CN-122018324-A - Photovoltaic bracket group form autonomous optimizing and cooperative control method for complex terrain

CN122018324ACN 122018324 ACN122018324 ACN 122018324ACN-122018324-A

Abstract

The invention provides a photovoltaic bracket group form autonomous optimizing and cooperative control method oriented to complex terrain, which realizes fine modeling of a coverage area irradiation field through high-resolution irradiation distribution map acquisition, multi-source data fusion and spatial interpolation, combines gradient sensitivity and action entropy of each bracket node to realize judgment and cooperative processing of high-conduction low-response risk nodes, utilizes a lightweight graph attention model and a consistency voting mechanism to generate locally consistent action correction suggestion, and automatically eliminates node-level action instruction conflict.

Inventors

HUANG SHENGHAI
LIU XIAN
HUANG PANPAN
LI YIBIAO

Assignees

广州市哲明惠科技有限责任公司

Dates

Publication Date: 20260512
Application Date: 20260407

Claims (10)

1. The photovoltaic bracket group form autonomous optimizing and cooperative control method for the complex terrain is characterized by comprising the following steps of: s1, acquiring a real-time irradiation distribution map of a coverage area of a photovoltaic bracket group; S2, constructing a space adjacency graph by taking the physical coordinates of each photovoltaic bracket as graph nodes based on the real-time irradiation distribution graph, calculating irradiation intensity differential vectors in the field of each node 8, and generating a three-dimensional irradiation gradient tensor; s3, carrying out gradient propagation weighted aggregation on the three-dimensional irradiation gradient tensor, and outputting node-level gradient sensitivity indexes; s4, calculating node action entropy based on action instruction types, amplitude intervals and execution result feedback of the last N control periods of each node; S5, judging whether each node simultaneously meets the conditions that the gradient sensitivity index is larger than a gradient sensitivity threshold and the node motion entropy is smaller than a motion entropy threshold, and if so, judging that the node is in a high conduction-low response risk state; s6, triggering a local cache mode and generating a neighborhood negotiation request for the node in the high-conduction low-response risk state; S7, responding to the neighborhood negotiation request, and generating a locally consistent action correction suggestion set through a graph attention model based on gradient sensitivity indexes and action entropy of k neighboring nodes; And S8, performing consistency voting on the action correction suggestion set, outputting a unique correction instruction, covering an original control instruction, and completing localization resolution of node action instruction conflict.
2. The method for autonomous optimizing and cooperatively controlling the morphology of a photovoltaic bracket group facing complex terrain according to claim 1, wherein the step of obtaining the real-time irradiation distribution map of the coverage area of the photovoltaic bracket group specifically comprises the following steps: And respectively calibrating, linearizing, registering, multi-source fusion and spatial interpolation on the original visible light image sequence acquired by the sky imager and the received meteorological satellite downlink real-time cloud image data, and finally outputting a real-time irradiation distribution diagram with the spatial resolution of 10 cm/pixel.
3. The complex terrain-oriented autonomous optimizing and cooperative controlling method for the morphology of the photovoltaic bracket group according to claim 1, wherein the three-dimensional irradiation gradient tensor consists of horizontal irradiation differences in the x and y directions of nodes and curvature approximations in the z direction.
4. The method for autonomous optimizing and cooperatively controlling the morphology of the photovoltaic bracket group facing to the complex terrain according to claim 1, wherein the step S3 specifically comprises: Performing distance reciprocal weight coefficient calculation processing on Euclidean distances of all adjacent edges in the space adjacent graph to generate a weight basis of gradient propagation weighting aggregation; Determining an 8 neighborhood node set of each photovoltaic support node based on the space adjacency graph, and extracting a three-dimensional irradiation gradient tensor of the node and the 8 neighborhood nodes thereof; performing weighted average aggregation operation on the three-dimensional irradiation gradient tensor by applying the distance reciprocal weight coefficient to generate a node-level aggregation gradient vector; Performing Euclidean norm calculation processing on the node level aggregation gradient vector to obtain a node gradient sensitivity preliminary value; and mapping the node gradient sensitivity preliminary value to a unified interval through linear normalization, and outputting a node level gradient sensitivity index.
5. The method for autonomous optimizing and cooperatively controlling the morphology of a photovoltaic bracket group oriented to complex terrains according to claim 4, wherein the node-level gradient sensitivity index characterizes the conduction position attribute of the node in the irradiation gradient field.
6. The method for autonomous optimizing and cooperatively controlling the morphology of the photovoltaic bracket group facing the complex terrain according to claim 1, wherein the step S4 specifically comprises: performing sliding window time sequence analysis processing on the action instruction type set, the amplitude interval division standard and the execution result feedback type of the photovoltaic support node, and constructing a node action condition probability distribution model; Based on the node action conditional probability distribution model, executing data acquisition operation of a local storage unit of the edge controller on the designated photovoltaic bracket node identifier, extracting action instruction types, amplitude intervals and execution result feedback data of the latest N control periods, and forming a node historical action record sequence; Applying a sliding window statistical mechanism to the node historical action record sequence to process, and calculating node action condition probability distribution of each type of action instruction type in a given state; Executing negative logarithm summation operation processing on the node action conditional probability distribution, and calculating a node action entropy value; And carrying out numerical range verification and standardization processing on the node action entropy value, and outputting standardized node action entropy.
7. The complex terrain-oriented photovoltaic bracket group morphology autonomous optimizing and cooperative control method according to claim 6, wherein the standardized node motion entropy is used as a quantitative measure of node response certainty.
8. The method for autonomous optimizing and cooperatively controlling the morphology of the photovoltaic bracket group facing the complex terrain according to claim 1, wherein the step S5 specifically comprises: Executing real-time data acquisition processing on the node-level gradient sensitivity index stored in the edge controller to acquire the current value of the node-level gradient sensitivity index; based on the current value of the node-level gradient sensitivity index, performing numerical comparison operation processing with a gradient sensitivity threshold value to generate a Boolean judgment result that the gradient sensitivity index is larger than the gradient sensitivity threshold value; executing real-time data acquisition processing on the standardized node action entropy stored in the edge controller to acquire the current numerical value of the standardized node action entropy; Based on the current value of the standardized node action entropy, executing numerical comparison operation processing with an action entropy threshold value, and generating a Boolean judgment result that the node action entropy is smaller than the action entropy threshold value; And performing logic AND operation processing on the Boolean judgment result with the gradient sensitivity index larger than the gradient sensitivity threshold and the Boolean judgment result with the node motion entropy smaller than the motion entropy threshold, and outputting a high-conduction-low-response risk state judgment signal.
9. The complex terrain oriented photovoltaic bracket group morphology autonomous optimizing and cooperative control method according to claim 8, wherein the gradient sensitivity threshold is based on pre-set calibration of photovoltaic power station early-stage terrain survey data, historical irradiation field distribution statistical results and bracket array arrangement parameters, and the action entropy threshold is based on pre-set calibration of historical action execution data, control period response characteristics and historical instruction conflict records of a photovoltaic bracket driving mechanism.
10. The complex terrain-oriented autonomous optimizing and cooperative control method for the photovoltaic support group morphology according to claim 1 is characterized in that the graph attention model is constructed by performing graph attention model framework definition processing on three-dimensional irradiation gradient tensors and node action entropy of photovoltaic support nodes, constructing a model input layer and an attention calculation layer based on action sensitivity characteristics and node response certainty measures on gradient propagation paths, and generating graph attention model examples.

