TCN-based short-time photovoltaic generating capacity prediction method and system, and medium
Through the method of time -based convolutional network, the problem of accuracy and real -time performance of photovoltaic power generation is solved, and higher prediction accuracy and real -time nature have been achieved, and operation and maintenance decision -making and user experience are optimized.
Patent Information
- Application Number
- CN202411394558.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-08
- Publication Date
- 2025-01-28
AI Technical Summary
The prediction accuracy and real-time performance of existing photovoltaic power generation prediction methods are difficult to guarantee.
The short -term photovoltaic power generation prediction method based on the timing convolution network (TCN) is adopted, and by obtaining the historical data and weather data of the photovoltaic power plant, pre -processing and building a power generation predictive model, and using the model to predict the future power generation.
Improve the accuracy and real -time of photovoltaic power generation, enhance the generalization ability and stability of the model, optimize the decision -making and efficiency of operation and maintenance, and improve the user experience and satisfaction.
Smart Images
Figure CN119382079A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of intelligent prediction and data analysis, and in particular to a method, system and medium for predicting short-term photovoltaic power generation based on TCN. Background Art
[0002] Photovoltaic power generation, as a renewable energy source, has been widely used in recent years. With technological advances and cost reductions, the installed capacity of photovoltaic power generation has increased dramatically. However, the prediction of photovoltaic power generation still faces major challenges. Photovoltaic power generation is affected by many factors, including environmental factors such as light intensity, temperature, cloud cover and humidity. Accurately predicting the impact of these environmental factors on photovoltaic power generation can help power systems optimize grid scheduling, improve the reliability of power supply, and maximize the economic benefits of photovoltaic power generation. In particular, in recent years, with a large number of industrial and commercial energy storage systems equipped with large-capacity photovoltaics, the prediction of large-capacity photovoltaic power generation in a short period of time has an important impact on the charging and discharging scheduling of energy storage systems.
[0003] Existing photovoltaic power generation prediction methods are mainly based on statistical analysis and physical models. These methods often rely on a large amount of historical data and complex mathematical models, and the prediction accuracy and real-time performance are difficult to guarantee. Summary of the invention
[0004] The purpose of the present invention is to provide a short-term photovoltaic power generation prediction method, system and medium based on TCN, aiming to solve the problem that the prediction accuracy and real-time performance of the existing photovoltaic power generation prediction methods are difficult to ensure.
[0005] The present invention is achieved through the following technical solutions:
[0006] A short-term photovoltaic power generation prediction method based on TCN includes the following steps:
[0007] Acquire historical data of the photovoltaic power station within a first preset time in the past and weather data of the photovoltaic power station within a second preset time in the future, and obtain a first initial data set and a second initial data set respectively;
[0008] Preprocessing the first initial data set and the second initial data set to obtain a first preprocessed data set and a second preprocessed data set respectively;
[0009] An initial power generation prediction model is constructed by using a time series convolutional network, and based on the first preprocessed data set, the initial power generation prediction model is trained to obtain a power generation prediction model;
[0010] Based on the second preprocessed data set, the future power generation of the photovoltaic power station is predicted by the power generation prediction model, and a prediction result is obtained;
[0011] Among them, TCN stands for Temporal Convolutional Network.
[0012] Optionally, the specific process of acquiring the historical data of the photovoltaic power station within a first preset time in the past and the weather data of the photovoltaic power station within a second preset time in the future to obtain the first initial data set and the second initial data set respectively is:
[0013] The real-time data collection module installed on the photovoltaic power station collects historical data of the photovoltaic power station within a first preset time in the past, and collects weather data of the photovoltaic power station within a second preset time in the future from the meteorological department;
[0014] A real-time data collection module is formed by a plurality of data collectors, and a plurality of data collectors adopt sentinel mode to perform data collection operations.
[0015] Optionally, the specific process of preprocessing the first initial data set and the second initial data set to obtain the first preprocessed data set and the second preprocessed data set respectively is:
[0016] The key feature data in the first initial data set and the second initial data are extracted, and the key feature data are normalized. The expression of the normalization processing is shown in the following formula (1):
[0017]
[0018] Among them, x norm is the normalized value, x is the original key eigenvalue, min(x) and max(x) are the minimum and maximum values of the original key eigenvalues, respectively.
