Photovoltaic power station power generation prediction method based on multi-model fusion

Abstract

The invention relates to the technical field of power generation prediction data processing, and discloses a photovoltaic power station power generation prediction method based on multi-model fusion, which combines a photovoltaic physical model, a power generation statistical model and a power generation prediction model, not only considers a physical rule, but also combines statistical learning and machine learning to improve the stability and accuracy of prediction, and improves the prediction efficiency. And predicting the generating capacity of the photovoltaic power station by using environmental meteorological data, electrical parameters of the photovoltaic power station and historical power generation data. Accurate photovoltaic power generation power prediction is beneficial to better arrangement of power generation plans and resource scheduling of the power grid. The prediction precision and stability are improved through multi-model fusion, so that power grid dispatching is more accurate and efficient, and power grid instability factors caused by photovoltaic power generation fluctuation are reduced. Therefore, photovoltaic power generation can be connected to a power grid more stably, the power demand is met, and wide application and development of clean energy are promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of power generation prediction data processing, and in particular to a photovoltaic power station power generation prediction method based on multi-model fusion. Background Art

[0002] Photovoltaic power generation is an important part of clean energy. The prediction of its power generation is of great significance to the stable operation of the power system, grid dispatching and the operation and maintenance of photovoltaic power stations. Accurate power forecasting can help the power grid better arrange power generation plans and reduce the instability of the power grid caused by photovoltaic power generation fluctuations. It also helps to improve the economic benefits of photovoltaic power stations. The photovoltaic power forecasting system is a system that uses modern digital communication technology, computer technology, real-time meteorological information monitoring technology, and numerical forecasting technology to predict the power generation of photovoltaic power stations. Its main purpose is to predict the power generation changes of photovoltaic power stations in the future through comprehensive analysis, and provide an important basis for power dispatching, network optimization, resource management and fault prevention.

[0003] The existing technology uses a single model prediction method, and a single model can often only use part of the useful information for prediction, while ignoring other information that may be equally important. In the prediction of photovoltaic power generation, this may lead to the model being unable to fully consider the various factors that affect photovoltaic power generation, such as solar radiation, temperature, humidity, wind speed, and equipment status. Due to the lack of comprehensive information processing capabilities, the accuracy and reliability of its prediction results may be affected, and a single model often finds it difficult to achieve high accuracy when predicting photovoltaic power generation. Especially under extreme weather conditions or when equipment fails, the prediction results of a single model may produce large deviations. A single model also has certain limitations in terms of scalability. With the continuous development of photovoltaic power generation technology and the expansion of its application scope, the requirements for prediction models are also getting higher and higher. However, a single model often finds it difficult to meet these new requirements and requires major improvements or redesigns. Summary of the invention

[0004] In order to solve the above problems, the present invention integrates photovoltaic physical models, power generation statistical models and power generation prediction models, and uses environmental meteorological data, photovoltaic power station electrical parameters and historical power generation data to predict the power generation of photovoltaic power stations. Not only physical laws are taken into account, but also statistical learning and machine learning are combined to improve the stability and accuracy of the prediction. Specifically, the present invention provides a photovoltaic power station power generation prediction method based on multi-model fusion, which specifically includes the following steps:

[0005] S1. Data collection and preprocessing: collecting environmental meteorological data, photovoltaic power station electrical parameters, historical power generation data and preprocessing the data;

[0006] S2. Build and fuse multiple models, including:

[0007] Photovoltaic physical model: Based on the photoelectric conversion principle of photovoltaic cells, the photovoltaic physical model is constructed according to factors such as temperature, light intensity, and component efficiency;

[0008] Power generation statistical model: Based on historical power generation data and environmental meteorological data, a time series model is constructed to perform power generation statistics;

[0009] Power generation prediction model: Based on machine learning methods, combined with environmental meteorological data, photovoltaic power station electrical parameters, and power generation statistical models, a power generation prediction model is established;

