Distributed photovoltaic power generation power prediction method

Abstract

The invention relates to a distributed photovoltaic power generation power prediction method, which comprises the following steps of: preprocessing collected historical power output data and meteorological data, segmenting and splitting the data according to unit time so as to obtain a plurality of storage packets of the power output data and the meteorological data according to the unit time, and storing the plurality of storage packets of the power output data and the meteorological data into a database; the method comprises the following steps of: integrating data in each storage packet, constructing a network model, training the network model to obtain a prediction model, finally, changing a numerical value of unit time to construct a plurality of prediction models corresponding to the numerical value of the unit time, and finally, constructing a plurality of prediction models corresponding to the numerical value of the unit time. According to the method, meteorological data information of a prediction period is divided according to numerical values of unit time in sequence and then substituted into prediction models corresponding to the numerical values of the unit time, so that a plurality of power prediction values are obtained, and finally, the plurality of power prediction values are compared and analyzed to obtain a final power prediction value. The method has the advantages of being accurate, efficient and high in applicability.

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy power generation, and particularly relates to a method for predicting the power of distributed photovoltaic power generation. Background Art

[0002] Photovoltaics, that is, a photovoltaic power generation system, is a power generation system that uses the photovoltaic effect of semiconductor materials to convert solar radiant energy into electrical energy. The photovoltaic power generation process does not pollute the environment or damage the ecology, and is a clean, safe and renewable energy source; however, due to the influence of solar radiation intensity, photovoltaic module temperature, weather and some random factors, the operation process of the photovoltaic power generation system is a non-equilibrium random process, and its power generation and output power have strong randomness, large fluctuations and are not easy to control, which is particularly prominent when the weather suddenly changes; at present, with more and more megawatt-level photovoltaic power generation systems connected to the grid for operation, many problems are brought to the stability and dispatching management of the grid; therefore, it is necessary to propose a method for predicting the power of distributed photovoltaic power generation to accurately predict the power of the photovoltaic power generation system, so as to take corresponding technical measures to smooth the power fluctuation of photovoltaic power generation and improve the safety and stability of the power system. Summary of the Invention

[0003] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a distributed photovoltaic power prediction method that is accurate, efficient and highly applicable.

[0004] The purpose of the present invention is achieved as follows: A method for predicting the power of distributed photovoltaic power generation includes the following steps: Step 1: Collect and store the power output data and meteorological data of the distributed photovoltaic power station through the monitoring system, and then preprocess the stored data. Step 2: Segment and split the collected power output data and meteorological data according to the unit time, so as to obtain multiple storage packets of power output data and meteorological data according to the unit time. Step 3: Integrate the data inside each storage packet, that is, through the method of weighted average, uniformly integrate the output power data corresponding to similar meteorological data, so as to form multiple corresponding average output power values according to the differences in meteorological data within the unit time. Step 4: Based on the meteorological data values and average output power values within the unit time in Step 3, construct multiple network models sequentially by time period. Step 5: Train each of the network models constructed in Step 4 to obtain the first prediction model. Step 6: In the same way, by changing the value of the unit time, re-segment and integrate the historical meteorological data and historical output power data, and then through model construction and training, obtain the second prediction model. Step 7: Determine the meteorological data prediction information for the prediction period. After splitting the meteorological data of the prediction period according to the unit time in Step 2, input it into the first prediction model to obtain the first power prediction value. Then, split the meteorological data of the prediction period again according to the unit time in Step 6 and input it into the second prediction model to obtain the second power prediction value. Step 8: Compare the first power prediction value and the second power prediction value. If the difference is small, the final power prediction value can be obtained through weighted average. If the difference is large, the unit time value needs to be replaced again, and then the data is split, integrated, and the model is constructed and trained again to calculate the power prediction value again.

[0005] Furthermore, the historical meteorological data of the photovoltaic power station includes: ground temperature, humidity, wind speed, and ground irradiance of the photovoltaic power station every day during the training period.

[0006] Furthermore, the preprocessing operation in Step 1 includes: removing outliers and interpolating and supplementing missing data in the historical power output data and historical meteorological data.

[0007] Furthermore, the integration process in Step 3 is to first perform attribute reduction on the environmental attributes using fuzzy rough sets, and then perform normalization processing on the meteorological data corresponding to the reduced environmental attributes.

[0008] Advantages of the present invention: By preprocessing the power output data and meteorological data collected and stored by the monitoring system of the distributed photovoltaic power station, outliers can be removed and missing data can be interpolated and supplemented. Then, according to the unit time, the collected power output data and meteorological data are segmented and split, so as to obtain multiple storage packets of power output data and meteorological data according to the unit time. After that, the data inside each storage packet is integrated, that is, the output power data corresponding to similar meteorological data is uniformly integrated by the method of weighted average, so as to form multiple corresponding average output power values according to the different meteorological data within the unit time. With this structure, the accuracy of the data can be improved through normalization processing. Then, by constructing a network model and training the network model, a prediction model is obtained. Finally, by changing the value of the unit time, multiple prediction models corresponding to the value of the unit time are constructed. After that, the meteorological data information of the prediction period is split according to the value of the unit time and then brought into the prediction model corresponding to the value of the unit time, so as to obtain multiple power prediction values. Finally, the multiple power prediction values are compared. If the difference is small, the final power prediction value can be obtained through weighted average; if the difference is large, it is necessary to re-change the unit time value, and then perform data splitting, integration, model construction and training again, and then calculate the power prediction value again. The structure adopted by the present invention can increase the accuracy of the power prediction value through the comparison of the final power value and weighted average. Generally, the present invention has the advantages of high precision, high efficiency and strong applicability. Detailed implementation mode

[0009] The present invention will be further described below.