[0019] Optionally, the specific process of constructing the initial power generation prediction model through the time series convolutional network is:
[0020] The initial power generation prediction model is constructed through the time series convolutional network. The initial power generation prediction model includes:
[0021] An input layer receives the first preprocessed data set as an input feature;
[0022] Convolutional layer, which extracts local features in the time series and sets the size, step size and padding of the convolution kernel;
[0023] The output layer maps the output of the convolutional layer to the final prediction target and obtains the predicted photovoltaic power generation.
[0024] Optionally, there are multiple convolutional layers, and a residual connection is set between each of the convolutional layers.
[0025] Optionally, a dilated convolution is provided in the convolution layer.
[0026] Optionally, the convolution layer uses a one-dimensional convolution layer and a ReLU activation function.
[0027] Optionally, the specific process of training the initial power generation prediction model based on the first preprocessed data set and obtaining the power generation prediction model is:
[0028] Dividing the first preprocessed data set into a training set, a validation set, and a test set;
[0029] Based on the training set, the initial power generation prediction model is trained;
[0030] In each training cycle, the data of the training set are input into the initial power generation prediction model in batches, and the prediction value is calculated by the forward propagation algorithm;
[0031] The difference between the predicted value and the actual value is calculated using the loss function, and the corresponding parameters of the initial power generation prediction model are updated through the back propagation algorithm;
[0032] According to the preset training termination conditions, the model training is completed, and based on the verification set, the trained initial power generation prediction model is verified, and the initial power generation prediction model is optimized according to the verification results;
[0033] Based on the test set, the initial power generation prediction model after model optimization is tested and the power generation prediction model is obtained.
[0034] The present invention also provides a short-time photovoltaic power generation prediction system based on TCN, which is used to implement the short-time photovoltaic power generation prediction method based on TCN, comprising:
[0035] Data acquisition module: used to obtain historical data of the photovoltaic power station within a first preset time in the past and weather data within a second preset time in the future, and generate a first initial data set and a second initial data set respectively;
[0036] Data preprocessing module: used for preprocessing the first initial data set and the second initial data set;
[0037] Model building module: used to build power generation prediction model;
[0038] Prediction and feedback module: used to predict the future power generation of photovoltaic power stations using the power generation prediction model, and present the prediction results to users in the form of charts, reports, etc., and give corresponding operation and maintenance measures based on the prediction results;
[0039] User interaction module: used to receive user input, display prediction results and operation and maintenance suggestions, and collect user feedback on the system.
[0040] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the TCN-based short-time photovoltaic power generation prediction method are implemented.
[0041] The technical solution of the present invention has at least the following advantages and beneficial effects:
[0042] Improve the accuracy and real-time performance of photovoltaic power generation forecasts: By building a power generation forecast model through a time series convolutional network, and utilizing its powerful processing capabilities on time series data, it can accurately capture the complex patterns of photovoltaic power generation changes over time, thereby improving the accuracy of the forecast; the introduction of the sentinel mode for data collection ensures the real-time collection and rapid processing of a large amount of photovoltaic data, further improving the real-time performance of the forecast, and enabling operation and maintenance personnel to obtain forecast information on power generation in a timely manner.
[0043] Enhance the generalization and stability of the model: Use normalization to preprocess the raw data, eliminate the impact of different dimensions and scales on model training, and improve the generalization ability of the model; through residual connection and dilated convolution and other technologies, effectively alleviate the common gradient vanishing problem in deep neural networks, so that the model can learn deeper features, improve the stability and prediction performance of the model.
[0044] Optimize operation and maintenance decision-making and efficiency: According to the prediction results and the actual situation of the photovoltaic power station, the corresponding operation and maintenance measures are automatically generated and pushed to relevant personnel in a timely manner through system messages, emails or mobile phone text messages, so that the operation and maintenance work can be responded to and executed quickly, improving the operation and maintenance efficiency; establish an execution tracking mechanism for operation and maintenance measures, collect and analyze feedback data after the execution of operation and maintenance measures, which is used to evaluate the effectiveness of operation and maintenance measures and improve operation and maintenance strategies, further optimizing the operation and maintenance decision-making process.
[0045] Improve user experience and satisfaction: Provide a user-friendly interactive interface, allowing users to easily enter query conditions, select forecast time periods, and view forecast results and operation and maintenance suggestions in real time, which improves user experience; forecast results are presented in the form of intuitive and easy-to-understand charts and reports, allowing users to quickly understand the future power generation of photovoltaic power stations and improve user satisfaction.
