葉艷珠楊 博Panitarn CHONGFUANGPRINYA

(日立美國(guó)有限公司,圣克拉拉95054,加州,美國(guó))

0 Introduction

In recent years,energy regulations,environmental policies,and technology advances have driven a significant rise in the penetration level of distributed energy resources (DERs)in distribution system[1].The distribution network is gradually evolving to a distributed network of various DERs,including solar photovoltaic panels(PV),energy storage system,EVs,etc.These transitions have brought various challenges to utilities,e.g.reverse power flow,complicated protection schemes,over-voltage profiles,etc.,which not only require regulation changes but also technology innovations.

There are emerging needs for a novel DER management platform to provide better situational awareness of these assets and effective ways to manage these DERs,which is essential to improve grid operation reliability,resiliency,and even market efficiency.When dealing with proliferation of DERs,most of which are considered as behindthe-meter assets,the direct observability and controllability becomes a big challenge.In other words,the efficient data acquisition system and communication infrastructure are critical for an effective DER monitoring and management system.

Currently these emerging DER management systems can be categorized as proprietary and nonproprietary (open)system.In most cases,the DER management systems are dominated by major vendors[2-3]e.g.,and operates as proprietary systems,which integrate hardware equipment and software from specific vendors,and had limited interoperability with open system and standards.Also,most of those conventional DER management systems are often developed in a vertically integrated way,that frontend graphic user interface and various analytical applications are deeply coupled with proprietary data infrastructure,which lacks the flexibility for system scale-up and functional expansions.A standardized,open architecture to aid the integration of DER is still not fully ready yet.

As an integration platform of different emerging technologies,the“internet of things”(Io T)enables many application solutions in various vertical domains,e.g.utilities,smart cities,healthcare,industries,etc.[4-5].The rapid adoption of networked intelligence devices(e.g.smart meters,DERs)greatly supports the Io T applications in energy industry.There exists some works which exploits the potential applications of Io T platform for energy management system[6-8],however most of them are either not utilizing the latest technologies,or have limited openness.In this paper,an innovative Io T data analytical framework is proposed featuring an open end-to-end architecture,integrated data management,interoperable interfaces with thirdparty systems and modular application design.This open architecture features with a hierarchical layered design,which consists of edge layer,Io T data management layer,and modular application layer.On the edge level,the edge gateway consists of physical devices and software for interfacing with field devices.It is critical that edge systems are hardware agnostic and interoperable with different communication standard and data protocols.Io T data management layer is optimized for large data throughputs and massive data storage due to the potential high volume of DER integrations.Coupled through common data interfaces,a variety of DER analytical applications,including third party apps,can be deployed and scaled-up under this framework.Different types of open-source software tools are utilized in the design.

The paper is organized as follows:in sectionⅠthe detailed architecture of this proposed open Io T end-to-end data analytical framework for DER management is introduced.In the sectionⅡ,a cloud-based PV management system is developed based on this proposed Io T analytical framework for a commercial PV fleet owner,which provides monitoring and management of PV fleets across multiple sites.Finally,section Ⅲ concludes the paper and future works are proposed.

1 Architecture of The Open IoT Data analytica l framework

1.1 System Architecture

Figure 1 shows the layered architecture of the proposed open Io T data analytical framework.The proposed framework mainly consists of three layers:edge layer,integrated data management layer,modular application layer.

1.1.1 Edge layer

圖1 開放式物聯(lián)網(wǎng)數(shù)據(jù)分析框架的分層架構(gòu)Fig.1 Layered architecture of the open IoT data analytical framework

The edge layer includes physical hardware devices and software components.We choose to use the open Io T hardware,e.g.Raspberry P i[9].or Intel Mini Box.The edge gateway is physically deployed in the field,which will directly interact with edge devices,e.g.sensors,actuators,data logger,meters,etc.It acts as an intermediate layer between the numerous edge devices and integrated data management layer.The placement of edge gateways alleviates the direct high-volume data throughput between a large amount of DERs and data servers.

Figure 2 illustrates the functional blocks of the software system design in edge gateway.The software system is designed to enable the interoperability with different communication standard and data protocols.The major function of the edge gateway is data collection,data concentration and data transmission through backhaul network to integrated data management layer.Considering that the data from edge device or other third-party system have various data format,the data conversion gateway usually needs to accommodate different field device set-ups.One implementation example is illustrated in Section II when the edge gateway is installed on PV fleet sites.The edge gateway can also serve as edge computing node to provide onsite data pre-processing,edge analytical functions,etc.