[0010] Fusion of multiple models through model integration method;

[0011] S3. Train and verify the constructed multiple models, use historical power generation data to train the models, evaluate the prediction performance of the power generation prediction model and adjust and optimize the parameters to improve the prediction accuracy;

[0012] S4, real-time prediction and monitoring, integrating real-time environmental meteorological data interface, ensuring the real-time and accuracy of the input data of the power generation prediction model, updating the prediction results in real time, monitoring the power generation of the power station, timely adjusting the power station operation strategy, and conducting remote monitoring and early warning management;

[0013] S5. Feedback the prediction results and perform multi-model optimization iterations, compare the real-time prediction results of the power generation prediction model with the actual power generation data, analyze the causes of the prediction errors, and continuously optimize the multiple models.

[0014] Furthermore, the environmental meteorological data collected in S1 include but are not limited to light intensity, temperature, humidity, wind speed, cloud cover, etc.; the electrical parameters of the photovoltaic power station include but are not limited to the installation angle, tilt angle, component type, component efficiency, array area, inverter efficiency, etc. of the photovoltaic array; the historical power generation data is the power generation data of the power station over a period of time in the past.

[0015] Furthermore, the data preprocessing in S1 is specifically as follows: first, the data is cleaned, data noise is reduced, missing values ​​and outliers are processed, and then the data is standardized or normalized to ensure that the data can be effectively compared and analyzed.

[0016] Furthermore, the model integration method in S2 includes but is not limited to weighted averaging method and stacking method.

[0017] Furthermore, in S3, the prediction performance of the power generation prediction model is evaluated through cross-validation, AUC, RMSE and other indicators.

[0018] Furthermore, S3 also includes:

[0019] S31. Model training: Use the preprocessed data to train the photovoltaic physical model, power generation statistical model, and power generation prediction model. During the training process, monitor the performance of each model and make adjustments as needed;

[0020] S32, Hyperparameter optimization: Use Bayesian optimization method to adjust the hyperparameters of each model, including learning rate, number of iterations, hidden layer dimension, etc., to find the optimal model configuration;

[0021] S33. Regularize and prevent overfitting: Add regularization terms during training to prevent the model from overfitting, monitor the performance on the validation set through an early stopping strategy, and stop training when the performance no longer improves.

[0022] Furthermore, the continuous optimization of multiple models in S5 specifically includes:

[0023] Adjust model parameters: adjust the hyperparameters of the power generation prediction model according to the verification results, and optimize the parameter settings of the photovoltaic physical model and the power generation statistical model;

[0024] Data quality improvement: Identify and improve deficiencies in data collection and processing processes to reduce data noise and missing values;

[0025] Introducing new features: Based on the prediction effect analysis, new environmental meteorological data or electrical parameters are introduced as model input features.

[0026] Furthermore, the multi-model fusion also includes integration with the power grid system, integrating the prediction results of the power generation prediction model with the power grid dispatching system, optimizing the power grid dispatching strategy, adjusting the power output of the power station according to the prediction results, optimizing the power generation plan, and providing early warning of possible faults based on the power generation prediction model.

[0027] Furthermore, the real-time prediction and monitoring in S4 also includes: regularly generating prediction reports, including power generation prediction results, prediction error analysis, model optimization suggestions, etc., and using data visualization tools to display and monitor power generation prediction results, environmental meteorological data, photovoltaic power station electrical parameters, etc. in the form of charts, dashboards, etc.