[0046] Support continuous optimization and improvement: Based on user feedback and actual application conditions, the performance of the power generation prediction model can be continuously optimized to improve prediction accuracy and stability, ensuring that the model always remains in the best state; the formulation and implementation process of operation and maintenance measures can be continuously improved to improve operation and maintenance efficiency and effectiveness, providing long-term support for the stable operation and efficient power generation of photovoltaic power stations. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1It is a schematic flow chart of a method for predicting short-time photovoltaic power generation based on TCN in Example 1 of the present invention;
[0048] Figure 2 Schematic diagram of the structure of the TCN-based short-time photovoltaic power generation prediction system in Example 2 of the present invention. DETAILED DESCRIPTION
[0049] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.
[0050] Example 1
[0051] Reference Figure 1 , a short-time photovoltaic power generation prediction method based on TCN, comprising the steps of:
[0052] Step 1: Obtain historical data of the photovoltaic power station within a first preset time in the past and weather data of the photovoltaic power station within a second preset time in the future, and obtain a first initial data set and a second initial data set respectively;
[0053] In this embodiment, a real-time data acquisition module is provided on the photovoltaic power station to collect historical data of the photovoltaic power station within a first preset time in the past, and to collect weather data of the photovoltaic power station within a second preset time in the future from the meteorological department; the historical data are relevant historical data of the photovoltaic power station, such as historical data such as the power generation of the power station; the weather data may be relevant weather data such as light intensity, temperature, cloud cover and humidity; the historical data within the first preset time in the past may be historical data such as power generation within the past n minutes; the weather data within the second preset time in the future may be weather data such as light intensity, temperature, cloud cover and humidity within the next n minutes.
[0054] A real-time data collection module is formed by a plurality of data collectors, and a plurality of data collectors adopt sentinel mode to perform data collection operations.
[0055] In this embodiment, the sentinel mode for data collection operation means: when multiple data collectors collect photovoltaic data at the same time, the data collector that collects data first is used as the sentinel data collector; when a problem occurs in a non-sentinel data collector, the sentinel data collector wakes up the spare non-sentinel data collector to replace the problematic non-sentinel data collector; when a problem occurs in the sentinel data collector, the non-sentinel data collector competes and obtains a new sentinel data collector, and the problematic sentinel data collector is converted into a non-sentinel data collector.
[0056] Data collection operations are performed in sentinel mode to ensure the real-time collection of large amounts of photovoltaic data; at the same time, it ensures that the power generation prediction model can be rapidly incrementally learned.
[0057] Step 2: preprocessing the first initial data set and the second initial data set to obtain a first preprocessed data set and a second preprocessed data set respectively;
[0058] In this embodiment, the specific process of preprocessing the first initial data set and the second initial data set to obtain the first preprocessed data set and the second preprocessed data set is as follows:
[0059] The key feature data in the first initial data set and the second initial data are extracted, and the key feature data are normalized. The expression of the normalization processing is shown in the following formula (1):
[0060]
[0061] Among them, x norm is the normalized value, x is the original key eigenvalue, min(x) and max(x) are the minimum and maximum values of the original key eigenvalues, respectively.
[0062] In this embodiment, the original key features include: photovoltaic power generation, light intensity, temperature, cloud cover and humidity. Normalization is to scale these original key features to a specific range in order to improve the training effect of the model. The specific range is usually 0 to 1.
[0063] Step 3: construct an initial power generation prediction model through a time series convolutional network, train the initial power generation prediction model based on the first preprocessed data set, and obtain a power generation prediction model;
[0064] In this embodiment, the specific process of constructing the initial power generation prediction model through the time series convolutional network is as follows:
[0065] The initial power generation prediction model is constructed through the time series convolutional network. The initial power generation prediction model includes:
[0066] The input layer receives the first preprocessed data set through the input layer and uses it as an input feature. In this embodiment, it is assumed that X represents a feature vector, X=[z,h,c,m], where z is a light intensity feature value, h is a temperature feature value, c is a cloud cover feature value, and m is a humidity feature value.
[0067] Convolutional layer, which extracts local features in the time series and sets the size, step size and padding of the convolution kernel;
[0068] The convolution operation can be expressed as the following formula (2):
[0069] Y=W*X+b(2)
[0070] Wherein, Y is the output of the convolution layer, W is the weight matrix of the convolution kernel, the shape of the convolution kernel is (K, d), K is the size of the convolution kernel, d is the dimension of the feature vector, in this embodiment, d is 4, the shape of the feature vector is (n, d), n is the number of time steps, b is the bias term, * is the convolution operation, and the convolution operation is used to extract the local time series features in the input data, and generate the output feature map by sliding the convolution kernel on the input data. In this embodiment, the convolution layer uses a one-dimensional convolution layer and a ReLU activation function.
