TW201740296A - Method and system for predicting power generation capacity of renewable energy using a neural network to accurately calculate the power generation capacity of renewable energy - Google Patents

Abstract

A system for predicting power generation capacity of renewable energy comprises a data retrieval module, a database, an abnormality point correction module, a prediction module, an input module, and an output module. The data retrieval module retrieves past weather data and past power generation data. The database stores the past weather data and the past power generation data. The abnormality point correction module removes abnormal data from the past weather data and the past power generation data. The prediction module performs training on neural network prediction model based on the past weather data and the past power generation data from which abnormal data has been filtered, so as to obtain a power generation capacity prediction model. The input module inputs future weather data, solar trajectory, and atmosphere mass to the prediction module so as to allow the prediction module to calculate a predicted power generation capacity. The output module outputs the predicted power generation capacity.

Description

Renewable energy generation quantity prediction method and system

This creation is about a method and system for predicting the amount of power generated by renewable energy, and in particular, a method and system for predicting the amount of power generated by a renewable neural network using a neural network.

Renewable energy is an inexhaustible source of energy from nature, such as solar energy, wind power, tidal energy, geothermal energy, biomass energy or water power. However, the amount of electricity generated by renewable energy power generation systems will be affected by many external factors, such as climate temperature, solar intensity, or seasonal factors. These external factors affect the power generation of renewable energy power generation systems to varying degrees, and renewable energy power generation systems. It can be regarded as an unstable power source. Therefore, it is important to study the randomness of renewable energy and the estimation technology of renewable energy power generation.

However, the shortcomings such as randomness, volatility, intermittentness and uncertainty of power generation in renewable energy power plants have caused certain difficulties in estimating renewable energy power generation. The difference between the estimation results and the actual results is too large, resulting in no reference value for the estimation results.

A neural network is a mathematical model or computational model that mimics the structure and function of a biological neural network. It is calculated by a large number of artificial neuron connections. A neural network is an operational model consisting of a large number of nodes (or neurons or units) connected to each other. Each node represents a specific output function, and the connection between each two nodes represents a weight for passing the connection signal. Neural networks have been successfully applied to predict stock indices or other commercial forecasts. If neural networks can be applied to the prediction of renewable energy generation, it is possible to estimate the amount of power generated by renewable energy plants in the future. The power generation of the power plant can be allocated in advance according to the predicted power generation amount, so as to avoid insufficient or excessive power supply due to excessive or too little power generation, resulting in low power utilization.

The purpose of this creation is to design a renewable energy generation forecasting system that uses a neural network to predict the amount of power generated by a renewable energy plant.

According to the above purpose, the present invention provides a renewable energy power generation quantity prediction system, comprising: a database, storing a plurality of historical climate data, a plurality of historical power generation data, a plurality of solar trajectory data, and a plurality of air quality data; Taking a module, which is connected to the database, and extracts the historical climate data, the historical power generation data, the solar trajectory data and the air quality data from the database; an abnormal point correction module, and The data capture module connection, filtering the historical climate data, the historical power generation data, the solar trajectory data and the plurality of abnormal data in the air quality data; a prediction module, and the abnormality Point correction module connection, the prediction module performs training on a neural network prediction model based on the historical climate data, the historical power generation data, the solar trajectory data and the air quality data of the abnormal data. Obtaining a power generation prediction model; an input module, connecting the prediction module, inputting a plurality of future climate data, and A male track and predict the air mass to the module so that the module calculating a predicted power generation amount prediction; an output module, connected to the forecast module, for outputting the predicted power generation amount.

Another object of the present invention is to provide a method for predicting the amount of renewable energy generation, by which a neural network prediction model is trained to obtain a better neural network prediction model, and then a predictive power plant with high accuracy can be calculated. the amount.

According to the above purpose, the present invention provides a method for predicting the amount of renewable energy power generation, comprising the following steps: obtaining a plurality of historical climate data, a plurality of historical power generation data, a plurality of solar trajectory data, and a plurality of air quality through a data acquisition module. Data; correcting the historical climate data, the historical power generation data, the solar trajectory data, and the plurality of abnormal data in the air quality data through an abnormal point correction module; The historical climate data of the abnormal data, the historical power generation data, the solar trajectory data and the air quality data are transmitted to a prediction module, and a neural network prediction model is trained to obtain a power generation prediction model. Entering a plurality of future time intervals to predict climatic data, a solar trajectory, and a mass of mass to the prediction module to perform power generation prediction of the future time interval; and predicting the power generation amount of the power plant through an output module Output.