[0028] The beneficial effects of the present invention are:

[0029] 1. Improve prediction accuracy Multi-model fusion can make full use of the advantages of different models in information processing and achieve information complementarity. The photovoltaic physical model can be modeled based on the physical principles of photovoltaic cells, considering the impact of physical factors such as solar radiation and temperature on photovoltaic power generation efficiency; the power generation statistical model can explore the periodicity and regularity of photovoltaic power generation through statistical analysis of historical data. By integrating these two models, physical factors and statistical laws can be comprehensively considered to improve the prediction accuracy of the power generation prediction model. Error correction: Multi-model fusion can improve the accuracy of prediction results by error correction between different models. Each model may have certain errors in the prediction process, but the errors of different models may have different characteristics and distribution laws. Through multi-model fusion, the complementarity of these errors can be used for correction, thereby reducing the overall prediction error.

[0030] 2. Enhance prediction stability. A single model may be affected by specific factors and produce large fluctuations during the prediction process, while multi-model fusion can enhance prediction stability by dispersing risks. When a model is subject to abnormal data or external interference, other models can provide relatively stable prediction results, thereby avoiding drastic fluctuations in the overall prediction results. Multi-model fusion can automatically adjust the weights and parameters of each model according to different environmental conditions and data characteristics to adapt to different prediction scenarios and needs. This highly adaptable feature makes the multi-model fusion method more reliable and stable in practical applications.

[0031] 3. Improve generalization ability. Multi-model fusion can cross different disciplines and technical categories, and organically combine models and methods in different fields. This cross-domain fusion can make full use of the professional knowledge and technical advantages of each field to improve the generalization ability of the prediction model. The photovoltaic physical model can be integrated with the power generation prediction machine learning model, and the physical meaning of the physical model and the powerful computing power of the machine learning model can be used to jointly improve the prediction accuracy. Photovoltaic power generation is affected by many complex factors, including weather, temperature, humidity, wind speed, and equipment status. Multi-model fusion can comprehensively consider the interaction between these complex factors, so as to more accurately predict the changing trend of photovoltaic power generation. This ability to cope with complex environments makes the multi-model fusion method more advantageous in practical applications.

[0032] 4. Optimize resource allocation. Accurate photovoltaic power generation prediction helps the power grid to better arrange power generation plans and dispatch resources. By improving the prediction accuracy and stability through multi-model fusion, the power grid dispatch can be more accurate and efficient, and the instability of the power grid caused by photovoltaic power generation fluctuations can be reduced. Photovoltaic power generation can be more stably connected to the power grid and meet electricity demand, thereby promoting the widespread application and development of clean energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a flow chart of a photovoltaic power station power generation prediction method based on multi-model fusion. DETAILED DESCRIPTION

[0034] The following is combined with Figure 1 The preferred embodiments of the present invention are described in detail so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0035] The present invention provides a photovoltaic power station power generation prediction method based on multi-model fusion, which realizes high-precision prediction of photovoltaic power station power generation by integrating the advantages of physical models, statistical models and machine learning models. At the same time, through real-time monitoring, optimization iteration and grid system integration, the stability and practicality of the prediction are further improved. Specifically:

[0036] First, data collection and preprocessing are carried out to collect environmental meteorological data, electrical parameters of photovoltaic power stations, and historical power generation data and preprocess the data; the collected environmental meteorological data include but are not limited to light intensity, temperature, humidity, wind speed, cloud cover, etc.; it covers the key factors affecting photovoltaic power generation, and these data have a direct impact on the power generation efficiency of photovoltaic power stations.

[0037] The electrical parameters of a photovoltaic power station include but are not limited to the installation angle, tilt angle, module type, module efficiency, array area, inverter efficiency, etc. of the photovoltaic array. These parameters reflect the physical characteristics and technical configuration of the power station.

[0038] Historical power generation data refers to the power generation data of the power plant over a period of time in the past. It is the basis for evaluating the performance of the prediction model. Through historical data, we can understand the power generation performance of the power plant under different conditions.

[0039] The specific data preprocessing is as follows: first clean the data, perform data noise reduction, deal with missing values ​​and outliers, and then standardize or normalize the data to ensure the consistency and comparability of the model input and ensure that the data can be effectively compared and analyzed.