[0071] In this embodiment, there are several convolutional layers, and a residual connection is set between each of the convolutional layers; the expression of the residual connection is shown in the following formula (3):
[0072] F(x)=H(x)+x(3)
[0073] Among them, F(x) is the output of the residual connection, H(x) is the output of the convolution layer, and x is the original data input to the convolution layer. Through the residual connection, the model can learn deeper features and avoid the gradient vanishing problem. The residual connection adds the original data x to the output H(x) of the convolution layer, so that each layer can directly learn the residual of the previous layer, that is, the difference.
[0074] In this embodiment, a dilated convolution is provided in the convolution layer; the expression of the dilated convolution is shown in the following formula (4):
[0075]
[0076] Among them, Y[i] is the output of the dilated convolution, W[k] is the kth weight of the convolution kernel, X[i+D·k] is the i+D·kth element of the input data, and D is the dilation factor, which determines the sampling interval of the input signal. The receptive field is increased by skipping some points in the middle of the input data. The skipped points are controlled by the dilation factor D, which helps the model capture long-term dependencies in the time series. Dilated convolution increases the receptive field by inserting zeros (i.e., "holes") between the elements of the convolution kernel, allowing the network to capture longer historical dependencies without increasing the number of parameters.
[0077] The output layer maps the output of the convolutional layer to the final prediction target and obtains the predicted photovoltaic power generation. The expression of the output layer is shown in the following formula (5):
[0078] P=W·X flat +b (5)
[0079] Among them, P is the predicted photovoltaic power generation output by the model, X flatIt is the flattened representation of the convolutional layer output.
[0080] In this embodiment, based on the first preprocessed data set, the specific process of training the initial power generation prediction model and obtaining the power generation prediction model is as follows:
[0081] The first preprocessed data set is divided into a training set, a validation set, and a test set; the first preprocessed data set is divided into a training set, a validation set, and a test set according to a certain ratio. A common division ratio may be 70% for the training set, 15% for the validation set, and 15% for the test set; the first preprocessed data set has been normalized and contains key features such as photovoltaic power generation, light intensity, temperature, cloud cover, and humidity.
[0082] Based on the training set, the initial power generation prediction model is trained; in each training cycle, the data of the training set is input into the initial power generation prediction model in batches, and the prediction value is calculated by the forward propagation algorithm; the difference between the prediction value and the actual value is calculated using the loss function, and the loss function can be the mean square error MSE, and the corresponding parameters of the initial power generation prediction model are updated through the back propagation algorithm.
[0083] Repeat the above training steps and complete the model training according to the preset training termination conditions. The training termination conditions can be the maximum number of iterations, the loss value is stable, or the performance of the validation set cannot be improved any further.
[0084] Based on the validation set, the initial power generation prediction model that has completed training is validated to evaluate the performance of the model on unseen data, such as calculating indicators such as loss value and accuracy on the validation set.
[0085] According to the verification results, the initial power prediction model is optimized, such as adjusting the network structure, learning rate, batch size and other hyperparameters. Hyperparameter tuning can be performed using methods such as Grid Search, Random Search or Bayesian Optimization.
[0086] Based on the test set, the initial power generation prediction model after model optimization is tested and the power generation prediction model is obtained. Its final performance is evaluated, and if the test results meet the requirements, the model is deployed as the power generation prediction model.
[0087] Step 4: Based on the second preprocessed data set, the future power generation of the photovoltaic power station is predicted by the power generation prediction model, and a prediction result is obtained;
[0088] In this embodiment, it is necessary to ensure that the second preprocessed data set has been properly prepared and meets the input requirements of the power generation prediction model; load the trained and optimized power generation prediction model from the storage medium. The power generation prediction model is obtained by training the first preprocessed data set through a time series convolutional network, and has the ability to predict photovoltaic power generation; the second preprocessed data set is used as input and passed to the subsequent convolution layer through the input layer of the model. In the convolution layer, the model processes the input data through the set convolution kernel size, step size and padding method, as well as dilated convolution and residual connection techniques, extracts local and global features in the time series, and the output layer maps the features extracted by the convolution layer to the final prediction target, that is, the future power generation of the photovoltaic power station. This process is implemented through the forward propagation algorithm to calculate the prediction results; organize the prediction results into an easy-to-understand form, such as time series data, charts, etc.; store the prediction results in a specified location, such as a database, file system, etc., for subsequent query and analysis.
