CN-122026801-A - Photovoltaic power minute-scale rolling prediction method and system based on internal perception

CN122026801ACN 122026801 ACN122026801 ACN 122026801ACN-122026801-A

Abstract

The invention relates to the technical field of new energy power generation operation control and intelligent prediction, in particular to a photovoltaic power minute-scale rolling prediction method and system based on internal perception, wherein the method comprises the steps of firstly constructing a second-level multi-source real-time perception network containing a multi-spectrum light intensity sensor and a micro barometer in a photovoltaic power station, extracting characteristics through an edge calculation server and triggering internal primary early warning; then establishing an RNN mixed prediction model (comprising delay perception modulation and internal early warning enhancement mechanism) which fuses the early warning characteristics and external hysteresis meteorological data; the method solves the problem of photovoltaic power prediction hysteresis distortion in short-time strong convection weather, improves the perception foresight and prediction reliability, strives for key early warning advance for a power grid dispatching and energy storage system, and enhances the operation safety and the power grid friendliness of a power station.

Inventors

WU MING
HUANG JIE
YANG FAN
HONG XING
RUAN JIANFEI
ZHAI XIAOTING
WANG LIANG
JIANG ZHAOXIA
JIN XINXIN
ZHAO ZHENCHAO

Assignees

南通沃太新能源有限公司

Dates

Publication Date: 20260512
Application Date: 20260119

Claims (10)

1. The photovoltaic power minute-scale rolling prediction method based on internal perception is characterized by comprising the following steps of: (1) A second-level multisource real-time sensing network is built in a photovoltaic array area of a photovoltaic power station, and solar total radiation intensity, assembly back plate temperature, environment temperature and humidity, multispectral irradiance and micro-air pressure data are collected and converged to an edge calculation server; (2) The edge calculation server performs feature extraction on the acquired data to generate a radiation related index, a spectrum distortion index and an air pressure disturbance index, and an internal primary early warning signal close to strong convection weather is triggered based on the indexes; (3) Constructing a hybrid prediction model fusing internal real-time perception features and external hysteresis meteorological data, wherein the model comprises a delay perception modulation mechanism and an internal early warning enhancement mechanism; (4) When the internal primary early warning signal is continuously triggered, a minute-level rolling prediction process is started, the internal early warning state and external meteorological data delay information are updated in real time, and the mixed prediction model is input; (5) And outputting a photovoltaic power prediction sequence in a future preset time window through the hybrid prediction model, generating a regulation and control instruction by combining the internal early warning grade, and sending the regulation and control instruction to the power station energy management system.
2. The method for predicting the power level of the photovoltaic power by rolling in minute based on internal perception according to claim 1, wherein in the step (1), the multisource real-time perception network adopts a meshing and functional layering deployment strategy, basic monitoring nodes are distributed at equal intervals along the component row-column direction, and a multispectral light intensity sensor and a micro barometer are distributed on the windward side of the photovoltaic array and in a region which is easily affected by cloud shadows.
3. The internal perception based photovoltaic power minute-scale rolling prediction method according to claim 1, wherein in step (2), the radiation-related index is used to quantify the synchronous trend of the total station radiation, the spectral distortion index is used to characterize the change of the solar radiation energy structure, and the barometric disturbance index is used to describe the micro-scale pressure anomaly.
4. The method for predicting the power level of photovoltaic power by rolling in minute based on internal perception according to claim 1, wherein in the step (3), the hybrid prediction model uses a cyclic neural network as a core frame, and inputs radar echo intensity and satellite cloud image data including an internal primary early warning signal intensity sequence, a solar total radiation intensity history sequence and a delay mark.
5. The photovoltaic power minute-level rolling prediction method based on internal perception according to claim 1, wherein in step (4), the minute-level rolling prediction flow processes the second-level internal feature by weighted time aggregation, so that the internal feature is adapted to the minute-level rolling rhythm.
6. The internal awareness based photovoltaic power minute-scale rolling prediction method according to claim 1, wherein in step (5), the hybrid prediction model dynamically adjusts the degree of dependence of internal awareness features on external meteorological data by a trust weight scheduling function, the trust weight scheduling function being related to internal pre-warning intensity and external data delay degree.
7. The method of claim 1, wherein in step (5), when the prediction result indicates that a power cliff drop will occur within a preset time in the future, the preset time is 10 minutes, an emergency control command of the highest priority is generated.
8. An internally-aware-based photovoltaic power minute-level roll prediction system for implementing the internally-aware-based photovoltaic power minute-level roll prediction method of any of claims 1-7, comprising: the multisource real-time sensing network is deployed in a photovoltaic array area of the photovoltaic power station and is used for acquiring second-level multisource sensing data; the edge computing server is in communication connection with the multi-source real-time sensing network and is used for feature extraction, internal primary early warning signal triggering and minute-level rolling prediction flow starting; The mixed prediction model is deployed on the edge calculation server and is used for fusing the internal real-time perception characteristics and external hysteresis meteorological data and outputting a photovoltaic power prediction sequence; And the power station energy management system is in communication connection with the edge calculation server and is used for receiving the photovoltaic power prediction sequence and the regulation and control instruction and executing corresponding operation.
9. The internal perception-based photovoltaic power minute-scale rolling prediction system according to claim 8, wherein the multi-source real-time perception network comprises a basic monitoring node, a multi-spectrum light intensity sensor and a micro barometer, wherein the basic monitoring node is used for collecting solar total radiation intensity, component back plate temperature and environment temperature and humidity data.
10. The internal perception based photovoltaic power minute-scale rolling prediction system of claim 8, wherein the edge calculation server is further configured to time-stamp correct data collected by the multi-source real-time perception network and evaluate a power dip risk of the photovoltaic power prediction sequence.

