave power generation in any area (i.e., the area to be predicted), a small amount of measured power data of the area to be predicted is obtained as the wave power generation measured power, so that the electronic device can predict the wave power generation in the area to be predicted based on the wave power generation measured power.

[0035] A small number of wave power generation devices may be deployed in the area to be predicted, and a small amount of wave power generation measured power may be collected by using the small number of wave power generation devices.

[0036] The measured power of wave power generation can be understood as the historical measured power before the current time, that is, the measured power data.

[0037] S120, using the fine-tuned wave power prediction model, processing the actually measured wave power of the prediction area to determine the predicted wave power of the prediction area, wherein the wave power prediction model uses the wave height data of the reference area to pre-train the initial neural network, and uses the actually measured wave power of the reference area to fine-tune the obtained pre-trained model to determine that the wave height data is positively correlated with the actually measured wave power of the reference area, and the relationship value of the positive correlation is greater than a preset positive correlation value.

[0038] It is understandable that since the wave height data is positively correlated with the measured power of wave power generation in the reference area, and the relationship value of the positive correlation is greater than the preset positive correlation value. In other words, the wave height data and the measured power of wave power generation are strongly positively correlated, and the wave height data contains information such as the measured power of wave power generation. Using wave height data to participate in model pre-training can enable the model to perform better in the fine-tuning stage, that is, it can reflect the relationship between the measured power of wave power generation and the predicted power of wave power generation. In addition, the amount of wave height data is abundant and easy to obtain. In this way, based on the wave height data and the measured power of wave power generation, a pre-trained model that conforms to the law of wave energy power generation prediction can be obtained.

[0039] Furthermore, a small amount of measured wave power in the reference area is used to fine-tune the pre-trained model to obtain a fine-tuned wave power prediction model, thereby achieving accurate wave power prediction in scenarios with insufficient data.

[0040] The reference area is an area with similar topographic conditions to the area to be predicted, and a small number of wave power generation devices are deployed in the reference area, so only a small amount of wave power generation measured power can be collected.

[0041] The preset positive correlation value refers to the minimum correlation value that can characterize the strong correlation between the wave height data and the measured power of wave power generation.

[0042] Among them, the predicted power of wave power generation can be understood as the predicted power after the current time, which can be specifically used to characterize the predictable power of the wave power generation device, which is used to participate in the power system scheduling, thereby promoting the development and utilization of wave energy.

[0043] In this embodiment, the fine-tuned wave power prediction model includes the first linear layer at the beginning and the end and the nonlinear layer in the middle. Accordingly, the specific implementation method of S120 includes but is not limited to the following methods: based on the first linear layer in the fine-tuned wave power prediction model, the wave power measured power in the prediction area is processed by dimensionality increase to obtain the wave power measured power in the first dimension; based on the nonlinear layer in the fine-tuned wave power prediction model, the wave power measured power in the first dimension is processed by time series feature reasoning to predict the power reasoning feature in the first dimension; based on the second linear layer in the fine-tuned wave power prediction model, the power reasoning feature in the first dimension is processed by dimensionality reduction to obtain the power prediction feature in the second dimension, and the power prediction feature in the second dimension is used as the wave power prediction power in the prediction area.

[0044] Among them, the first linear layer and the second linear layer can be fully connected layers, and the nonlinear layer can be a recurrent neural network (RNN for short), and specifically, a long short-term memory network (LSTM), a gated recurrent unit (GRU), etc. can be selected.

[0045] Therefore, the fine-tuned wave power prediction model uses a combination of linear and nonlinear layers to predict wave power generation, which can improve the prediction accuracy of wave power generation.

[0046] In this embodiment, it is considered that the measured power of wave power generation in the area to be predicted often has power missing problems, and there are data differences in the measured power of wave power generation in the area to be predicted, which affects the prediction accuracy of wave power generation to a certain extent. In order to improve the prediction accuracy of wave power generation, the specific implementation method of S120 includes but is not limited to the following methods: identifying the position of the missing power from the measured power of wave power generation in the area to be predicted; based on the measured power of wave power generation before the position of the missing power and the measured power of wave power generation after the position of the missing power, the missing power is supplemented to obtain the corrected measured power of wave power generation; the corrected measured power of wave power generation is normalized to obtain the normalized measured power; the normalized measured power is processed using the fine-tuned wave power generation power prediction model to determine the predicted power of wave power generation in the area to be predicted.