[0089] In this embodiment, it also includes: Step 5, feeding back the prediction results to the user, and providing corresponding operation and maintenance measures. The prediction results are presented to the user in an intuitive and easy-to-understand manner. The predicted value of photovoltaic power generation and its changing trend in the future period of time (such as the next n minutes, hours, days) can be displayed through charts (such as line charts, bar charts, etc.); on the user interface, a special display area can be designed to update the prediction results in real time to ensure that users can grasp the future power generation of photovoltaic power stations at any time; a detailed report of the prediction results can also be generated, including the prediction time range, predicted power generation, prediction error rate, weather condition analysis, etc. The report can be saved in PDF, Excel and other formats for users to further analyze and archive.
[0090] In this embodiment, corresponding operation and maintenance measures are formulated based on the prediction results and the actual situation of the photovoltaic power station. For example, if it is predicted that the light intensity will drop significantly in a certain period of time in the future, which may affect the power generation, operation and maintenance work such as cleaning the photovoltaic panels and checking the inverter status can be arranged in advance; the operation and maintenance measures are pushed to relevant personnel in a timely manner through system messages, emails or mobile phone text messages to ensure that the operation and maintenance work can be responded to and executed quickly. An execution tracking mechanism for operation and maintenance measures is established to record the execution status, execution time and execution effect of operation and maintenance measures; feedback data after the execution of operation and maintenance measures is collected, such as the increase in power generation after cleaning the photovoltaic panels and the stability of the inverter after maintenance, etc., to evaluate the effectiveness of operation and maintenance measures and improve operation and maintenance strategies.
[0091] In this embodiment, based on user feedback and actual application conditions, the performance of the power generation prediction model is continuously optimized to improve prediction accuracy and stability; the formulation and implementation process of operation and maintenance measures is continuously improved to improve operation and maintenance efficiency and effectiveness, providing strong support for the stable operation and efficient power generation of photovoltaic power stations.
[0092] Embodiment 2
[0093] Based on Example 1, refer to Figure 2 In this embodiment, a short-time photovoltaic power generation prediction system based on TCN is provided, which is used to implement a short-time photovoltaic power generation prediction method based on TCN, including:
[0094] Data collection module: used to obtain historical data of the photovoltaic power station within the first preset time in the past and weather data within the second preset time in the future, and generate the first initial data set and the second initial data set respectively; specifically, the real-time data collection module set on the photovoltaic power station collects the historical data of photovoltaic power generation within the first preset time in the past, and obtains the weather data within the second preset time in the future from the meteorological department, and these data are integrated into the first initial data set and the second initial data set. The real-time data collection module is composed of several data collectors, which perform data collection operations in sentinel mode.
[0095] Data preprocessing module: used to preprocess the first initial data set and the second initial data set; extract key feature data and perform normalization processing. The normalization processing is performed using formula (1) to ensure the consistency and comparability of the data. The preprocessed data form the first preprocessing data set and the second preprocessing data set respectively.
[0096] Model building module: used to build a power generation prediction model; build an initial power generation prediction model based on a time series convolutional network. The model includes an input layer, a convolutional layer, and an output layer. The input layer receives the preprocessed data set as input features; the convolutional layer extracts local and global features in the time series by setting the size, step size, and padding of the convolution kernel, and using technologies such as dilated convolution and residual connection; the output layer maps the features output by the convolutional layer to the final prediction target, that is, the predicted photovoltaic power generation.
[0097] Prediction and feedback module: used to predict the future power generation of photovoltaic power stations using the power generation prediction model, and present the prediction results to users in the form of charts, reports, etc.; according to the prediction results, corresponding operation and maintenance measures are recommended to help photovoltaic power stations optimize operation and management.
[0098] User interaction module: used to receive user input, display prediction results and operation and maintenance suggestions, and collect user feedback on the system. Provide users with a friendly interactive interface, through which users can enter query conditions, select prediction time periods, etc., and the system returns corresponding prediction results and operation and maintenance suggestions based on user requests.
[0099] Embodiment 3
[0100] Based on Example 1, in this embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the short-time photovoltaic power generation prediction method based on TCN are implemented.
[0101] It is understandable that the processor in the embodiments of the present invention may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.
[0102] The method steps in the embodiments of the present invention can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC.
[0103] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a storage medium or transmitted through the storage medium. The computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state disk (SSD)), etc.