Through the creation of the renewable energy generation forecasting method and system, a reliable prediction of the power generation of the power plant can be calculated, because the creation of the power generation quantity prediction model is based on historical climate data and historical power generation data through a neural network. Some historical climate data contain various power generation environment states, and these historical power generation data are closely related to various power generation environment states. Therefore, the power generation quantity prediction model established by this creation will randomly generate the power generation of renewable energy power plants due to environmental climate factors. Uncertainty factors such as sex, volatility and intermittentness are taken into consideration, so that the predicted power generation can be referenced, allowing plant managers to effectively allocate power generation capacity by predicting power generation of power plants and improve utilization of power generation. .

The technical means adopted by the present invention for achieving the intended purpose of creation are further explained below in conjunction with the drawings and the preferred embodiment of the present invention.

Figure 1 is a block diagram of the regenerative energy generation forecasting system of the present invention. As shown in FIG. 1 , the regenerative energy prediction system 10 of the present invention includes a data acquisition module 11 , a database 12 , an abnormal point correction module 13 , a prediction module 14 , a neural network prediction model 15 , and an input module 16 . And the output module 17.

Historical climate data, historical power generation data, historical solar trajectory data and historical air quality data of renewable energy power plants can be recorded in the database by staff, 12 or historical climate data, historical solar trajectory data or history of renewable energy power plants. The air quality data can be measured by the power plant climate detection component or by the computer system from the weather website. The power plant climate detection component can be a thermometer, gyroscope or electronic device meteorological data application software, etc., can obtain temperature, solar azimuth or It consists of tools such as the solar elevation angle. The historical power generation data can be transmitted from the power generation component of the power plant to the computer system, and then stored in the database 12 by the computer system, which is not limited thereto. Historical climate data, historical power generation data, historical solar trajectory data and historical air quality data of renewable energy power plants are used as the basis for predicting future power generation. Historical climate data can be solar radiation, temperature, humidity, air pressure, wind speed, wind direction, and rainfall. , cloud cover or soil temperature, etc., but not limited here. The historical power generation data may be direct current (DC) data of a renewable energy power plant, such as voltage, current, renewable energy device power generation efficiency (kWh/kWp/h), kilowatts (kW), per kilowatt hour (kWh), or per unit device. Power generation (kWh/kWp), etc. or AC (AC) data, such as voltage, current, regenerative power generation efficiency (kWh/kWp/h), kilowatts (kW), per kWh (kWh) or per unit Power generation (kWh/kWp), etc., but not limited here.

The data capture module 11 is connected to the database 12 and the abnormal point correction module 13. The data capture module 11 can be an execution instruction or application software of the computer or the mobile device, and the data capture module 11 can be set in a At a fixed time (per minute, hourly or daily, etc.), the climate data, power generation data, solar trajectory and air quality of the solar power plant are extracted. The design of the data acquisition module 11 and the execution instruction are designed as basic software designers in the field. It is well known and will not be described here. When the neural network prediction model 15 is to be trained, the data acquisition module 11 takes the historical climate data, historical power generation data, historical solar trajectory data and historical air quality data of the renewable energy power plant from the database 12 The historical climate data, the historical power generation data, the historical solar trajectory data, and the historical atmospheric quality data taken out are transmitted to the abnormal point correction module 13. The abnormal point correction module 13 is configured to analyze the historical climate data, historical power generation data, historical solar trajectory and historical air quality, and find out the abnormal data through linear regression, normal distribution or data comparison, and then perform Filter or correct to improve the accuracy of the prediction. For example, the power generation efficiency (PR (kWh/kWp/h)) of the renewable energy power plant can be applied as the basis for the correction of the abnormal point, and the average power generation efficiency of the previous power generation device is used as the abnormal point judgment and correction. The benchmark is considered to be an abnormal point if the average power generation efficiency of other weeks is too large compared with the average power generation efficiency of the week. Further correcting or filtering out the abnormal point, it is also possible to compare the power generation efficiency of different power generation devices of the same power plant at the same time. The basis for the correction of the abnormal point is that if the difference between the power generation device of the same power plant and the other power generation device (power generation efficiency or unit power generation) is too large, it is also regarded as an abnormal point, and the abnormal point is further corrected or filtered. The correction method may be the same as the average power generation efficiency of the power generation device in the past period of time as the correction reference, or compare the average power generation efficiency or the average unit power generation amount of the other single or plural power generation devices of the same power plant at the same time as the reference further complements or corrects the reference. The power generation device may be one or more Maximum Power Point Tracking (MPPT) devices, one or more inverters, or the entire power plant, and is not limited herein. In addition, or in different embodiments, by setting a threshold, the abnormal point correction module 13 determines whether the historical climate data, the historical power generation data, the historical solar trajectory data, or the historical air quality data exceeds the threshold value, so as to exceed the threshold. Historical climate data, historical power generation data, historical solar trajectory data, or historical air quality data are considered as anomalous data and filtered or corrected.