[0047] Specifically, for the missing power in the wave power generation measured power, the average value of the wave power generation measured power before the position of the missing power and the wave power generation measured power after the position of the missing power is calculated to make up for the missing power and obtain the corrected wave power generation measured power. Then, the corrected wave power generation measured power is normalized by using but not limited to the maximum and minimum value normalization method, and the corrected wave power generation measured power is controlled to be scaled between [0,1] to obtain the normalized measured power. Then, the normalized measured power is input into the fine-tuned wave power generation prediction model to perform wave power generation prediction, thereby obtaining the wave power generation predicted power of the area to be predicted.

[0048] It should be noted that the fine-tuned wave power prediction model can specifically use a combination of linear layers and nonlinear layers to predict the normalized measured power and obtain the predicted wave power in the area to be predicted. The specific implementation method can refer to the above process and will not be repeated here.

[0049] In this way, before using the fine-tuned wave power prediction model to predict wave power, the missing values ​​in the measured wave power are filled with the mean, and then normalized to eliminate the problems of missing data or data differences that affect the prediction accuracy of wave power to a certain extent, thereby further improving the accuracy of wave power prediction in the area to be predicted.

[0050] The disclosed embodiment provides a method for predicting wave power generation, which obtains the measured power of wave power generation in the area to be predicted; uses a fine-tuned wave power generation prediction model to process the measured power of wave power generation in the area to be predicted, and determines the predicted power of wave power generation in the area to be predicted; wherein the wave power generation prediction model uses the wave height data of the reference area to pre-train the initial neural network, and uses the measured power of wave power generation in the reference area to fine-tune the obtained pre-trained model, and the wave height data is positively correlated with the measured power of wave power generation in the reference area, and the relationship value of the positive correlation is greater than the preset positive correlation value. Therefore, using publicly available wave height data for model pre-training can solve the current problem of lack of measured power generation, and quickly fine-tune the pre-trained model based on a small amount of measured power of wave power generation to adapt to the wave power prediction scenario. In this way, accurate prediction of wave power generation is achieved using a small amount of measured power of wave power generation.

[0051] In another embodiment of the present disclosure, the pre-training process and fine-tuning process of the wave power prediction model are explained in detail.

[0052] Figure 2A schematic flow chart of a pre-training method and a fine-tuning method for a wave power prediction model provided by an embodiment of the present disclosure is shown.

[0053] like Figure 2 As shown, the pre-training method and fine-tuning method of the wave power prediction model may specifically include the following steps.

[0054] S210, obtaining a first number of first pre-trained sample sets and a first number of second pre-trained sample sets from the wave height data, and copying the initial neural network to obtain a first number of copy models, wherein the first pre-trained sample set and the second pre-trained sample set contain different wave height data.

[0055] In this embodiment, before performing model pre-training and fine-tuning, the electronic device obtains public buoy data from a public platform, and extracts time series data of wave height from the buoy data as wave height data, and then fills in missing values ​​for the wave height data, specifically based on the average value of the wave height data before the position of the missing wave height data and the wave height data after the position of the missing wave height data, to fill in the missing wave height data, thereby obtaining corrected wave height data, and then, using but not limited to the maximum and minimum value normalization method, normalize the corrected wave height data, control the corrected wave height data to be scaled between [0,1], and obtain normalized wave height data.

[0056] Furthermore, a pre-trained data set is prepared based on the normalized wave height data. Specifically, a first number of sample sets are randomly selected from the normalized wave height data and shuffled in two different ways to obtain a first number of first pre-trained sample sets I and a first number of second pre-trained sample sets II. At the same time, the initial neural network is copied to obtain a first number of copied models. It can be understood that since the first number of first pre-trained sample sets I and the first number of second pre-trained sample sets II are shuffled in two different ways, the specific data included in the first number of first pre-trained sample sets I and the first number of second pre-trained sample sets II are different.

