Photovoltaic power prediction method based on DNN-physical combination model

Abstract

The invention discloses a photovoltaic power prediction method based on a DNN-physical combination model. The method comprises the following steps: S1, acquiring historical power and meteorological data of a region; s2, the given data set is preprocessed, and whether the data set is complete or not is judged; s3, weight factors for building the DNN model and the physical model are distributed, the weight factors tend to the DNN model under the condition that the data set is complete, the factors distributed to the physical model tend to the physical model if the data set is incomplete; s4, a parallel modeling strategy is adopted for the DNN model and the physical model, and model building and verification evaluation are carried out at the same time; s5, combining the DNN model and the physical model into a hybrid model for evaluation and optimization; and S6, outputting a photovoltaic power prediction result based on the hybrid model. According to the method, the limitation of a single model on data processing and prediction capability is overcome, and adaptive weight distribution under different data set conditions is realized, so that the accuracy and stability of photovoltaic power prediction are improved.

Description

Technical Field

[0001] The present invention belongs to the field of new energy prediction for electric power systems, and specifically relates to a photovoltaic power prediction method based on a DNN-physical combined model. Background Art

[0002] As a clean, renewable and unlimited energy source, solar energy has no pollution and noise in its power generation process, which can significantly reduce greenhouse gas emissions and environmental pollution, and help improve air quality and water resource protection. In addition, solar power generation can promote economic development, create jobs, and improve energy security and independence. Among solar power generation, photovoltaic power generation is one of the most important forms and has become a hot topic of research in countries around the world. Photovoltaic power generation systems are affected by multiple natural factors such as solar radiation intensity, temperature, humidity and cloud cover, showing significant randomness and nonlinear characteristics, which makes the prediction of photovoltaic power generation system output power extremely challenging. And this volatility and randomness are exacerbated by the constant changes in meteorological and environmental conditions. These characteristics have brought adverse effects on the grid connection of large-scale photovoltaic power generation and posed a threat to the stable operation of the power grid.

[0003] The present application is different from the prior art as follows:

[0004] Technical comparison with patent CN112507603B "Identification method for extreme scenarios of robust optimization of power system based on DNN algorithm"

[0005] 1. The input part of the data set of the DNN algorithm constructed by patent CN112507603B is mainly composed of box-type uncertainty sets, including m load powers and n photovoltaic power generation output parameters. The input layer of our DNN model specially designed five neurons to cover key influencing factors such as ambient temperature, system capacity, wind speed, solar irradiance, photovoltaic panel type, and additional ambient humidity. This makes the design close to the actual physical process of photovoltaic power generation and accurately transmits environmental parameters for calculation.

[0006] Second, patent CN112507603B is a method for identifying extreme scenarios of robust optimization of power systems based on DNN algorithms, which can simplify the steps of solving extreme scenarios to improve the efficiency of identifying extreme scenarios. The photovoltaic power prediction method based on the DNN-physical combined model designed by us combines the physical meaning clarity of the physical model with the high-precision characteristics of the DNN model, aiming to achieve a dual improvement in the prediction accuracy and practicality of photovoltaic power technology.

[0007] Technical comparison with patent CN117708716B "Photovoltaic power prediction method and equipment based on regression and time series fusion"

[0008] 1. Patent CN 117708716B adopts a prediction method that combines random forest, histogram gradient boosting regression model HGBR and deep neural network DNN, and integrates the regression results of the above models through Stacking integration technology. We innovatively integrate physical models with clear physical meanings and high-precision DNN models. In the prediction process, we implement dynamic weight analysis on the two methods: when data is scarce or DNN performance is limited, increase the weight of the physical model, guided by physical principles, to make up for the limitations of the data-driven model; when data is sufficient, increase the weight of the DNN model. This strategy ensures efficient and accurate predictions in different scenarios.

[0009] Second, patent CN 117708716B uses the Informer model for long-term time series forecasting, using large-scale data sets and long sequence information to improve the accuracy and efficiency of forecasting. The combined model we designed can not only achieve accurate forecasting when the amount of data is sufficient, but also show a strong forecasting ability enhancement effect when the amount of data is scarce, effectively coping with the complexity and uncertainty in photovoltaic power forecasting.

[0010] Technical comparison with patent CN118643328A "Power transfer learning prediction method for new photovoltaic stations based on coupled irradiance and temperature"

[0011] 1. Patent CN118643328A analyzes the principle of photovoltaic modules absorbing solar irradiance, and first physically models the module to extract the slope irradiance feature. After the physical modeling extracts the feature information, the DNN model is trained to further extract the temperature features, and then used in combination with other models for prediction. We specifically carry out precise physical modeling for the data provided by the manufacturer, and cleverly integrate the DNN model with it, adopt a collaborative strategy, model in parallel at the same time, and then assign weights to combine into a hybrid model for power prediction, which not only retains the accuracy of the physical model, but also gives play to the advantages of DNN in complex feature extraction. It significantly reduces the dependence on field-collected data and improves prediction efficiency and accuracy.