[0040] Then, multiple models are constructed and fused, including:

[0041] Photovoltaic physical model: Based on the photoelectric conversion principle of photovoltaic cells, a photovoltaic physical model is constructed according to factors such as temperature, light intensity, and component efficiency. Based on the photoelectric conversion principle of photovoltaic cells, the power generation process of the photovoltaic power station is directly simulated to provide a physical basis for prediction.

[0042] Power generation statistical model: Based on historical power generation data and environmental meteorological data, a time series model is constructed to perform power generation statistics. Using time series analysis technology, the statistical laws of power generation are extracted from historical data to provide data support for prediction.

[0043] Power generation prediction model: Based on machine learning methods, combined with environmental meteorological data, photovoltaic power station electrical parameters, and power generation statistical models, a power generation prediction model is established to perform high-precision power generation prediction.

[0044] Multiple models are integrated through model integration methods; model integration methods include but are not limited to weighted average method and stacking method. Each model has its own unique advantages and limitations, and combining them can improve the accuracy and robustness of predictions.

[0045] The integration of multiple models can be performed by weighted averaging. Specifically, a weight is assigned to the prediction results of each model, and then the weighted prediction results are added together to obtain the final prediction value. The weight assignment is usually based on the performance of the model on historical data, and the model with better performance will receive a higher weight. This method is simple to implement and has high computational efficiency. It can adjust the weight according to the model performance and make full use of the advantages of photovoltaic physical model, power generation statistical model, and power generation prediction model.

[0046] You can also use the stacking method to integrate multiple models. First, train multiple base models using the collected original data sets (environmental meteorological data, photovoltaic power station electrical parameters, historical power generation data). Then generate new features: use the base model to predict the training set, and add the power generation prediction results as new features to the training set. Train the metamodel: Use the training set containing the new features of the power generation prediction results to train the metamodel. The input of the metamodel is the prediction result of the base model, and the output is the final prediction value. Prediction: Use the trained metamodel to predict the power generation of the test set, and then input the prediction results into the trained metamodel to obtain the final prediction results.

[0047] Then, the constructed multiple models are trained and verified, and the models are trained using historical power generation data. The prediction performance of the power generation prediction model is evaluated and the parameters are adjusted and optimized to improve the prediction accuracy.

[0048] The prediction performance of the power generation prediction model is evaluated through cross-validation, AUC, RMSE and other indicators. These indicators help to understand the prediction accuracy and stability of the model and provide a basis for subsequent optimization. Each basic model is trained and tested through the cross-validation method to obtain the prediction error of each model, and a weight is assigned to each model according to the size of the prediction error. Weight allocation usually uses optimization algorithms such as least squares method to minimize the prediction error of the fusion model.

[0049] Model training: Use preprocessed data to train the photovoltaic physical model, power generation statistical model, and power generation prediction model. During the training process, monitor the performance of each model and make adjustments as needed;

[0050] Hyperparameter optimization: Use Bayesian optimization to adjust the hyperparameters of each model, including learning rate, number of iterations, hidden layer dimensions, etc., to find the optimal model configuration;

[0051] Regularization and prevention of overfitting: Add regularization terms (such as L1 or L2 regularization, Dropout, etc.) during training to prevent model overfitting, monitor the performance on the validation set through early stopping strategy, and stop training when the performance no longer improves.

[0052] Then, real-time prediction and monitoring are integrated with real-time environmental meteorological data interfaces to ensure the real-time and accuracy of the input data of the power generation prediction model, and the real-time data collection and transmission are realized through the Internet of Things technology. The prediction results are updated in real time, the power generation of the power station is monitored, the power station operation strategy is adjusted in time, and remote monitoring and early warning management are carried out to timely discover and deal with potential problems. Regularly generate forecast reports, including power generation forecast results, forecast error analysis, model optimization suggestions, etc., and use data visualization tools (such as Tableau, PowerBI, Matplotlib, etc.) to display and monitor power generation forecast results, environmental meteorological data, photovoltaic power station electrical parameters, etc. in the form of charts, dashboards, etc. Regularly generate forecast reports, including power generation forecast results, forecast error analysis, and model optimization suggestions, etc., to facilitate management personnel's decision-making. This helps to quickly understand the operation status of the power station, monitor the prediction accuracy, and discover potential problems or trends.