[0057] S220. For each pre-training number, based on the first number of first pre-training sample sets and the first number of second pre-training sample sets, update the model parameters of the first number of replication models, and determine the loss values ​​corresponding to the first number of replication models respectively.

[0058] In this embodiment, the electronic device can learn the wave height variation characteristics of different buoy positions based on the meta-learning strategy, thereby improving the generalization performance of pre-training and the ability to capture time series features. Specifically, the electronic device successively trains the first number of replication models using the first number of first pre-training sample sets and the first number of second pre-training sample sets, and determines the loss values ​​corresponding to the first number of replication models during the training process, so that the initial neural network iteration is pre-trained based on the loss values ​​corresponding to the first number of replication models.

[0059] Correspondingly, the specific implementation method of S220 includes but is not limited to the following methods: for each pre-training number, use each first pre-training sample set to update the parameters of the corresponding copy model to obtain the copy model with updated parameters; use each second pre-training sample set to train the corresponding copy model with updated parameters to determine the loss values ​​corresponding to the first number of copy models.

[0060] Specifically, first, the electronic device respectively uses a specific initialization method (such as the Xavier initialization method) to initialize the parameters of the first number of replica models. In each pre-training number, each first pre-training sample set is used to update the parameters of the corresponding replica models. The parameter updating process includes loss value calculation, gradient calculation, back propagation, and updating the parameters of each replica model, so as to obtain the replica models with updated parameters. Then, each second training sample set is used to train the replica models with updated parameters, specifically calculating the loss values ​​corresponding to the first number of replica models.

[0061] S230. Average the loss values ​​corresponding to the first number of updated replication models to obtain an average loss value.

[0062] In this embodiment, after determining the loss values ​​corresponding to the first number of replica models, the electronic device adds the loss values ​​corresponding to the first number of updated replica models and averages them to obtain an average loss value, so that the average loss value is used to characterize the overall loss of the first number of updated replica models.

[0063] S240: Based on the average loss value, update the parameters of the initial neural network to obtain the initial neural network after parameter update.

[0064] In this embodiment, the electronic device performs gradient calculation, gradient back propagation and parameter update on the initial neural network in sequence based on the average loss value, thereby obtaining the initial neural network after parameter update.

[0065] S250, continue to copy the initial neural network after the parameter update to execute the step of continuing to update the parameters of the initial neural network after the parameter update, until the average value of the loss value is less than or equal to the preset loss value threshold, and obtain the pre-trained model.

[0066] It is understandable that at each pre-training number, it is determined whether the average loss value is less than or equal to the preset loss value threshold. If the average loss value is greater than the preset loss value threshold, it is necessary to continue pre-training the initial neural network after the parameter update, that is, continue to copy the initial neural network after the parameter update, return to execute S220~S240, and continue to determine whether the average loss value is less than or equal to the preset loss value threshold. If the average loss value is less than or equal to the preset loss value threshold, it means that the network parameter area of ​​the initial neural network after continuing to update the parameters is stable, and the pre-training model is obtained. In actual situations, the pre-training model can also be determined by combining the average value of the loss value and the maximum number of pre-training times.

[0067] In this way, since the wave height data and wave power generation data are strongly positively correlated, participating in model pre-training based on wave height data can enable the model to perform better in the fine-tuning stage. Moreover, through the meta-learning training method, the model can learn the temporal characteristics of different buoys and improve the generalization ability of the model.

[0068] S260, based on the linear layer and the nonlinear layer of the pre-trained model, the measured power of wave power generation in the reference area is processed to obtain the predicted power of wave power generation in the reference area.

[0069] In this embodiment, before the electronic device fine-tunes the pre-trained model, a fine-tuning data set is prepared. Specifically, the measured power of wave power generation in the reference area is obtained, and the measured power of wave power generation in the reference area is divided into training samples, verification samples and test samples according to a specific ratio, which may be 4:2:4.

[0070] Furthermore, in the model fine-tuning stage, the electronic device inputs the measured power of wave power generation in the reference area into the pre-training model, and uses the linear layer and nonlinear layer of the pre-training model to process the measured power of wave power generation in the reference area, and the pre-training model outputs the predicted power of wave power generation in the reference area.