[0012] 2. Patent CN118643328A In order to predict photovoltaic power, it is necessary to first obtain relevant parameter data through the slope irradiance feature extraction module. However, in the DNN-physics combination model designed by us, the physical model part does not rely on a large amount of data to operate efficiently. It only relies on the manufacturer's data sheet that is easily accessible to the manufacturer, which greatly reduces the burden of data collection and improves the convenience and practicality of the model.

[0013] 3. Patent CN118643328A builds a prediction model by integrating multiple feature extraction modules. First, the model is pre-trained on the data of photovoltaic power station A, and then the model parameters are adaptively fine-tuned based on the data of photovoltaic power station B, and finally the power generation of photovoltaic power station B in the future period is accurately predicted. The DNN-physics combined model we designed is based on the deep learning of historical data by the DNN model, combined with the precise description of photovoltaic panels by the physical model. This combined model can more effectively capture the dynamic changes in the output power of photovoltaic power stations and effectively deal with the randomness and intermittency of meteorological conditions. This method reduces the dependence on massive historical data. Summary of the invention

[0014] In response to the above problems, the present invention proposes a photovoltaic power prediction method based on a DNN-physical combined model. This method combines the physical meaning clarity of the physical model and the high-precision characteristics of the DNN model, aiming to achieve a dual improvement in prediction accuracy and practicality. Specifically, the physical model, as a supplement, can provide guidance based on physical principles when data is scarce or the DNN model performs poorly, ensuring the stability and reliability of the prediction results. At the same time, the DNN model uses its powerful data processing and feature extraction capabilities to accurately capture the complex characteristics of the photovoltaic system in actual operation, such as spectral loss, dirt accumulation, and component mismatch, thereby correcting the deviation of the physical model under ideal conditions and improving prediction accuracy.

[0015] To achieve the above object, the technical solution adopted by the present invention is:

[0016] The photovoltaic power prediction method based on the DNN-physical combined model is characterized by comprising the following steps:

[0017] Step S1: Obtain historical power and meteorological data of the area;

[0018] Step S2: preprocess the given data set and determine whether the data set is complete;

[0019] Step S3: Allocate weight factors for building the DNN model and the physical model. When the data set is complete, the weight factors tend to favor the DNN model. If the data set is incomplete, the weight factors allocated to the physical model should be distributed in a way that favors the physical model.

[0020] Step S4: The DNN model and the physical model adopt a parallel modeling strategy, and the model construction and verification evaluation are carried out simultaneously;

[0021] Step S5: The DNN model and the physical model are combined into a hybrid model for evaluation and optimization;

[0022] Step S6: Based on the hybrid model, output the photovoltaic power prediction result.

[0023] As a further improvement of the present invention, step S3 is specifically as follows: weight factors of the DNN model and the physical model are allocated according to the completeness of the data set. When the data set is complete, the weight factor is more inclined to the DNN model to take advantage of its strong data fitting ability; when the data set is incomplete, the weight of the physical model is increased to take advantage of its prediction stability and reliability based on physical principles. This dynamic weight allocation strategy helps to improve the overall prediction performance.

[0024] As a further improvement of the present invention, the specific step S4 is specifically:

[0025] The parallel modeling method and strategy are adopted to build the physical model and DNN model together. According to the data provided by the manufacturer, the physical model is built as follows:

[0026]

[0027] As a further improvement of the present invention, the DNN model is used for photovoltaic power prediction, specifically as follows: five neurons are specially designed in the model input layer to cover key influencing factors such as ambient temperature, system capacity, wind speed, solar irradiance, photovoltaic panel type, and additional ambient humidity. The input data is divided into numerical and categorical types. The categorical data is encoded by OneHotEncoder and input. The data is processed and transformed in the hidden layer of the DNN, the neuron output is standardized by the activation function, and the data stream is formed by layer-by-layer transmission. Finally, the photovoltaic power prediction result is generated in the output layer to verify the accuracy of the model.

[0028] As a further improvement of the present invention, step S5 is specifically as follows: organically integrating the two, giving full play to the advantages of the DNN model in processing complex nonlinear relationships, and the accuracy of the physical model in explaining physical phenomena, thereby effectively alleviating the limitations of each of them, and then strictly verifying the accuracy of the hybrid model.

