Modelling Residential supply tasks based on digital orthophotography using machine learning

2025-05-06 0 0 397.56KB 5 页 10玖币

侵权投诉

1/5

MODELLING RESIDENTIAL SUPPLY TASKS BASED ON DIGITAL

ORTHOPHOTOGRAPHY USING MACHINE LEARNING

Klemens SCHUMANN1,2, Luis Böttcher2, Philipp Hälsig1, Daniel Zelenak2, Andreas Ulbig1,2

1Fraunhofer Center Digital Energy – Germany

2IAEW at RWTH Aachen University – Germany

klemens.schumann@fit.fraunhofer.de

ABSTRACT

In order to achieve the climate targets, electrification of

individual mobility is essential. However, grid integration

of electrical vehicles poses challenges for the electrical

distribution network due to high charging power and

simultaneity. To investigate these challenges in research

studies, the network-referenced supply task needs to be

modeled. Previous research work utilizes data that is not

always complete or sufficiently granular in space. This is

why this paper presents a methodology which allows a

holistic determination of residential supply tasks based on

orthophotos. To do this, buildings are first identified from

orthophotos, then residential building types are classified,

and finally the electricity demand of each building is

determined. In an exemplary case study, we validate the

presented methodology and compare the results with

another supply task methodology. The results show that

the electricity demand deviates from the results of a

reference method by an average 9%. Deviations result

mainly from the parameterization of the selected

residential building types. Thus, the presented

methodology is able to model supply tasks similarly as

other methods but more granular.

INTRODUCTION

One important aspect of achieving the climate targets is the

electrification of the mobility sector. Therefore, Germany

has set a target of ten million electric vehicles and one

million charging stations by 2030 [1].

Due to the high charging power and the simultaneity of

electric vehicle charging, the growth poses challenges for

the distribution network infrastructure, which needs to be

investigated especially on low and medium voltage levels.

To study the impact, network-referenced supply tasks are

needed. These consist of geo-referenced electricity

demands, load and generation time series as well as

charging profiles for the electric vehicles, among others.

While the modelling of synthetic time series (load and

generation) and charging profiles is continuously

discussed in research, the geo-referencing of electricity

demands is still a challenge due to limited data availability.

Previous studies such as [2–6] which tackle this topic use

socioeconomic data, geospatial data, digital

orthophotographs (DOP), or measurement data.

Socioeconomic data contains information on the

population and building structure of a region. Among other

things, they indicate number of buildings by building size

and construction year. For data privacy reasons, the data is

aggregated to grids (e.g. 100𝑚 𝑥 100𝑚) [7]. Supply task

modelling based on socioeconomic data uses population

and building statistics. Due to the low granularity,

conclusions cannot be drawn about the actual location of

individual buildings in the distribution network [2].

Geospatial data represents georeferenced shapes of

buildings provided with information about the building.

Among others, geospatial data is collected on the map

service Open Street Maps (OSM). Besides the outlines of

buildings, it sometimes is also indicated whether buildings

are residential or commercial buildings [8]. Besides OSM,

geospatial data can be accessed from cadastral offices (e.g.

[9]). Depending on the data set, the data includes house

perimeters up to information about building height and

roof shape. Methodologies that use geospatial data as data

basis make use of available building information [3, 4].

However, data in OSM is uploaded by users non-

commercially so that completeness and accuracy cannot be

guaranteed. Cadastral data often needs to be licensed.

DOP are georeferenced aerial photographs. Although

information cannot be derived directly from these images,

buildings or other objects can be identified with the aid of

suitable methods [10]. In Germany, DOP are maintained

by cadastral offices (e.g. [11]). This ensures high

resolution and availability but needs to be licensed in some

cases. Other providers offer georeferenced satellite

imagery that can be used similarly to DOP (e.g. [12]). [5]

presents a methodology to model supply tasks using

various data sources. Besides using geospatial data and

socio-economic data, the authors also use DOP. However,

they apply the DOP in order to identify existing PV-

rooftop systems.

Studies that use measurement data-based supply tasks are

often conducted in collaboration with distribution system

operators (DSOs) [6]. Here, the DSO’s data is used.

However, these studies are always limited to the network

area and data of the DSO.

Since the methods mentioned are not fully suitable to

model georeferenced supply tasks, we present a method to

determine georeferenced supply tasks based on DOP using

image recognition and different machine learning

techniques. As training data, DOP and geospatial data is

used. The trained models can be applied just using DOP.

Adding information from socioeconomic or statistical

data, the electricity demand can be determined. Rather

than determining the real supply task, the aim is to

determine realistic supply tasks. The result of the method

can therefore be used for research studies instead of studies

for network operators.

In an exemplary case study, we analyze the quality of the

trained models. For this purpose, we first train the models

on a training region that has a very high quality of open

source data. Then, we apply the trained models to a study

region. Finally, we discuss different findings, and compare

and evaluate the identified energy demands with results of

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

1/5MODELLINGRESIDENTIALSUPPLYTASKSBASEDONDIGITALORTHOPHOTOGRAPHYUSINGMACHINELEARNINGKlemensSCHUMANN1,2,LuisBöttcher2,PhilippHälsig1,DanielZelenak2,AndreasUlbig1,21FraunhoferCenterDigitalEnergy–Germany2IAEWatRWTHAachenUniversity–Germanyklemens.schumann@fit.fraunhofer.deABSTRACTInordertoachievetheclim...

展开>> 收起<<

Modelling Residential supply tasks based on digital orthophotography using machine learning.pdf

共5页,预览1页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Modelling Residential supply tasks based on digital orthophotography using machine learning

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: