Photovoltaic power generation power prediction method based on micrometeorological model

Abstract

The invention relates to a photovoltaic power generation power prediction method based on a micrometeorological model, and the method comprises the steps: obtaining the historical power generation power data of target photovoltaic power generation equipment and corresponding environment monitoring data, and carrying out the preprocessing and correlation detection of the obtained data; therefore, non-corresponding historical generation power data and environment monitoring data are removed. Thirdly, constructing an information database, and obtaining a micrometeorological model by adopting a crisscross optimization mode based on the environment monitoring data in the information database; then, the detected numerical value is input into the micro-meteorological model, so that a predicted meteorological value of the photovoltaic power station is obtained, and finally, according to the predicted meteorological value, environment monitoring data similar to the predicted meteorological value and historical power generation power data corresponding to the environment monitoring data are selected in an information database; therefore, through normalization and weighted average processing, an actual photovoltaic power generation power prediction value can be obtained. The method has the advantages of being convenient to use, accurate and efficient.

Description

Technical Field

[0001] This invention belongs to the field of photovoltaic power generation technology, specifically relating to a method for predicting photovoltaic power generation based on a micro-meteorological model. Background Technology

[0002] Photovoltaics, or photovoltaic power generation systems, are power generation systems that utilize the photovoltaic effect of semiconductor materials to convert solar radiation energy into electrical energy. The photovoltaic power generation process does not pollute the environment or damage the ecosystem; it is a clean, safe, and renewable energy source. However, due to the influence of solar radiation intensity, photovoltaic module temperature, weather, and other random factors, the operation of a photovoltaic power generation system is a non-equilibrium stochastic process. Its power generation and output are highly random, fluctuate greatly, and are difficult to control, which is particularly pronounced during sudden weather changes. Therefore, to avoid the impact of photovoltaic power generation systems on grid stability and dispatch management, it is necessary to propose a photovoltaic power generation prediction method based on micro-meteorological models to accurately predict the power output of photovoltaic power generation systems. This allows for the implementation of corresponding technical measures to smooth out photovoltaic power generation fluctuations and improve the safety and stability of the power system. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide a convenient, accurate and efficient method for predicting photovoltaic power generation based on a micro-meteorological model.

[0004] The objective of this invention is achieved as follows: a photovoltaic power generation prediction method based on a micro-meteorological model, comprising the following steps: Step 1: Obtain historical power generation data of the target photovoltaic power generation equipment, and then, based on the time points of the historical power generation data, obtain environmental monitoring data corresponding to the historical power generation data; Step 2: Preprocess the data obtained in Step 1 to remove error messages generated during the transition when the operating mode changes; Step 3: Detect the correlation between the preprocessed historical power generation data and environmental monitoring data, and remove historical power generation data and environmental monitoring data that do not correspond. Then, perform noise reduction, integration registration and structuring processing on the remaining data to increase the compactness and density of the data while reducing the amount of data transmission. Step 4: Based on the historical power generation data and environmental monitoring data detected in Step 3, an information database is constructed. Then, based on the environmental monitoring data within the information database, the initial neural network is trained using a cross-optimization approach to obtain the micro-meteorological model. Step 5: Detect the environment around the photovoltaic power station using a detection device, and input the detected values ​​into the micro-meteorological model to obtain the predicted meteorological values ​​for the photovoltaic power station; Step 6: Based on the predicted meteorological value, select environmental monitoring data that is similar to it from the information database. After the environmental monitoring data is selected, obtain the historical power generation data corresponding to the selected environmental monitoring data from the information database. Then, normalize and weighted average the obtained historical power generation data to obtain the actual photovoltaic power generation prediction value.

[0005] Furthermore, the environmental monitoring data in step 1 includes solar irradiance, ambient temperature, relative humidity, total cloud cover, and horizontal visibility.

[0006] Furthermore, the integration registration and structuring processing in step 3 involves first using fuzzy rough sets to reduce the environmental attributes in the environmental monitoring data, and then normalizing the historical power generation data corresponding to the reduced environmental monitoring data.

[0007] Furthermore, the correlation detection of historical power generation data and environmental monitoring data in step 3 specifically involves: obtaining a first correlation coefficient based on the historical power generation data and environmental monitoring data; and then classifying data whose absolute value of the first correlation coefficient is lower than a threshold as having no corresponding data.