The prediction module 14 is connected to the abnormal point correction module 13 , and the abnormal point correction module 13 transmits the historical climate data, the historical power generation data, the historical solar trajectory and the historical air quality after filtering or correcting the abnormal data to the prediction module 14 for class. The training of the neural network prediction model 15 obtains a power generation prediction model. Similarly, the prediction module 14, the neural network prediction model 15 and the power generation prediction model may be an application or an execution instruction of a computer system or a mobile device, how to design and construct a prediction module 14, a neural network prediction model 15 and The power generation prediction model is well known to those skilled in the art with software design or information engineering, and will not be described here. As shown in FIG. 2, the neural network prediction model 15 includes an input layer 151, a hidden layer 152, and an output layer 153. The input layer 151 receives historical climate data and historical power generation data (or training data or testing data) of the renewable energy power plant, and the output layer 153 outputs the training result or calculation of the neural network prediction model 15 . The result is shown in Figure 2. The hidden layer 152 is located between the input layer 151 and the output layer 153. The input layer 151, the hidden layer 152 and the output layer 153 respectively comprise one or more nodes 154, and the nodes 154 of the input layer 151 are respectively connected to the nodes 154 of the hidden layer 152. The nodes 154 of the hidden layer 152 are also connected to the nodes 154 of the output layer 153, respectively. When the prediction module 14 performs the training of the neural network prediction model 15, the historical climate data, the historical power generation data, the historical solar trajectory data and the historical air quality data are cut into training materials, training materials and test data according to the training mode. Then, using the training data and the training data, the training and tuning of the neural network prediction model 15 are performed. Since the technology of the neural network is well known to those of ordinary skill in the art, a detailed description of the neural network is not described here. In addition, the number of nodes 154 or the number of layers of the hidden layer 152 may have different number of nodes 154 or layers according to different requirements, and is not limited to the aspect shown in FIG. 2. Moreover, the methods for evaluating errors used in the neural network of the present invention include Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute. Percentage Error (MAPE) or Mean Relative Error (MRE), etc., with the following formula:

Where N of the above equation is the number of training materials, In order to predict the amount of power generation, It is the actual power generation value. When the predicted performance of the neural network prediction model 15 is not as expected, the prediction module 14 replaces the input variables, adjusts the neural network prediction model 15 structure or parameter settings, and gradually improves the system prediction performance. For example, when training the neural network model 15, the historical climate data, historical power generation data, solar azimuth angle (Azimuth Angle), and the sun of the abnormal data from the abnormal point correction module 13 are filtered from January to August last year. Elevation Angle and Air Mass (AM) are input to the neural network prediction model 15 for training, and then the data is tested from September to December. The output of the neural network model 15 is output. The power plant's predicted power generation from September to December is compared with the actual power generation of the power plant from September to December last year, and the error is evaluated by the above-mentioned root mean square error method, average absolute error method or average absolute proportional error method. If the error is too large, adjust the input parameters of the input layer 151 of the neural network prediction model 15 and the number of layers 152 nodes 154 of the hidden layer, or adjust the input historical climate data and historical power generation data, and then generate historical climate data and historical power generation. The data is input to the adjusted neural network prediction model 15 for another model training error evaluation, which is pre-processed through the neural network described above. 15 model training process, and finally get a higher prediction accuracy neural network prediction model 15 as a generating capacity of this creation of predictive models.