[0071] S270, based on the predicted wave power generation in the reference area and the measured wave power generation in the reference area, fine-tune the parameters of the pre-trained model to obtain a wave power generation prediction model.

[0072] This embodiment freezes some parameters to retain the core time series features learned in pre-training. In some embodiments, the specific implementation method of S270 includes the following steps: based on the predicted power of wave power generation in the reference area and the measured power of wave power generation in the reference area, fine-tuning the parameters of the linear layer and controlling the parameters of the nonlinear layer to remain unchanged, to obtain a wave power generation power prediction model.

[0073] Specifically, this embodiment uses the method of freezing the nonlinear layer parameters to fine-tune the pre-trained model, specifically, fine-tuning the parameters of the linear layer and controlling the parameters of the nonlinear layer to remain unchanged, thereby avoiding the problem of reduced model stability caused by drastic changes in model parameters due to simultaneous changes in the linear layer and the nonlinear layer.

[0074] In other embodiments, the specific implementation method of S270 includes the following steps: based on the predicted wave power generation power in the reference area and the measured wave power generation power in the reference area, fine-tuning the parameters of the nonlinear layer and controlling the parameters of the linear layer to remain unchanged, thereby obtaining a wave power generation power prediction model.

[0075] Specifically, this embodiment uses the method of freezing the linear layer parameters to fine-tune the pre-trained model, specifically, fine-tuning the parameters of the nonlinear layer and controlling the parameters of the linear layer to remain unchanged, thereby avoiding the problem of reduced model stability caused by drastic changes in model parameters due to simultaneous changes in the linear layer and the nonlinear layer.

[0076] In this way, the measured wave power data is fine-tuned based on the parameter freezing strategy to obtain a fine-tuned wave power prediction model, thereby avoiding drastic changes in model parameters during the fine-tuning process, which may lead to a decrease in model stability.

[0077] Furthermore, after the fine-tuned wave power prediction model is determined, the fine-tuned wave power prediction model can be verified and tested based on the verification samples and the test samples to improve the reliability and prediction accuracy of the wave power prediction model.

[0078] The present disclosure also provides a wave energy power prediction device for implementing the above-mentioned wave energy power prediction method. Figure 3 In the embodiment of the present disclosure, the wave energy power generation prediction device is configured in an electronic device or a server. The electronic device may include a device with communication function such as a tablet computer, a desktop computer, a laptop computer, etc., and may also include a device simulated by a virtual machine or a simulator. The server may include a server cluster and a cloud server.

[0079] Figure 3 A schematic structural diagram of a wave energy power generation prediction device provided by an embodiment of the present disclosure is shown.

[0080] like Figure 3 As shown, the wave energy power generation prediction device 300 may include:

[0081] An acquisition module 310 is used to acquire the actual wave power generation power in the area to be predicted;

[0082] The prediction module 320 is used to process the measured wave power of the area to be predicted by using the fine-tuned wave power prediction model to determine the predicted wave power of the area to be predicted;

[0083] The wave power prediction model uses wave height data of a reference area to pre-train an initial neural network, and uses the measured wave power of the reference area to fine-tune the obtained pre-trained model, wherein the wave height data is positively correlated with the measured wave power of the reference area, and the relationship value of the positive correlation is greater than a preset positive correlation value.

[0084] A wave power generation prediction device of an embodiment of the present disclosure obtains the measured power of wave power generation in the area to be predicted; uses a fine-tuned wave power generation prediction model to process the measured power of wave power generation in the area to be predicted, and determines the predicted power of wave power generation in the area to be predicted; wherein the wave power generation prediction model uses the wave height data of the reference area to pre-train the initial neural network, and uses the measured power of wave power generation in the reference area to fine-tune the obtained pre-trained model, and the wave height data is positively correlated with the measured power of wave power generation in the reference area, and the relationship value of the positive correlation is greater than the preset positive correlation value. Therefore, using publicly available wave height data for model pre-training can solve the current problem of lack of measured power generation, and quickly fine-tune the pre-trained model based on a small amount of measured power of wave power generation to adapt to the wave power prediction scenario. In this way, accurate prediction of wave power generation is achieved using a small amount of measured power of wave power generation.