Description

Photovoltaic bracket group form autonomous optimizing and cooperative control method for complex terrain Technical Field The invention relates to the technical field of intelligent cooperative control of photovoltaic power stations, in particular to an autonomous optimizing and cooperative control method for a photovoltaic bracket group form oriented to complex terrain. Background Currently, a cluster cooperative control technology of a photovoltaic power station in a complex terrain and non-uniform cloud shadow environment is continuously and widely focused. The mainstream photovoltaic support control system in the industry generally adopts centralized instruction distribution and an action scheduling mechanism based on a priori strategy or priority table to cope with the collaborative light following of the distributed support group, but the traditional technical paths have obvious limitations under the dynamic and asymmetric irradiation conditions. Due to factors such as space shielding, rapid movement of local cloud shadow, transient irradiation intensity and the like, the action response demands of individual support nodes are frequently changed, and sensitivity of each node to irradiation gradient presents strong space heterogeneity, so that the consistency and robustness of overall response are difficult to ensure in the prior art; Currently, part of the industry attempts to introduce methods such as multi-agent control, global scheduling based Yu Yunying prediction, reinforcement learning distributed strategies, and the like, but the mainstream technology generally depends on complex cloud shadow dynamic modeling or wide area environment sensing, a central server with high calculation power is needed to participate in decision making, or a static priority distribution and centralized anomaly detection mechanism is adopted. The scheme has the following typical application scenes and application ranges in actual deployment, is suitable for small and medium-sized photovoltaic arrays with relatively uniform topography and relatively stable illumination distribution, and can not meet the engineering requirements of high availability and high robustness on the problems of high-frequency action conflict, local priority mismatch, excessively high global re-optimization time delay and the like of an asymmetric irradiation scene with extremely complex topography and rapid change. Particularly, in a large-scale distributed bracket cooperative system, due to the lack of dynamic quantization constraint on actual gradient sensitivity and response certainty of nodes, the fuzzy or erroneous judgment of the bracket nodes on the priority judgment can cause frequent collision of the action instructions, and local cooperative failure and overall power generation efficiency reduction can be caused when the action instructions are serious. In addition, the existing methods such as abnormal cluster clustering, static weight mapping, rolling optimization and the like can not realize real-time response and conflict localization resolution of physical coupling dynamic characteristics among nodes, are easy to be limited by measurement and control bandwidth, data delay and calculation power resources, and are difficult to support intelligent operation and maintenance requirements of future heterogeneous complex terrain large-scale photovoltaic power stations. Disclosure of Invention The invention aims to solve the technical problems and provides a photovoltaic bracket group form autonomous optimizing and cooperative control method for complex terrain. The technical scheme of the invention is realized in such a way that the photovoltaic bracket group form autonomous optimizing and cooperative control method facing to complex terrain comprises the following steps: S1, acquiring a real-time irradiation distribution diagram of a coverage area of a photovoltaic bracket group, wherein the spatial resolution of the irradiation distribution diagram reaches 10 cm/pixel and the irradiation distribution diagram is used as an input data source for asymmetric irradiation field modeling; S2, based on the real-time irradiation distribution map, constructing a space adjacency graph by taking the physical coordinates of each photovoltaic bracket as graph nodes, calculating 8 neighborhood irradiation intensity differential vectors of each node, and generating a three-dimensional irradiation gradient tensor as basic data of gradient field topological coding; s3, carrying out gradient propagation weighted aggregation on the three-dimensional irradiation gradient tensor, and outputting a node-level gradient sensitivity index, wherein the index represents the conduction position attribute of the node in an irradiation gradient field; s4, calculating node action entropy based on action instruction types, amplitude intervals and execution result feedback of the latest N control periods of each node, wherein the action entropy is used as a quantized m