[0010] Embodiment: A distributed photovoltaic power generation prediction method includes the following steps: Step 1: Collect and store the power output data and meteorological data of the distributed photovoltaic power station through the monitoring system. Then, preprocess the stored data. The preprocessing operation includes: removing outliers and interpolating and supplementing missing data in the historical power output data and historical meteorological data. The historical meteorological data of the photovoltaic power station includes: ground temperature, humidity, wind speed and ground irradiance of the photovoltaic power station every day during the training period. Step 2: Segment and split the collected power output data and meteorological data according to the unit time, so as to obtain multiple storage packets of power output data and meteorological data according to the unit time. Step 3: Integrate the data inside each storage packet, that is, uniformly integrate the output power data corresponding to similar meteorological data by the method of weighted average, so as to form multiple corresponding average output power values according to the different meteorological data within the unit time. Step 4: Based on the meteorological data values and average output power values within a unit time in Step 3, construct multiple network models sequentially by time periods; Step 5: Train each of the network models constructed in Step 4 one by one to obtain the first prediction model; Step 6: In the same way, by changing the value of the unit time, re - split and integrate the historical meteorological data and historical output power data, and then through model construction and training, obtain the second prediction model; Step 7: Determine the meteorological data prediction information for the prediction period. After splitting the meteorological data of the prediction period according to the unit time in Step 2, input it into the first prediction model to obtain the first power prediction value; then, after splitting the meteorological data of the prediction period again according to the unit time in Step 6 and inputting it into the second prediction model, obtain the second power prediction value; Step 8: Compare the first power prediction value and the second power prediction value. If the difference is small, the final power prediction value can be obtained through weighted average; if the difference is large, it is necessary to re - change the unit time value, and then re - perform data splitting, integration, model construction and training, and then recalculate the power prediction value.

[0011] When the present invention is in use, first, the monitoring system is used to collect and store the power output data and meteorological data of the distributed photovoltaic power station. After that, the stored data is preprocessed, that is, the outliers in the historical power output data and historical meteorological data are removed and the missing data is interpolated and supplemented. Then, according to the unit time, the collected power output data and meteorological data are segmented and split, so as to obtain multiple storage packages of power output data and meteorological data according to the unit time. After that, the data inside each storage package is integrated, that is, the output power data corresponding to similar meteorological data is uniformly integrated by means of weighted average, so as to form multiple corresponding average output power values according to the different meteorological data within the unit time. With this structure, the accuracy of the data can be improved through normalization processing. Then, through the meteorological data values and average output power values within the unit time, multiple network models are constructed sequentially in different time periods and the network models are trained to obtain the prediction model. Finally, in the same way, by changing the value of the unit time, multiple prediction models corresponding to the value of the unit time are constructed. After that, the meteorological data information of the prediction period is split according to the value of the unit time and then brought into the prediction model corresponding to the value of the unit time to obtain multiple power prediction values. Finally, the multiple power prediction values are compared. If the difference is small, the final power prediction value can be obtained through weighted average; if the difference is large, it is necessary to re-change the unit time value and perform data splitting, integration, model construction and training again, and then calculate the power prediction value again. The structure adopted by the present invention can increase the accuracy of the power prediction value through the comparison of the final power value and weighted average. Generally, the present invention has the advantages of high precision, high efficiency and strong applicability.

[0012] The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that the specific implementation manners of the present invention can still be modified or equivalently replaced. Any modification or equivalent replacement without departing from the spirit and scope of the present invention shall be covered by the scope of the claims of the present invention.

Claims

1. A distributed photovoltaic power generation prediction method, characterized in that It includes the following steps: Step 1: Collect and store the power output data and meteorological data of the distributed photovoltaic power station through the monitoring system, and then preprocess the stored data; Step 2: Segment and split the collected power output data and meteorological data according to the unit time, so as to obtain multiple storage packages of power output data and meteorological data according to the unit time; Step 3: Integrate the data inside each storage package, that is, through the weighted average method, uniformly integrate the output power data corresponding to similar meteorological data, so as to form multiple corresponding average output power values according to the different meteorological data within the unit time; Step 4: Based on the meteorological data values and average output power values within the unit time in Step 3, construct multiple network models sequentially by time period; Step 5: Train each network model constructed in Step 4 to obtain the first prediction model; Step 6: In the same way, by changing the value of the unit time, re - split and integrate the historical meteorological data and historical output power data, and then through model construction and training, obtain the second prediction model; Step 7: Determine the meteorological data prediction information of the prediction period, and after splitting the meteorological data of the prediction period according to the unit time in Step 2, input it into the first prediction model to obtain the first power prediction value; then, split the meteorological data of the prediction period again according to the unit time in Step 6 and input it into the second prediction model to obtain the second power prediction value; Step 8: Compare the first power prediction value and the second power prediction value. If the difference is small, the final power prediction value can be obtained through weighted average; if the difference is large, it is necessary to re - change the unit time value, and then perform data splitting, integration, model construction and training again, and then calculate the power prediction value again.

2. The distributed photovoltaic power generation power prediction method according to claim 1, wherein: The historical meteorological data of the photovoltaic power station includes: ground temperature, humidity, wind speed and ground irradiance of the photovoltaic power station every day during the training period.

3. The distributed photovoltaic power generation power prediction method according to claim 1, characterized in that: The preprocessing operation in Step 1 includes: removing outliers and interpolating and supplementing missing data in the historical power output data and historical meteorological data.

4. The distributed photovoltaic power generation power prediction method according to claim 1, wherein: The integration process in Step 3 is to first use the fuzzy rough set to perform attribute reduction on the environmental attributes, and then perform normalization processing on the meteorological data corresponding to the reduced environmental attributes.