[0008] Furthermore, in step 4, when training the initial neural network, a maximum number of iterations is set, and the difference between the environmental prediction value and the sample label of the training sample in each iteration is specified as the loss function.

[0009] The beneficial effects of this invention are as follows: This invention acquires historical power generation data of the target photovoltaic power generation equipment and corresponding environmental monitoring data. Then, it preprocesses and performs correlation detection on the acquired data to remove historical power generation data and environmental monitoring data without corresponding data. Next, it integrates, registers, and structures the remaining data, specifically by using fuzzy rough set theory to reduce the environmental attributes in the environmental monitoring data and normalizing the historical power generation data corresponding to the reduced environmental monitoring data. Finally, based on the detected historical power generation data and environmental monitoring data, an information database is constructed. Then, based on the environmental monitoring data within the information database, a cross-optimization approach is used to train an initial neural network, thereby obtaining a micro-meteorological model. After completing the above operations, the environment around the photovoltaic power station is monitored by a detection device, and the detected values ​​are input into a micro-meteorological model to obtain the predicted meteorological value for the photovoltaic power station. Finally, based on the predicted meteorological value, environmental monitoring data that is similar to it is selected from the information database. After the environmental monitoring data is selected, the historical power generation data corresponding to the selected environmental monitoring data is obtained from the information database. Thus, by normalizing and weighting the obtained historical power generation data, the actual predicted value of photovoltaic power generation can be obtained. This invention, by adopting this structure, can avoid the problem of photovoltaic power generation systems affecting the stability and dispatch management of the power grid by connecting to the grid, thereby improving the safety and stability of the power system. In summary, this invention has the advantages of being convenient to use and being accurate and efficient. Detailed Implementation

[0010] The present invention will now be further described.

[0011] Example: A method for predicting photovoltaic power generation based on a micro-meteorological model, comprising the following steps: Step 1: Obtain historical power generation data of the target photovoltaic power generation equipment, and then, based on the time points of the historical power generation data, obtain environmental monitoring data corresponding to the historical power generation data; the environmental monitoring data includes solar irradiance, ambient temperature, relative humidity, total cloud cover and horizontal visibility; Step 2: Preprocess the data obtained in Step 1 to remove error messages generated during the transition when the operating mode changes; Step 3: Detect the correlation between the preprocessed historical power generation data and environmental monitoring data, and remove historical power generation data and environmental monitoring data without corresponding data. Then, perform denoising, integration registration, and structuring processing on the remaining data to increase the compactness and density of the data while reducing the amount of data transmission. Specifically, the correlation detection between historical power generation data and environmental monitoring data involves obtaining a first correlation coefficient based on the historical power generation data and environmental monitoring data. Then, data with an absolute value of the first correlation coefficient lower than a threshold are classified as having no corresponding data. The integration registration and structuring processing involves first using fuzzy rough set theory to reduce the environmental attributes in the environmental monitoring data, and then normalizing the historical power generation data corresponding to the reduced environmental monitoring data.

[0012] Step 4: Based on the historical power generation data and environmental monitoring data detected in Step 3, an information database is constructed. Then, based on the environmental monitoring data within the information database, the initial neural network is trained using a cross-optimization approach to obtain a micro-meteorological model. During this process, when training the initial neural network, a maximum number of iterations is set, and the difference between the environmental prediction value and the sample label of the training sample in each iteration is specified as the loss function.

[0013] Step 5: Detect the environment around the photovoltaic power station using a detection device, and input the detected values ​​into the micro-meteorological model to obtain the predicted meteorological values ​​for the photovoltaic power station; Step 6: Based on the predicted meteorological value, select environmental monitoring data that is similar to it from the information database. After the environmental monitoring data is selected, obtain the historical power generation data corresponding to the selected environmental monitoring data from the information database. Then, normalize and weighted average the obtained historical power generation data to obtain the actual photovoltaic power generation prediction value.

[0014] In the use of this invention, firstly, historical power generation data of the target photovoltaic power generation equipment and corresponding environmental monitoring data are acquired. Then, the acquired data undergoes preprocessing and correlation detection to remove historical power generation data and environmental monitoring data without corresponding data. The preprocessing operation involves removing error information generated during transitional states when the operating mode changes. The correlation detection of historical power generation data and environmental monitoring data specifically involves obtaining a first correlation coefficient based on the historical power generation data and environmental monitoring data, and then classifying data with an absolute value below a threshold as having no corresponding data. Next, the remaining data undergoes integration, registration, and structuring processing. This involves using fuzzy rough sets to reduce the environmental attributes in the environmental monitoring data, and normalizing the historical power generation data corresponding to the reduced environmental monitoring data, thereby increasing the compactness and density of the data while reducing the amount of data transmitted. Finally, based on the completed historical power generation data and environmental monitoring data, an information database is constructed, and subsequently, based on the environmental monitoring data within the information database... According to this invention, a cross-optimization approach is used to train an initial neural network to obtain a micro-meteorological model. During training, a maximum number of iterations is set, and the difference between the predicted environmental value and the sample label of the training sample in each iteration is used as the loss function. After this operation, the environment around the photovoltaic power station is detected by a detection device, and the detected values ​​are input into the micro-meteorological model to obtain the predicted meteorological value for the photovoltaic power station. Finally, based on the predicted meteorological value, environmental monitoring data that approximates it is selected from the information database. After the environmental monitoring data selection is completed, historical power generation data corresponding to the selected environmental monitoring data is obtained from the information database. Thus, by normalizing and weighted averaging the obtained historical power generation data, the actual predicted photovoltaic power generation value can be obtained. This invention, using this structure, can avoid the problem of photovoltaic power generation systems affecting grid stability and dispatch management when connected to the grid, thereby improving the safety and stability of the power system. In summary, this invention has the advantages of being convenient to use and highly accurate and efficient.

[0015] The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A method for photovoltaic power generation power prediction based on a microclimate model, characterized by, The method comprises the following steps: Step 1: obtaining historical power generation data of a target photovoltaic power generation device, and then, based on time nodes of the historical power generation data, obtaining environmental monitoring data corresponding to the historical power generation data; Step 2: pre-processing the data obtained in step 1 to remove error information generated in a transition state when a running mode is changed; Step 3: detecting the relevance of the historical power generation data and the environmental monitoring data after pre-processing, removing data without corresponding historical power generation data and environmental monitoring data, and then, performing denoising, integration registration and structural processing on the remaining data to increase the compactness and density of the data while reducing the transmission amount of the data; Step 4: based on the historical power generation data and the environmental monitoring data after detection in step 3, constructing an information database, and then, based on the environmental monitoring data inside the information database, training an initial neural network in a vertical and horizontal cross-optimization manner to obtain a micro-meteorological model; Step 5: detecting the environment around the photovoltaic power station through a detection device and inputting the detected values into the micro-meteorological model to obtain predicted meteorological values of the photovoltaic power station; Step 6: selecting environmental monitoring data similar to the predicted meteorological values in the information database, and after the selection of the environmental monitoring data is completed, obtaining historical power generation data corresponding to the selected environmental monitoring data in the information database; then, performing normalization and weighted average processing on the obtained historical power generation data to obtain actual photovoltaic power generation power prediction values.

2. The method for photovoltaic power generation power prediction based on microclimate model according to claim 1, characterized in that: The environmental monitoring data in step 1 comprises solar irradiance, environmental temperature, relative humidity, total cloud cover and horizontal visibility.

3. The method for photovoltaic power generation power prediction based on microclimate model according to claim 1, characterized in that: The integration registration and structural processing in step 3 are to first perform attribute reduction on environmental attributes in the environmental monitoring data by using a fuzzy rough set, and then perform normalization processing on historical power generation data corresponding to the reduced environmental monitoring data.

4. The method for photovoltaic power generation power prediction based on microclimate model according to claim 1, characterized in that: The relevance detection of the historical power generation data and the environmental monitoring data in step 3 is to obtain a first correlation coefficient according to the historical power generation data and the environmental monitoring data, and then, take the absolute value of the first correlation coefficient below a threshold value as a non-corresponding class.

5. The method for photovoltaic power generation power prediction based on microclimate model according to claim 1, characterized in that: In step 4, when training the initial neural network, a maximum number of iterations is set, and the difference between the environmental prediction value and the sample label of the training sample in each iteration is specified as a loss function.