[0085] In some embodiments of the present disclosure, the prediction module 320 includes:

[0086] A first processing unit is used to perform dimensionality-upgrading processing on the wave power measured power in the area to be predicted based on the first linear layer in the fine-tuned wave power prediction model to obtain the wave power measured power in the first dimension;

[0087] A second processing unit is used to perform time series feature inference processing on the measured power of wave power generation in the first dimension based on the nonlinear layer in the fine-tuned wave power generation prediction model, and predict the power inference feature of the first dimension;

[0088] The third processing unit is used to perform dimensionality reduction processing on the power inference features of the first dimension based on the second linear layer in the fine-tuned wave power prediction model to obtain power prediction features of the second dimension, and use the power prediction features of the second dimension as the wave power prediction power of the area to be predicted.

[0089] In some embodiments of the present disclosure, the prediction module 320 includes:

[0090] An identification unit, used to identify the location of missing power from the measured power of wave power generation in the area to be predicted;

[0091] A correction unit, configured to make up for the missing power based on the wave power generation measured power before the position of the missing power and the wave power generation measured power after the position of the missing power, so as to obtain a corrected wave power generation measured power;

[0092] A normalization processing unit, used for normalizing the corrected wave power generation measured power to obtain normalized measured power;

[0093] The prediction unit is used to process the normalized measured power by using the fine-tuned wave power prediction model to determine the wave power prediction power of the area to be predicted.

[0094] In some embodiments of the present disclosure, the device further includes: a pre-training module; the pre-training module includes:

[0095] a sample acquisition unit, configured to acquire a first number of first pre-trained sample sets and a first number of second pre-trained sample sets from the wave height data, and to copy the initial neural network to acquire a first number of copy models, wherein the first pre-trained sample set and the second pre-trained sample set contain different wave height data;

[0096] A first determining unit is used to update the model parameters of the first number of replication models based on the first number of first pre-training sample sets and the first number of second pre-training sample sets for each pre-training number of times, and determine the loss values ​​respectively corresponding to the first number of replication models;

[0097] A second determining unit is used to average the loss values ​​corresponding to the first number of updated replication models to obtain an average loss value;

[0098] A parameter updating unit, used to update the parameters of the initial neural network based on the average loss value to obtain the initial neural network after parameter update;

[0099] The pre-training unit is used to continue to copy the initial neural network after the parameter update to execute the step of continuing to update the parameters of the initial neural network after the parameter update until the average value of the loss value is less than or equal to the preset loss value threshold, thereby obtaining the pre-training model.

[0100] In some embodiments of the present disclosure, the first determining unit is specifically configured to:

[0101] For each of the pre-training times, using each first pre-training sample set to update the parameters of the corresponding replication model to obtain a replication model with updated parameters;

[0102] Use each second pre-training sample set to train the corresponding replication model with updated parameters to determine the loss values ​​corresponding to the first number of replication models respectively.

[0103] In some embodiments of the present disclosure, the device further includes: a fine-tuning module; the fine-tuning module includes:

[0104] a fourth processing unit, configured to process the measured power of wave power generation in the reference area based on the linear layer and the nonlinear layer of the pre-trained model to obtain the predicted power of wave power generation in the reference area;

[0105] A parameter fine-tuning unit is used to fine-tune the parameters of the pre-trained model based on the wave power generation prediction power of the reference area and the wave power generation measured power of the reference area to obtain the wave power generation prediction model.

[0106] In some embodiments of the present disclosure, the parameter fine-tuning unit is specifically used to:

[0107] Based on the predicted wave power generation power of the reference area and the measured wave power generation power of the reference area, the parameters of the linear layer are fine-tuned and the parameters of the nonlinear layer are controlled to remain unchanged, so as to obtain the wave power generation power prediction model; or

[0108] Based on the predicted wave power generation power of the reference area and the measured wave power generation power of the reference area, the parameters of the nonlinear layer are fine-tuned and the parameters of the linear layer are controlled to remain unchanged, so as to obtain the wave power generation power prediction model.

[0109] It should be noted that Figure 3 The wave energy power generation prediction device 300 shown can execute Figures 1-2 The various steps in the method embodiment shown in the figure are implemented Figures 1-2 The various processes and effects in the method embodiment shown are not described in detail here.

[0110] Figure 4A schematic structural diagram of an electronic device provided by an embodiment of the present disclosure is shown.

[0111] like Figure 4 As shown, the electronic device may include a processor 401 and a memory 402 storing computer program instructions.

[0112] Specifically, the processor 401 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0113] The memory 402 may include a large capacity memory for information or instructions. By way of example and not limitation, the memory 402 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a tape, or a universal serial bus (USB) drive or a combination of two or more of these. Where appropriate, the memory 402 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 402 may be inside or outside the integrated gateway device. In a particular embodiment, the memory 402 is a non-volatile solid-state memory. In a particular embodiment, the memory 402 includes a read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (Electrically Erasable Programmable ROM, EEPROM), an electrically rewritable ROM (EAROM) or a flash memory, or a combination of two or more of these.

[0114] The processor 401 reads and executes the computer program instructions stored in the memory 402 to perform the steps of the wave energy power generation prediction method provided in the embodiment of the present disclosure.

[0115] In one example, the electronic device may further include a transceiver 403 and a bus 404. Figure 4 As shown, the processor 401, the memory 402 and the transceiver 403 are connected via a bus 404 and communicate with each other.

[0116] The bus 404 includes hardware, software, or both. For example, but not limitation, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industrial Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a Memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses or a combination of two or more of these. Where appropriate, the bus 404 may include one or more buses. Although embodiments of the present application describe and illustrate a particular bus, the present application contemplates any suitable bus or interconnect.

[0117] Of course, the storage medium containing computer executable instructions provided in the embodiment of the present disclosure is not limited to the above method operations, and the computer executable instructions can also execute related operations of the wave energy power generation prediction method provided in any embodiment of the present disclosure.

[0118] Through the above description of the implementation methods, the technical personnel in the relevant field can clearly understand that the present disclosure can be implemented by means of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present disclosure, or the part that contributes to the prior art, can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer cloud platform (which can be a personal computer, a server, or a network cloud platform, etc.) to execute the wave energy power generation power prediction method provided by each embodiment of the present disclosure.

[0119] Note that the above are only preferred embodiments of the present disclosure and the technical principles used. Those skilled in the art will understand that the present disclosure is not limited to the specific embodiments herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present disclosure. Therefore, although the present disclosure is described in more detail through the above embodiments, the present disclosure is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present disclosure, and the scope of the present disclosure is determined by the scope of the appended claims.

Claims

1. A wave energy power generation prediction method, characterized in that: include: Obtain the measured power of wave power generation in the area to be predicted; Using the fine-tuned wave power prediction model, the measured wave power of the area to be predicted is processed to determine the predicted wave power of the area to be predicted; The wave power prediction model uses wave height data of a reference area to pre-train an initial neural network, and uses the measured power of wave power generation in the reference area to fine-tune the obtained pre-trained model, wherein the wave height data is positively correlated with the measured power of wave power generation in the reference area, and the relationship value of the positive correlation is greater than a preset positive correlation value; The method of using the fine-tuned wave power prediction model to process the measured wave power of the area to be predicted to determine the predicted wave power of the area to be predicted includes: Based on the first linear layer in the fine-tuned wave power prediction model, the wave power measured power in the area to be predicted is subjected to dimensionality-upgrading processing to obtain the wave power measured power in the first dimension; Based on the nonlinear layer in the fine-tuned wave power prediction model, performing time series feature inference processing on the wave power measured power in the first dimension to predict the power inference feature of the first dimension; Based on the second linear layer in the fine-tuned wave power prediction model, the power inference features of the first dimension are reduced in dimension to obtain power prediction features of the second dimension, and the power prediction features of the second dimension are used as the wave power prediction power of the area to be predicted.

2. The method according to claim 1, characterized in that: The method of using the fine-tuned wave power prediction model to process the measured wave power of the area to be predicted to determine the predicted wave power of the area to be predicted includes: Identifying the location of missing power from the measured power of wave power generation in the area to be predicted; Based on the wave power generation measured power before the position of the missing power and the wave power generation measured power after the position of the missing power, the missing power is supplemented to obtain the corrected wave power generation measured power; Normalizing the corrected wave power generation measured power to obtain normalized measured power; The normalized measured power is processed by using the finely tuned wave power prediction model to determine the wave power prediction power of the area to be predicted.

3. The method according to claim 1, characterized in that The pre-training method of the wave power prediction model includes: Acquire a first number of first pre-training sample sets and a first number of second pre-training sample sets from the wave height data, and copy the initial neural network to obtain a first number of copy models, wherein the first pre-training sample set and the second pre-training sample set contain different wave height data; For each pre-training number, based on the first number of first pre-training sample sets and the first number of second pre-training sample sets, updating the model parameters of the first number of replication models, and determining the loss values ​​respectively corresponding to the first number of replication models; averaging the loss values ​​corresponding to the first number of updated replication models to obtain an average loss value; Based on the average loss value, updating the parameters of the initial neural network to obtain the initial neural network after parameter update; The initial neural network after the parameter update is continuously copied to execute the step of continuing to update the parameters of the initial neural network after the parameter update until the average value of the loss value is less than or equal to the preset loss value threshold, thereby obtaining the pre-trained model.

4. The method according to claim 3, characterized in that The method of updating the model parameters of the first number of replication models based on the first number of first pre-training sample sets and the first number of second pre-training sample sets for each pre-training number of times, and determining the loss values ​​respectively corresponding to the first number of replication models, comprises: For each of the pre-training times, using each first pre-training sample set to update the parameters of the corresponding replication model to obtain a replication model with updated parameters; Use each second pre-training sample set to train the corresponding replication model with updated parameters to determine the loss values ​​corresponding to the first number of replication models respectively.

5. The method according to claim 1, characterized in that Fine-tuning methods for pre-trained models, including: Based on the linear layer and the nonlinear layer of the pre-trained model, the measured power of wave power generation in the reference area is processed to obtain the predicted power of wave power generation in the reference area; Based on the predicted wave power generation power in the reference area and the measured wave power generation power in the reference area, the parameters of the pre-trained model are fine-tuned to obtain the wave power generation power prediction model.

6. The method according to claim 5, characterized in that The method of fine-tuning parameters of the pre-trained model based on the predicted wave power generation in the reference area and the measured wave power generation in the reference area to obtain the wave power generation prediction model includes: Based on the predicted wave power generation power of the reference area and the measured wave power generation power of the reference area, the parameters of the linear layer are fine-tuned and the parameters of the nonlinear layer are controlled to remain unchanged, so as to obtain the wave power generation power prediction model; or Based on the predicted wave power generation power of the reference area and the measured wave power generation power of the reference area, the parameters of the nonlinear layer are fine-tuned and the parameters of the linear layer are controlled to remain unchanged, so as to obtain the wave power generation power prediction model.

7. A wave energy power generation prediction device, characterized in that: include: An acquisition module is used to obtain the actual wave power generation power in the area to be predicted; A prediction module, used to process the measured wave power of the area to be predicted by using the fine-tuned wave power prediction model, and determine the predicted wave power of the area to be predicted; The wave power prediction model uses wave height data of a reference area to pre-train an initial neural network, and uses the measured power of wave power generation in the reference area to fine-tune the obtained pre-trained model, wherein the wave height data is positively correlated with the measured power of wave power generation in the reference area, and the relationship value of the positive correlation is greater than a preset positive correlation value; The prediction module comprises: A first processing unit is used to perform dimensionality-upgrading processing on the wave power measured power in the area to be predicted based on the first linear layer in the fine-tuned wave power prediction model to obtain the wave power measured power in the first dimension; A second processing unit is used to perform time series feature inference processing on the measured power of wave power generation in the first dimension based on the nonlinear layer in the fine-tuned wave power generation prediction model, and predict the power inference feature of the first dimension; The third processing unit is used to perform dimensionality reduction processing on the power inference features of the first dimension based on the second linear layer in the fine-tuned wave power prediction model to obtain power prediction features of the second dimension, and use the power prediction features of the second dimension as the wave power prediction power of the area to be predicted.

8. An electronic device, characterized in that: include: processor; A memory for storing executable instructions; The processor is used to read the executable instructions from the memory and execute the executable instructions to implement the method described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the method according to any one of claims 1 to 6.

Images