[0029] As a further improvement of the present invention, step S6 specifically includes: comprehensively analyzing the performance of the hybrid model through evaluation indicators, and optimizing and adjusting the model accordingly. Finally, the model successfully outputs accurate photovoltaic power prediction results, demonstrating its superior performance in the field of photovoltaic power prediction.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] This method combines the powerful data processing capabilities of DNN and the accuracy of physical models, and can significantly improve the accuracy and stability of photovoltaic power forecasting. Through DNN's deep learning of historical data and the precise description of photovoltaic panels by physical models, this combined model can better capture the dynamic characteristics of photovoltaic power station output power and effectively deal with the randomness and intermittent effects of meteorological conditions such as solar radiation and temperature. In addition, this method also reduces the dependence on a large amount of historical data, and is suitable for power forecasting of newly built and existing photovoltaic power stations, providing strong support for the safe and stable operation of power systems and energy management. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 is a flow chart of the method of the present invention;

[0033] Figure 2 is a DNN model framework diagram for photovoltaic prediction in one embodiment of the present invention;

[0034] Figure 3 is the corresponding photovoltaic output curve in one embodiment of the present invention. DETAILED DESCRIPTION

[0035] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0036] Flowchart as Figure 1 As shown, a photovoltaic power prediction method based on a DNN-physical combined model is characterized by comprising the following steps:

[0037] Step S1: Obtain historical power and meteorological data of the area;

[0038] Step S2: preprocess the given data set and determine whether the data set is complete;

[0039] Step S3: Allocate weight factors for building the DNN model and the physical model. When the data set is complete, the weight factors tend to favor the DNN model. If the data set is incomplete, the weight factors allocated to the physical model should be distributed in a way that favors the physical model.

[0040] The step S3 is specifically: assigning weight factors of the DNN model and the physical model according to the completeness of the data set. When the data set is complete, the weight factor is more inclined to the DNN model to take advantage of its strong data fitting ability; when the data set is incomplete, the weight of the physical model is increased to take advantage of its prediction stability and reliability based on physical principles. This dynamic weight allocation strategy helps to improve the overall prediction performance.

[0041] Step S4: The DNN model and the physical model adopt a parallel modeling strategy, and the model construction and verification evaluation are carried out simultaneously;

[0042] The specific step S4 is specifically: using the parallel modeling method and strategy to build the physical model and the DNN model together. According to the data provided by the manufacturer, the physical model is built as follows:

[0043]

[0044] The DNN model is used for photovoltaic power prediction. The DNN model framework is shown in the figure below. Figure 2 As shown in the figure. Five neurons are specially designed in the model input layer to cover key influencing factors such as ambient temperature, system capacity, wind speed, solar irradiance, photovoltaic panel type, and additional ambient humidity. The input data is divided into numerical and categorical types, and the categorical data is encoded by OneHotEncoder before input. The data is processed and transformed in the hidden layer of the DNN, the neuron output is standardized by the activation function, and is passed layer by layer to form a data stream. Finally, the photovoltaic power prediction result is generated in the output layer to verify the accuracy of the model.

[0045] Step S5: The DNN model and the physical model are combined into a hybrid model for evaluation and optimization;

[0046] The step S5 is specifically: organically integrating the two. By giving full play to the advantages of the DNN model in processing complex nonlinear relationships and the accuracy of the physical model in explaining physical phenomena, the limitations of each are effectively reduced. Subsequently, the hybrid model is rigorously verified for accuracy.

[0047] Step S6: Based on the hybrid model, output the photovoltaic power prediction result.

[0048] The step S6 is specifically: comprehensively analyzing the performance of the hybrid model through evaluation indicators, and optimizing and adjusting the model accordingly. Finally, the model successfully outputs accurate photovoltaic power prediction results, demonstrating its superior performance in the field of photovoltaic power prediction.

[0049] As a specific embodiment of the present invention, the present invention provides a photovoltaic power prediction method based on a DNN-physical combined model, comprising the following steps:

[0050] Step S1: Obtain historical electricity and meteorological data for the area:

[0051] In this example, the photovoltaic power prediction problem in a certain region is carried out. In order to predict the local photovoltaic power situation, the historical power and meteorological data of the region are first collected comprehensively, including temperature, humidity, cloud cover, solar irradiance, etc., and the type and capacity of photovoltaic panels and wind speed at the installation site are recorded. The data collection time resolution is 20 minutes, aiming to capture subtle changes and provide a rich and accurate information basis for subsequent model training.

[0052] Step S2: Preprocess the given data set and determine whether the data set is complete:

[0053] In this example, after obtaining the original data set, a series of preprocessing tasks such as data cleaning and removing missing values ​​are required to improve data quality and model training results. It is also necessary to evaluate the integrity of the data set to check whether it contains all necessary parameters and information, as well as the continuity of the time series.

