Solar Radiation Deep Learning Prediction Model on a Photovoltaic Solar Energy Analysis Platform

View Statistics Email Alert RSS Feed

Information

Details

Project title

Code/計畫編號

MOST107-2111-M019-001-CC3

Translated Name/計畫中文名

太陽輻射深度學習預測模式及光伏太陽能分析平台之研發

Project Coordinator/計畫主持人

Chih-Chiang Wei

Funding Organization/主管機關

National Science and Technology Council

Department/Unit

Department of Marine Environmental Informatics

Website

https://www.grb.gov.tw/search/planDetail?id=12556277

Year

2018

Start date/計畫起

01-06-2018

Expected Completion/計畫迄

31-05-2019

Bugetid/研究經費

475千元

ResearchField/研究領域

大氣科學

在臺灣，太陽能產業是一項未來的熱門產業，建築物可搭配各式太陽能發電系統提高能源的使用效率。太陽能經電能轉換技術後可將太陽能電池直接轉換為電能，直接取用的太陽能，因不造成污染且無噪音，是非常乾淨的能源，因此太陽能是具潛力的新興能源之一。一般，當太陽輻射進入大氣層後，主要受到雲層覆蓋及厚度的影響，造成太陽輻射被雲層吸收、反射及散射，致使太陽輻射能來源具有不確定性，當太陽能板上所收集到的能源密度偏低時，則須裝設大面積的太陽能板並增加投資成本；另外，隨著季節、日夜長短的變化、地球公轉與自轉，甚至隨機不易預測的天氣變化，皆會影響到太陽能源之接收。所以必須先掌握太陽輻射能量，才能設計最佳的太陽能板系統容量、裝設位置和傾斜角度。本計畫將採用人工智慧最新型深度學習法研發地表太陽輻射預測模式，再利用大數據平行分散式運算技術發展光伏太陽能分析平台，以達到降低因大氣因素所造成之不確定性，使太陽能板發電功效達到最佳化的目的。大數據運算平台採Apache Hadoop Spark最新雲端技術，以有效率的解決：1)同時接收來自各地遠端伺服器的資料來源；2)在主機端以太陽輻射深度學習預測模式即時預測；3)設計最佳的太陽能板系統容量、裝設位置和傾斜角度。預測模式建立採用三種數據，地面氣象、大氣遙測和太陽位置參數，其中衛星遙測來源為NOAA MODIS (選用大氣層中雲相關參數)。本計畫所開發的整合分析平台包含太陽輻射預測模式、大數據運算以及光伏太陽能分析模組。整合系統將有利於營建業或太陽能板業者於屋頂裝置固定/追日式太陽能板之最佳斜度；另外，亦提供地表太陽輻射的預測值以估算實際進入固定/追日式太陽能光電系統的輻射量，藉以避免光電系統過度頻繁的充電/放電，提供高效能的供電可靠度。本期將選定台南市或適合發展太陽能縣市為場址。計畫主要工作有：1)開發新一代人工智慧型深度學習法建立太陽輻射預測模式，深度學習是人工智慧成長最快的領域，讓電腦更接近人類的思考。本計畫採用人工智慧的深度學習模式包含深度神經網路、遞迴神經網路等；2)開發大數據Hadoop Spark平行且分散式運算平台以解決同時接收來自各地遠端伺服器的資料來源，並且在主機端利用所發展的地表太陽輻射預測模式進行即時預測工作；3)建立光伏太陽能分析平台以利光電能量轉換估算，並評估供電可靠度時之運用。Southern Taiwan has an excellent solar energy resource that remains largely unused. A methodology is proposed for developing a surface solar radiation real-time prediction model for photovoltaic (PV) generation system installed in the rooftop of buildings. The surface solar radiation prediction model deals with meteorological uncertainty of photovoltaic generation. The proposed surface solar radiation prediction model for simulating the hourly solar irradiation considers solar orbit/angle, atmospheric cloud status, and ground meteorological information using hourly solar irradiation as inputs. This platform will employ the AI-based deep neural network (DNN) and recurrent neural network (RNN) for predicting hourly surface solar radiation, and formulate photovoltaic generation system for tracking the angle of the sun, and adjust the angle for tracking leading to direct sunlight could be acquired to increase power output.We propose a Hadoop Spark Big Data platform applied to surface solar radiation real-time prediction in photovoltaic generation system installed in the buildings. There are various tools which can be used in Big Data management from data acquisition to data analysis. Hadoop brings the ability to cheaply process large amounts of data, regardless of its structure. Hadoop is made up of two core projects: Hadoop Distributed File System (HDFS) and MapReduce. Hadoop splits files into large blocks and distributes them across nodes in a cluster. To process data, Hadoop transfers packaged code for nodes to process in parallel based on the data that needs to be processed. In addition, Apache Spark provides programmers with an application programming interface centered on a data structure called the resilient distributed dataset, a read only multiset of data items distributed over a cluster of machines, that is maintained in a fault-tolerant way. It was developed in response to limitations in the MapReduce cluster computing paradigm, which forces a particular linear dataflow structure on distributed programs: MapReduce programs read input data from disk, map a function across the data, reduce the results of the map, and store reduction results on disk. In practice, solar collectors (flat plate thermal or photovoltaic collectors) are tilted, thus it is necessary to know the solar radiation incident on such tilted planes. Moreover, the particularity of solar conversion systems, as for all energy systems using a phenomenological input, comes from the uncontrollable character of the energy input as a result of non-foreseeable meteorological variations. Then, even for a perfectly known system, from a mathematical point of view, the efficiency or the productivity of such a system is dependent on the temporal fluctuations of the energy input and output. Thus, the solar irradiation must be known with a time step at most equal to an hour. To ensure the accuracy of the Hadoop-based machine learning-based models, the traditional regressions will be used as the benchmarks. A forecasting horizon ranging from 1 to 12 h will be used for real-time photovoltaic generation system operations for solar energy industry.

Keyword(s)

太陽輻射
光伏太陽能
大氣遙測
人工智慧
巨量資料
Solar Irradiation
Photovoltaic Solar Energy
Remote Sensing
Artificial Intelligence
Big Data

DSpace CRIS

Solar Radiation Deep Learning Prediction Model on a Photovoltaic Solar Energy Analysis Platform

Details

Description