圖2 邊際網(wǎng)關(guān)的功能塊Fig.2 Functional blocks on edge gateway

1.1.2 Data management layer

On the data management layer,it provides data storage,data consolidation,data processing functions.We choose to use open source software tools.One exemplary implementation is to use an ELK stack for data ingestor,database and data visualization.The ELK stack consists of Elasticsearch,Logstash and Kibana[10].Elasticsearch is a powerful RESTful,JSON-based search and analytics engine.It is easy to use,scalable and flexible.It supports extensive REST API which can be fully utilized by the analytical applications for data retrieval.The Kibana is a data visualization plugin for Elasticsearch.It provides visualization capabilities on top of the content indexed on an Elasticsearch cluster.

As shown in figure 3,an API gateway is implemented to provide access to back-end data sources and some application functionalities,e.g.authentication,data logging,user management,device control,etc.The API gateway sits on the communication path between different application modules and back-end resources,or commonly used features.API gateways can aggregate,manage,distribute difference service requests to and from these resources.One implementation example is to use Django REST framework,which is a powerful and flexible toolkit for building Web APIs.On top of Django,different functionalities can be developed,e.g.data requests from a front-end GUI can be provisioned through RESTful call through Django REST framework.

圖3 API服務(wù)說(shuō)明Fig.3 Illustration of API services

1.1.3 Modular application layer

Through the common data interfaces with data management layer or third-party APIs,a variety of DER analytical services are developed as application modules and deployed on this open data platform.These modular applications are referred to as microapps.These modular applications can be organized around specific business goals,user groups,DER types,or various analytical functions,etc.The modular application design approach enables system scalability and reusability,as well as efficiency.These individual micro-apps communicate with each other over standard protocols with well-defined accessible interfaces,typically over REST protocols.A large application can be easily built up as a suite of small,individual micro-apps.

Figure 4 illustrates the graphical web portal of Hitachi distributed energy resource management suites,which includes a variety of analytical modular applications targeting at DER monitoring,management,and operation,etc.

圖4 日立分布式能源管理解決方案Fig.4 Illustration of graphical web portal of Hitachi DER management solution suite

1.2 Data communication infrastructure

The detailed data communication infrastructure in the proposed IoT analytical framework is illustrated in figure 5,the data messages are distributed through message brokers.It receives data from edge devices,store them,package and distribute them out as a publisher.The messages are delivered using publish/subscribe mechanism instead of point-to-point data transmission.Due to the publish-subscribe mechanism,the entire data communication system can be easily expanded with new connections from new components to the message bus,which either publish or receive messages as clients.

圖5 數(shù)據(jù)通信基礎(chǔ)架構(gòu)圖Fig.5 Diagram of data communication infrastructure

One of our implementations is to use Rabbit MQ with MQTT as message brokers,considering the MQTT is a very popular protocol for Io T device and then library software is mature[11].MQTT is also a lightweight,open,simple and easy-toimplement protocol that is suitable for constrained network environments and resource-constrained IoT device.Rabbit MQ is a widely deployed open source message broker.

The data communication between analytical application modules and the data management layers is realized through standardized APIs,e.g.RESTful Web Services interface with HTTP protocol.

1.3 System deployment

For the system deployment,the loosely-coupled data management layer and modular application layer can be flexibly deployed either on customer premises,private server or public cloud.Furthermore,to solve the operating environment compatibility issue and reduce deployment complexity,the functional components on different layers can be built-up with container technology[12].These functional components are well separated,dockerized and run within docker containers.These dockers can be deployed easily,run consistently and reliably regardless of the computing environment,no matter on edge gateway,cloud,or private server.

Along with modular application design approach,we apply the continuous integration and continuous delivery (CI/CD)pipeline to develop,deploy and scale-up respective applications independently,speed up the production cycle[13].The open-source tools can be fully utilized in this process,e.g.Jenkins,CircleCI,etc.As shown in figure 6,the combination of CI/CD and container technology is an innovative and effective way to develop,deploy and manage various DER management applications at scale,and enhance the cloud capabilities.

2 Use Case

In this section,a cloud-based PV management system is developed for a commercial PV fleet owner based on the proposed open Io T analytical framework.This Proof-of-Concept(PoC)PV management system provides one uniform monitoring/management interface for multi-site commercial PV fleets.On different fleet sites,the PV inverters are provided by different vendors,so are the data logging system,network environment,etc.The open Io T data analytical framework well accommodates the discrepancies across different sites.Currently this online PV management system provides the following major functions:

?Uniform real-time monitoring interface for PV fleets across multi-sites at various levels,from entire fleet level down to each inverter modular level.

?Performance analytics/assessment at various levels

?System alerting and reporting

?Short-term temporal PV power generation forecasting

2.1 System architecture

The architecture of this cloud-based PV management system is shown in figure 7.On the edge layer,customized edge gateways are deployed on customer sites,the data management layer and the application layer are deployed on Amazon cloud.User access control are managed through user management module.The edge gateway is connected to cloud server through LAN network.

The edge gateway is developed based on RaspberryPi.An in-house edge processing software is developed and installed on RaspberryPi.The edge gateway is programmed to have automatic reboot or restart function to prevent unexpected power shown-down or other abnormal events.This edge gateway operates as a small server system and can be accessed through web protocol remotely.Software updates or code revision will be remotely pushed through Internet.

圖6 用于系統(tǒng)開發(fā)的CI/CD說(shuō)明Fig.6 Illustration of CI/CD for system development

圖7 基于云的光伏管理系統(tǒng)結(jié)構(gòu)Fig.7 System architecture of cloud-based PV management system

On the field,the edge gateway directly interacts with various edge devices,e.g.data loggers,energy meters,etc.As shown in figure 7,the data loggers and energy meters are physically located in multiple sites.The data logger collects and concentrates data from each PV inverters,including voltage,current,power,frequency,temperatures,etc.On different PV installation sites,the PV inverters are provided by different vendors,data loggers or meter settings also differ depending on vendors.

These edge devices are connected into the gateway via different communication standards and data protocols,e.g.LAN cable,USB cable,etc.The data protocols used in this implementation include Modbus over TCP/IP,some proprietary data protocol,etc.As shown in figure 8,a Modbus module is deployed on the data conversion gateway,which will retrieve edge device register data,parse and wrap the field measurements as JSON packet,then publish through MQQT to Amazon cloud.A local data storage can be optionally set up on premises as data backup to counteract communication lost between edge gateway and cloud.The entire software package in figure 8 can be dockerized,and easily deployed and updated on gateway remotely.

圖8 光伏管理系統(tǒng)邊緣層的數(shù)據(jù)轉(zhuǎn)換Fig.8 Data conversion at edge layer in PV management system

On the application layer,the functional structure for this PV management system are illustrated in figure 9.The information exchange between these micro-applications and back-end databases are through JSON-based REST API,which specifies the target resource location using HTTP URI and defines the functions to be performed on that resources by HTTP method.As shown in figure 10,a userfriendly graphical web portal is created to provides real-time monitoring and visualization of PV resources,performance analytic,system reporting,etc.

圖9 光伏管理系統(tǒng)的功能模塊Fig.9 Functional blocks of PV management system

圖10 光伏管理系統(tǒng)圖形用戶界面說(shuō)明Fig.10 Illustration of graphical user interface of PV management system

2.2 PV gene ration forecasting

Besides the basic monitoring and visualization functions,a short-term temporal PV power generation forecasting module is implemented and integrated into this system.The PV power forecasting module mainly consists of three sub-modules:offline training,online training,and online forecasting.These three functional sub-modules run in parallel and interact with each other,as shown in figure 11.Historical and forecasting weather data is integrated through third-party data API,e.g.national weather webservice,and applied as input for PV generation forecasting.

圖11 短期光伏發(fā)電預(yù)測(cè)時(shí)序圖Fig.11 Diagram of short-term temporal PV power generation forecasting

As shown in figure 12,the forecasting model parameters will be periodically trained and updated to accommodate the latest available information.The online PV generation forecast is set to operate every 10 minutes with forecasting horizon of future 24 hour.The forecasting resolution is 15 mins.

圖12 三階段PV電量預(yù)測(cè)時(shí)間線Fig.12 Timeline of three-step PV power forecasting

3 Conclusion

In this paper,an open IoT end-to-end analytical framework is proposed,which consists of three layers:edge layer,data management layer and modular application layer.This open analytical framework mostly relies on open hardware and open software tools.The open source protocols and standardized data infrastructure enable the future integration of various edge devices,and system scale-up with various advanced analytical applications.Also,the dockerizing of the framework components with latest container-based technology provides fast system deployment,consistent functionalities,and flexible compatibility with various computing environment.

An experimental validation of this proposed Io T end-to-end data analytical framework is implemented as a cloud-based PV management system for one commercial PV fleet owner.The cloud-based management system provides one uniform monitoring/management interface of multi-site commercial PV fleets,which bridges the silos between different vendor-specific monitoring interfaces.In the future,this proposed open end-to-end Io T data analytical framework will be further validated in other realworld use cases.