[0054] Step S3: Allocate weight factors for building the DNN model and the physical model. When the data set is complete, the weight factors tend to favor the DNN model. If the data set is incomplete, the weight factors assigned to the physical model tend to favor the physical model:

[0055] In this example, when the data set is relatively complete, the DNN model will be given a higher weight due to its strong data fitting ability to make full use of the laws it has learned from large-scale data. For example, when the data set is complete, the weight factors of the DNN model and the physical model are 0.9352 and 0.0648 respectively. On the contrary, if the data set is incomplete or there are many missing values, the physical model will get more weight due to its prediction stability and reliability based on physical principles. This dynamic weight allocation strategy aims to balance the advantages of the two models to adapt to the prediction needs under different data set conditions. For example, when the data set is incomplete, the weight factors of the DNN model and the physical model are 0.8087 and 0.1913 respectively.

[0056] Step S4: The DNN model and the physical model adopt a parallel modeling strategy, and the model construction and verification evaluation are carried out simultaneously:

[0057] In this example, after determining the weight factors of the DNN model and the physical model, a parallel modeling strategy is used to build the model. The two models are independent of each other during the training process, but will be combined and optimized later to achieve better prediction results.

[0058] Step S5: The DNN model and the physical model are combined into a hybrid model for evaluation and optimization:

[0059] In this example, the prediction performance of the hybrid model is verified by comparing it with the actual observed data to evaluate its accuracy and stability. If the prediction results are biased or insufficient, the parameters of the hybrid model need to be adjusted and optimized to improve its prediction ability.

[0060] Step S6: Based on the hybrid model, output the photovoltaic power prediction result:

[0061] Figure 3It is the prediction curve (the red line indicates fitting) and model evaluation index when the data set is incomplete. After the training and optimization in the above steps, the hybrid model based on the DNN-physical combination model has the ability to accurately predict the photovoltaic power. At this time, new meteorological data and photovoltaic system parameters can be input and the hybrid model can be used for real-time prediction. The prediction results will include the photovoltaic output power value in the future (such as hourly or daily level) and the corresponding prediction error range.

[0062] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. A photovoltaic power prediction method based on a DNN-physical combined model, characterized by: The steps include: Step S1: Obtain historical power and meteorological data of the area; Step S2: preprocess the given data set and determine whether the data set is complete; Step S3: Allocate weight factors for building the DNN model and the physical model. When the data set is complete, the weight factors tend to favor the DNN model. If the data set is incomplete, the weight factors allocated to the physical model should be distributed in a way that favors the physical model. Step S4: The DNN model and the physical model adopt a parallel modeling strategy, and the model construction and verification evaluation are carried out simultaneously; Step S5: The DNN model and the physical model are combined into a hybrid model for evaluation and optimization; Step S6: Based on the hybrid model, output the photovoltaic power prediction result.

2. The photovoltaic power prediction method based on the DNN-physical combined model according to claim 1 is characterized in that: The step S3 is specifically as follows: weight factors of the DNN model and the physical model are allocated according to the completeness of the data set. When the data set is complete, the weight factor is more inclined to the DNN model to utilize its powerful data fitting ability; when the data set is incomplete, the weight of the physical model is increased to utilize its prediction stability and reliability based on physical principles.

3. The photovoltaic power prediction method based on the DNN-physical combined model according to claim 1 is characterized in that: The specific step S4 is specifically as follows: The parallel modeling method and strategy are adopted to build the physical model and DNN model together. According to the data provided by the manufacturer, the physical model is built as follows:

4. The photovoltaic power prediction method based on the DNN-physical combined model according to claim 3 is characterized in that: The DNN model is used for photovoltaic power prediction. Specifically, five neurons are specially designed in the model input layer to cover key influencing factors such as ambient temperature, system capacity, wind speed, solar irradiance, photovoltaic panel type, and additional ambient humidity. The input data is divided into numerical and categorical types. The categorical data is encoded by OneHotEncoder and input. The data is processed and transformed in the hidden layer of the DNN, and the neuron output is standardized by the activation function, and is passed layer by layer to form a data stream. Finally, the photovoltaic power prediction result is generated in the output layer to verify the accuracy of the model.

5. The photovoltaic power prediction method based on the DNN-physical combined model according to claim 1 is characterized in that: The step S5 is specifically as follows: organically integrating the two, giving full play to the advantages of the DNN model in processing complex nonlinear relationships, and the accuracy of the physical model in explaining physical phenomena, effectively reducing the limitations of each, and then strictly verifying the accuracy of the hybrid model.

6. The photovoltaic power prediction method based on the DNN-physical combined model according to claim 1, characterized in that: The step S6 specifically includes: comprehensively analyzing the performance of the hybrid model through evaluation indicators, and optimizing and adjusting the model accordingly. Finally, the model successfully outputs accurate photovoltaic power prediction results, demonstrating its superior performance in the field of photovoltaic power prediction.

Images