[0104] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A short-term photovoltaic power generation prediction method based on TCN, characterized in that: Includes steps: Acquire historical data of the photovoltaic power station within a first preset time in the past and weather data of the photovoltaic power station within a second preset time in the future, and obtain a first initial data set and a second initial data set respectively; Preprocessing the first initial data set and the second initial data set to obtain a first preprocessed data set and a second preprocessed data set respectively; An initial power generation prediction model is constructed by using a time series convolutional network, and based on the first preprocessed data set, the initial power generation prediction model is trained to obtain a power generation prediction model; Based on the second preprocessed data set, the future power generation of the photovoltaic power station is predicted by the power generation prediction model, and a prediction result is obtained.
2. The short-term photovoltaic power generation prediction method based on TCN according to claim 1 is characterized in that: The specific process of obtaining the historical data of the photovoltaic power station within a first preset time in the past and the weather data of the photovoltaic power station within a second preset time in the future to obtain the first initial data set and the second initial data set respectively is as follows: The real-time data collection module installed on the photovoltaic power station collects historical data of the photovoltaic power station within a first preset time in the past, and collects weather data of the photovoltaic power station within a second preset time in the future from the meteorological department; A real-time data collection module is formed by a plurality of data collectors, and a plurality of data collectors adopt sentinel mode to perform data collection operations.
3. The short-term photovoltaic power generation prediction method based on TCN according to claim 1 is characterized in that: The specific process of preprocessing the first initial data set and the second initial data set to obtain the first preprocessed data set and the second preprocessed data set respectively is: The key feature data in the first initial data set and the second initial data are extracted, and the key feature data are normalized. The expression of the normalization processing is shown in the following formula (1): Among them, x norm is the normalized value, x is the original key eigenvalue, min(x) and max(x) are the minimum and maximum values of the original key eigenvalues, respectively.
4. The short-term photovoltaic power generation prediction method based on TCN according to claim 1 is characterized in that: The specific process of constructing the initial power generation prediction model through the time series convolutional network is as follows: The initial power generation prediction model is constructed through the time series convolutional network. The initial power generation prediction model includes: An input layer receives the first preprocessed data set as an input feature; Convolutional layer, which extracts local features in the time series and sets the size, step size and padding of the convolution kernel; The output layer maps the output of the convolutional layer to the final prediction target and obtains the predicted photovoltaic power generation.
5. The short-term photovoltaic power generation prediction method based on TCN according to claim 4 is characterized in that: There are several convolutional layers, and residual connections are set between each of the convolutional layers.
6. The short-term photovoltaic power generation prediction method based on TCN according to claim 4, characterized in that: The convolution layer is provided with a dilated convolution.
7. The short-term photovoltaic power generation prediction method based on TCN according to claim 4, characterized in that: The convolution layer uses a one-dimensional convolution layer and a ReLU activation function.
8. The short-term photovoltaic power generation prediction method based on TCN according to claim 1, characterized in that: The specific process of training the initial power generation prediction model based on the first preprocessed data set and obtaining the power generation prediction model is as follows: Dividing the first preprocessed data set into a training set, a validation set, and a test set; Based on the training set, the initial power generation prediction model is trained; In each training cycle, the data of the training set are input into the initial power generation prediction model in batches, and the prediction value is calculated by the forward propagation algorithm; The difference between the predicted value and the actual value is calculated using the loss function, and the corresponding parameters of the initial power generation prediction model are updated through the back propagation algorithm; According to the preset training termination conditions, the model training is completed, and based on the verification set, the trained initial power generation prediction model is verified, and the initial power generation prediction model is optimized according to the verification results; Based on the test set, the initial power generation prediction model after model optimization is tested and the power generation prediction model is obtained.
9. A short-time photovoltaic power generation prediction system based on TCN, used to implement the short-time photovoltaic power generation prediction method based on TCN as described in any one of claims 1 to 8, characterized in that: include: Data acquisition module: used to obtain historical data of the photovoltaic power station within a first preset time in the past and weather data within a second preset time in the future, and generate a first initial data set and a second initial data set respectively; Data preprocessing module: used for preprocessing the first initial data set and the second initial data set; Model building module: used to build power generation prediction model; Prediction and feedback module: used to predict the future power generation of photovoltaic power stations using the power generation prediction model, and present the prediction results to users in the form of charts, reports, etc., and give corresponding operation and maintenance measures based on the prediction results; User interaction module: used to receive user input, display prediction results and operation and maintenance suggestions, and collect user feedback on the system.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the TCN-based short-time photovoltaic power generation prediction method as described in any one of claims 1 to 8 are implemented.