Then, after the best model is trained, the input climate module is used to input the predicted climate data for a certain period of time, the solar trajectory data for a certain period of time (such as the altitude angle and azimuth of the sun, etc.) and the future of the air mass (Air Mass). The data is sent to the prediction module 14 for prediction of the amount of power generated by the renewable energy power plant. The input module 16 may be a keyboard, and the staff of the regenerative power plant inputs the predicted climate data, the solar trajectory data and the atmospheric quality to the regenerative energy generation quantity prediction system 10, or the input module 16 may be a program installed in a computer system. Future climate data can be obtained from climate websites or relevant climate prediction units (such as academic units), and solar trajectory data and air quality can be calculated for a period of time, and then input to prediction module 14 for regeneration. Forecast of power generation by energy power plants. Among them, the predicted climate data in the future time can be obtained from the meteorological bureau's forecast data or the relevant climate prediction unit, and the Azimuth Angle A refers to the angle between the projection of the sun's rays on the ground plane and the local meridian, the sun. The Elevation Angle is the angle between the line connecting the observer's location and the center of the sun and the ground plane. The air mass (AM) is the degree to which the atmosphere affects the earth's surface to receive sunlight, and the solar azimuth of the future time. The solar elevation angle or air quality can be calculated by mathematical calculation formula. The output module 16 is connected to the prediction module 14 and outputs the predicted power generation amount calculated by the prediction module 14 through the neural network prediction model 15 through the output module 16. Through the above-described renewable energy prediction system 10, the application-like neural network prediction model 15 can effectively predict the power generation amount of the power plant, and the power plant manager can efficiently distribute the power according to the predicted power generation amount.

Figure 3 is a flow chart of the method for predicting renewable energy in the present invention. As shown in FIG. 3, in step S301, the historical climate data, the historical power generation data, the historical solar trajectory, and the historical air quality are obtained through the data acquisition module 11. The regenerative energy prediction method of this creation draws on historical climate data of renewable energy power plants, historical solar trajectories of historical power generation data and historical air quality as the basis for predicting future power generation. In step S302, the abnormal point correction module 13 filters out or corrects the abnormal data in the historical climate data, the historical power generation data, the historical solar trajectory, and the historical air quality. For example, the abnormal point correction module 13 can filter or correct the outlier data by means of Regression Analysis, and the outlier value refers to one or more values and other values in the data. The numerical values are quite different. Or, as shown in FIG. 4, the abnormal point correction module 13 can use the normal distribution method to establish a normal distribution curve of the historical climate data and the historical power generation data, and further filter the data points larger than 3 standard deviations (±3 s). Or fix it. The so-called Standard Deviation (SD) is the most commonly used measure of statistical dispersion in probability statistics. The standard deviation is a measure of the degree to which a set of values is spread out from the mean. A larger standard deviation represents a larger difference between most of the values and their average values; a smaller standard deviation means that these values are closer to the average. In various embodiments, it is even possible to design a data whose threshold value will exceed or exceed the threshold value. It should be noted that the manner of filtering the abnormal points by applying the regression analysis or the normal distribution method is well known to those skilled in the art and will not be described herein. The modified method may be used to correct the average power generation efficiency of the same power generation device over a period of time, or to compare the average power generation efficiency or the average unit power generation amount of the plurality of power generation devices of the same power generation plant at the same time, and further supplement or correct the reference. Historical climate data, historical power generation data, the solar trajectory data, and multiple abnormal data in the air quality data.

In step S303, the historical climate data, the historical power generation data, the solar trajectory and the air quality of the filtered or corrected abnormal data are transmitted to the prediction module 14, and the neural network prediction model 15 is trained by the prediction module 14 to A power generation prediction model is obtained. The neural network prediction model 15 needs to perform repeated calculations and corrections by inputting historical climate data and historical power generation data (plus solar trajectory or air quality, etc.) that have been filtered or corrected to the prediction module 14 and through Mathematical formulas such as root mean square error method (RMSE), mean absolute error method (MAE) or mean absolute proportional error method (MAPE) and historical climate data and historical power generation data for evaluation are used to evaluate the prediction of neural network prediction model 15 Performance. If the difference between the historical predicted power generation predicted by the neural network prediction model 15 and the historical actual power generation is too large, the historical climate data and historical power generation data (sun trajectory or air quality, etc.) input to the neural network prediction model 15 are replaced. Adjusting the architecture or parameter setting of the neural network prediction model 15 , for example, changing the input layer 151 parameters of the neural network prediction model 15 , the number of layers of the hidden layer 152 and the number of nodes 154 , and gradually upgrading the neural network prediction model 15 The predicted performance, and finally the power generation forecasting model with better predictive performance.