[0053] Finally, the prediction results are fed back and multi-model optimization iterations are performed. The real-time prediction results of the power generation prediction model are compared with the actual power generation data to analyze the causes of prediction errors, which may be caused by data quality, limitations of model assumptions, uncertainties in environmental changes, etc. The multi-models are continuously optimized.

[0054] Adjust model parameters: adjust the hyperparameters of the power generation prediction model according to the verification results, and optimize the parameter settings of the photovoltaic physical model and the power generation statistical model;

[0055] Data quality improvement: Identify and improve deficiencies in data collection and processing processes to reduce data noise and missing values;

[0056] Introducing new features: Based on the prediction effect analysis, new environmental meteorological data or electrical parameters are introduced as model input features to improve the generalization and prediction capabilities of the model.

[0057] Multi-model fusion also includes integration with the power grid system, integrating the prediction results of the power generation prediction model with the power grid dispatching system, optimizing the power grid dispatching strategy, and improving the power grid's ability to accept renewable energy. Deeply integrate the power generation prediction results of the photovoltaic power station with the power grid dispatching system to achieve coordinated dispatching of photovoltaic power generation and power grid load, and improve the stability and economy of the power grid. Adjust the power output of the power station according to the prediction results, optimize the power generation plan, reduce fluctuations, and improve the stability of the power grid. And according to the power generation prediction model, early warning of possible faults, optimize equipment maintenance plans, and reduce downtime. Based on the prediction results and data analysis, provide decision support for the operation, maintenance, expansion, and transformation of the power station.

[0058] As technology advances and the operating environment of PV power plants changes, the technology and algorithms of the prediction system should be regularly evaluated and updated. As new machine learning algorithms or data processing technologies are developed, the potential improvement of these new technologies on prediction performance can be evaluated and technology upgrades can be carried out in a timely manner.

[0059] Any example of the present invention can be used as an independent technical solution or combined with other examples. All patents and publications mentioned in the specification of the present invention indicate that these are public technologies in the field and can be used in the present invention. All patents and publications cited here are also listed in the references, just like each publication is specifically cited separately. The present invention here can be implemented in the absence of any element or elements, one limitation or multiple limitations, and this limitation is not specifically stated here. The terms and expressions used here are descriptive methods, but not limited by them. There is no intention to indicate that these terms and explanations described in this book exclude any equivalent features, but it can be known that any appropriate changes or modifications can be made within the scope of the present invention and the claims. It can be understood that the embodiments described in the present invention are embodiments and features in some embodiments, and any person skilled in the art can make some changes and variations based on the essence of the present invention, and these changes and variations are also considered to belong to the scope of the present invention and the scope limited by the independent claims and the appended claims.

Claims

1. A photovoltaic power station power generation prediction method based on multi-model fusion, characterized in that: The specific steps include: S1. Data collection and preprocessing: collecting environmental meteorological data, photovoltaic power station electrical parameters, historical power generation data and preprocessing the data; S2. Build and fuse multiple models, including: Photovoltaic physical model: Based on the photoelectric conversion principle of photovoltaic cells, the photovoltaic physical model is constructed according to factors such as temperature, light intensity, and component efficiency; Power generation statistical model: Based on historical power generation data and environmental meteorological data, a time series model is constructed to perform power generation statistics; Power generation prediction model: Based on machine learning methods, combined with environmental meteorological data, photovoltaic power station electrical parameters, and power generation statistical models, a power generation prediction model is established; Fusion of multiple models through model integration method; S3. Train and verify the constructed multiple models, use historical power generation data to train the models, evaluate the prediction performance of the power generation prediction model and adjust and optimize the parameters to improve the prediction accuracy; S4, real-time prediction and monitoring, integrating real-time environmental meteorological data interface, ensuring the real-time and accuracy of the input data of the power generation prediction model, updating the prediction results in real time, monitoring the power generation of the power station, timely adjusting the power station operation strategy, and conducting remote monitoring and early warning management; S5. Feedback the prediction results and perform multi-model optimization iterations, compare the real-time prediction results of the power generation prediction model with the actual power generation data, analyze the causes of the prediction errors, and continuously optimize the multiple models.