Description

Photovoltaic power minute-scale rolling prediction method and system based on internal perception Technical Field The invention relates to the technical field of new energy power generation operation control and intelligent prediction, in particular to a photovoltaic power minute-scale rolling prediction method and system based on internal perception. Background Along with the rapid increase of the capacity of the photovoltaic installation, the fluctuation and uncertainty of the photovoltaic output bring higher requirements to the safe operation and the dispatching precision of the power grid, wherein the rapid movement of cloud clusters, the sudden change of radiation and the drop of power cliff caused by strong convection weather become important factors for restricting the grid-connected operation of the photovoltaic in high proportion, so that how to accurately sense the sudden change of weather and predict the power change in advance in a minute scale or even a shorter time scale is a key technical problem to be solved urgently in the current photovoltaic prediction field. Through retrieval, the patent with the application number of WO2014190651A1 provides a photovoltaic power prediction method based on a ground cloud picture, cloud picture movement is predicted through a cloud picture sequence, further ground irradiance change is predicted, photovoltaic power prediction is realized by combining a photoelectric conversion model, digital image processing and multi-factor modeling means are comprehensively utilized, the refinement level of photovoltaic power prediction is improved to a certain extent, effective supplement is formed by the traditional statistical prediction method, the method has good universality and robustness, but the method mainly depends on the ground cloud picture and relevant meteorological data as core input, and the prediction effect is limited by cloud picture acquisition frequency, image processing delay and cloud picture recognition precision to a great extent. The traditional photovoltaic power prediction method is generally focused on analysis of external weather information or cloud image data, the utilization of real-time physical quantity inside a photovoltaic power station is insufficient, the inherent time lag of the external data is not subjected to explicit modeling, local sudden change characteristics are difficult to reflect in time under a strong convection rapid development scene, minute-level rolling prediction results are easy to lag and even distort, strong convection weather impact cannot be responded in advance, local radiation sudden change caused by rapid movement cloud clusters and photovoltaic power 'cliff' falling problems are difficult to accurately and timely sense and predict, and further effective early warning advance cannot be provided for power grid dispatching and power station self regulation, so that the development of the photovoltaic power minute-level rolling prediction method and system based on internal sensing is urgently needed to overcome the defects in the current practical application aiming at the current situation. Disclosure of Invention The invention aims to provide a photovoltaic power minute-scale rolling prediction method and system based on internal perception, which are used for solving the problems in the background technology. In order to achieve the above purpose, the present invention provides the following technical solutions: an internal perception-based photovoltaic power minute-scale rolling prediction method comprises the following steps: (1) A second-level multisource real-time sensing network is built in a photovoltaic array area of a photovoltaic power station, and solar total radiation intensity, assembly back plate temperature, environment temperature and humidity, multispectral irradiance and micro-air pressure data are collected and converged to an edge calculation server; (2) The edge calculation server performs feature extraction on the acquired data to generate a radiation related index, a spectrum distortion index and an air pressure disturbance index, and an internal primary early warning signal close to strong convection weather is triggered based on the indexes; (3) Constructing a hybrid prediction model fusing internal real-time perception features and external hysteresis meteorological data, wherein the model comprises a delay perception modulation mechanism and an internal early warning enhancement mechanism; (4) When the internal primary early warning signal is continuously triggered, a minute-level rolling prediction process is started, the internal early warning state and external meteorological data delay information are updated in real time, and the mixed prediction model is input; (5) And outputting a photovoltaic power prediction sequence in a future preset time window through the hybrid prediction model, generating a regulation and control instruction by combining the internal early warning g