In step S304, a predicted climate data of a future time interval, a solar trajectory (such as a height angle or azimuth angle, etc.), and an air quality to prediction module 14 are input to perform power generation amount prediction for a future time interval. Among them, the predicted climate data in the future time can be obtained from the meteorological bureau's forecast data or the relevant climate prediction unit, and the Azimuth Angle A refers to the angle between the projection of the sun's rays on the ground plane and the local meridian, the sun. The height or solar elevation angle is the angle between the line connecting the observer's location and the center of the sun and the ground plane. As shown in Figure 5, the solar azimuth A can be obtained by the following formula:

Where h is the azimuth of the sun, δ _A is the solar declination at the time, and θ _A is the local geographic latitude.

The solar height or solar elevation angle α is the angle between the line connecting the observer's location and the center of the sun and the ground plane. As shown in Fig. 5, the solar elevation angle α can be calculated by the following formula to obtain an approximate value:

Where α is the solar elevation angle and ħ is the time angle under the local star system. Is the current sun declination, It is the local latitude.

Air Mass (AM) is the degree to which the atmosphere affects the Earth's surface to receive sunlight. The formula is:

Where θ _AM is the angle between the incident direction of the sun and the direction of the zenith, as shown in Fig. 6, and θ _{AM is} also called the zenith angle.

The climatic data, the solar azimuth angle, the solar elevation angle, and the air quality of the available future time interval are input to the regenerative energy prediction system 10 of the present creation through the input module 16 to perform power generation prediction of the regenerative power plant, and then in step S305 The predicted power generation amount of the power plant is output through the output module 17. The predicted power generation amount calculated by the prediction module 14 is output via the output module 17, so that the power plant manager can effectively allocate the power generation amount of the power plant.

The regenerative energy prediction method and system of this creation can predict the power generation capacity of the high-accuracy power plant. According to the predicted power generation capacity of the regenerative energy power plant, the power plant manager can adjust the power generation capacity of the power plant to improve the power generation of the renewable energy power plant. Effective utilization.

The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way. Although the present invention has been disclosed above in the preferred embodiment, it is not intended to limit the present invention, and any technique familiar to the art. A person skilled in the art can make some modifications or modifications to equivalent embodiments by using the technical content disclosed above, but the content of the present invention is not deviated from the present invention. Technical simplifications Any simple modifications, equivalent changes and modifications made to the above embodiments are still within the scope of the present technical solution.

10‧‧‧Renewable Energy Forecasting System
11‧‧‧Data Capture Module
12‧‧‧Database
13‧‧‧Anomaly correction module
14‧‧‧ Forecasting Module
15‧‧‧ Neural Network Prediction Model
151‧‧‧Input layer
152‧‧‧ hidden layer
153‧‧‧ Output layer
154‧‧‧ nodes
16‧‧‧Input module
17‧‧‧Output module

Figure 1 is a block diagram of the regenerative energy generation forecasting system of the present invention. Figure 2 is a schematic diagram of a neural network prediction model such as the creation. Figure 3 is a flow chart of the method for predicting the amount of renewable energy generated by the present invention. Fig. 4 is a distribution curve diagram of the normal distribution method of the present creation. Figure 5 is a schematic diagram of the calculation of the solar azimuth angle of the present creation. Figure 6 is a schematic diagram of the calculated air quality of the present creation.

10‧‧‧Renewable Energy Generation Forecasting System

11‧‧‧Data Capture Module

12‧‧‧Database

13‧‧‧Anomaly correction module

14‧‧‧ Forecasting Module

15‧‧‧ Neural Network Prediction Model

16‧‧‧Input module

17‧‧‧Output module

Claims (10)