2. A photovoltaic power station power generation prediction method based on multi-model fusion according to claim 1, characterized in that: The environmental meteorological data collected in S1 include but are not limited to light intensity, temperature, humidity, wind speed, cloud cover, etc.; the electrical parameters of the photovoltaic power station include but are not limited to the installation angle, tilt angle, component type, component efficiency, array area, inverter efficiency, etc. of the photovoltaic array; the historical power generation data is the power generation data of the power station over a period of time in the past.

3. A photovoltaic power station power generation prediction method and method based on multi-model fusion according to claim 1, characterized in that: The data preprocessing in S1 is specifically as follows: first, the data is cleaned, data noise is reduced, missing values ​​and outliers are processed, and then the data is standardized or normalized to ensure that the data can be effectively compared and analyzed.

4. A photovoltaic power station power generation prediction method and method based on multi-model fusion according to claim 1, characterized in that: The model integration method in S2 includes but is not limited to weighted average method and stacking method.

5. A photovoltaic power station power generation prediction method and method based on multi-model fusion according to claim 1, characterized in that: In S3, the prediction performance of the power generation prediction model is evaluated by cross-validation, AUC, RMSE and other indicators.

6. A photovoltaic power station power generation prediction method and method based on multi-model fusion according to claim 1, characterized in that: The S3 also includes: S31. Model training: Use the preprocessed data to train the photovoltaic physical model, power generation statistical model, and power generation prediction model. During the training process, monitor the performance of each model and make adjustments as needed; S32, Hyperparameter optimization: Use Bayesian optimization method to adjust the hyperparameters of each model, including learning rate, number of iterations, hidden layer dimension, etc., to find the optimal model configuration; S33. Regularize and prevent overfitting: Add regularization terms during training to prevent the model from overfitting, monitor the performance on the validation set through an early stopping strategy, and stop training when the performance no longer improves.

7. A photovoltaic power station power generation prediction method and method based on multi-model fusion according to claim 1, characterized in that: The continuous optimization of multiple models in S5 specifically includes: Adjust model parameters: adjust the hyperparameters of the power generation prediction model according to the verification results, and optimize the parameter settings of the photovoltaic physical model and the power generation statistical model; Data quality improvement: Identify and improve deficiencies in data collection and processing processes to reduce data noise and missing values; Introducing new features: Based on the prediction effect analysis, new environmental meteorological data or electrical parameters are introduced as model input features.

8. A photovoltaic power station power generation prediction method and method based on multi-model fusion according to claim 1, characterized in that: The multi-model fusion also includes integration with the power grid system, integrating the prediction results of the power generation prediction model with the power grid dispatching system, optimizing the power grid dispatching strategy, adjusting the power output of the power station according to the prediction results, optimizing the power generation plan, and providing early warning of possible faults based on the power generation prediction model.

9. A photovoltaic power station power generation prediction method and method based on multi-model fusion according to claim 1, characterized in that: The real-time prediction and monitoring in S4 also includes: regularly generating prediction reports, including power generation prediction results, prediction error analysis, model optimization suggestions, etc., and using data visualization tools to display and monitor power generation prediction results, environmental meteorological data, photovoltaic power station electrical parameters, etc. in the form of charts, dashboards, etc.

Images