A renewable energy power generation quantity prediction system, comprising: a database, storing a plurality of historical climate data, a plurality of historical power generation data, a plurality of solar trajectory data and a plurality of air quality data; a data acquisition module, and the data a library connection, the historical climate data, the historical power generation data, the solar trajectory data and the air quality data are extracted from the database; an abnormal point correction module is connected to the data acquisition module, Separating the historical climate data, the historical power generation data, the solar trajectory data, and the plurality of abnormal data in the air quality data; a prediction module connected to the abnormal point correction module, the prediction The module performs training on a neural network prediction model based on the historical climate data, the historical power generation data, the solar trajectory data, and the air quality data that have been corrected for the abnormal data to obtain a power generation prediction model; An input module, connected to the prediction module, inputting a plurality of future climate data, a solar trajectory and an air quality to the pre- Module, so that the predicted module calculating a predicted power generation amount; an output module, connected to the forecast module, for outputting the predicted power generation amount. The regenerative power generation quantity prediction system according to claim 1, wherein the abnormal point correction module applies the regression analysis, the normal distribution method or the data comparison method to perform the historical climate data, the historical power generation data, and the sun. Judgment of the trajectory data or the abnormal data of the air quality data, and filtering or correcting the data points, wherein the correction method is to use the average power generation efficiency of the same power generation device for a period of time as a correction reference, or compare the same time at the same time The average power generation efficiency or average unit power generation of other single or multiple power generation units of the power plant is further supplemented or corrected by the baseline. The regenerative power generation quantity prediction system according to claim 1, wherein the abnormal point correction module is configured to set a threshold value, the historical climate data exceeding the threshold value, the historical power generation data, and the solar trajectory data. Or the air quality data is regarded as the abnormal data. The regenerative energy generation amount prediction system according to claim 1, wherein the solar trajectory includes a solar altitude angle and a solar azimuth angle. The regenerative power generation amount prediction system according to claim 1, wherein the prediction module uses a Root Mean Square Error (RMSE), a Mean Absolute Error (MAE), and an average absolute ratio. Mean Absolute Percentage Error (MAPE) or Mean Relative Error (MRE) to evaluate the predictive performance of this type of neural network prediction model. A method for predicting the amount of renewable energy power generation includes the following steps: obtaining a plurality of historical climate data, a plurality of historical power generation data, a plurality of solar trajectory data, and a plurality of atmospheric quality data through a data acquisition module; The group corrects the historical climate data, the historical power generation data, the solar trajectory data, and the plurality of abnormal data in the air quality data; the historical climates of the abnormal data that have been corrected The data, the historical power generation data, the solar trajectory data and the air quality data are transmitted to a prediction module, and a type of neural network prediction model is trained to obtain a power generation prediction model; and a complex time interval is input The pen predicts the climatological data, a solar trajectory, and a mass of the air mass to the predictive module to predict the power generation of the power plant in the future time interval; and outputs the predicted power generation amount of the power plant through an output module. The method for predicting the amount of renewable energy generated according to claim 6, wherein the historical climate data captured by the abnormal point correction module, the historical power generation data, the solar trajectory data, and the air quality are obtained. The step of filtering out the abnormal data in the data is performed by using a regression analysis or a normal distribution method to perform the historical climate data and the historical power generation data, the solar trajectory data, and the abnormal data of the air quality data. Judging, and performing data point filtering or correction, wherein the correction method can be the same as the average power generation efficiency of the power generating device in the past period of time, or the average power generation efficiency or average unit of other single or multiple power generating devices in the same power plant at the same time. The amount of power generated is further supplemented or corrected as a basis. The method for predicting the amount of renewable energy generated according to claim 6, wherein the historical climate data captured by the abnormal point correction module, the historical power generation data, the solar trajectory data, and the air quality are obtained. The step of filtering out the abnormal data in the data is to set a threshold value, and the historical climate data exceeding the threshold value and the historical power generation data are regarded as the abnormal data. The method for predicting the amount of renewable energy power generation according to claim 6, wherein the historical climate data, the historical power generation data, the solar trajectory data, and the air quality data of the abnormal data that have been filtered out are transmitted. To the prediction module, the training of the neural network prediction model is performed to obtain the power generation prediction model by using Root Mean Square Error (RMSE) and Mean Absolute. Error, MAE), Mean Absolute Percentage Error (MAPE) or Mean Relative Error (MRE) to train this type of neural network prediction model. The method for predicting the amount of renewable energy generated according to claim 6, wherein the historical climate data captured by the abnormal point correction module, the historical power generation data, the solar trajectory data, and the air quality are obtained. The step of correcting the plurality of abnormal data in the data is to use the average power generation efficiency of the same power generation device as the correction reference, or to compare the average power generation efficiency or the average unit power generation amount of the plurality of power generation devices of the same power generation plant at the same time as the reference. Further supplementing or correcting the historical climate data, the historical power generation data, the solar trajectory data, and the plurality of abnormal data in the air quality data.