Commercial PV System Data Monitoring, Part One

Project owners, developers, contractors and financers all have a significant monetary stake in PV system performance. Without accurate and reliable data monitoring, evaluating and maintaining optimal system performance is just a guessing game.

For years, PV system data monitoring was mainly an afterthought, used primarily as an educational tool or for publicity value. However, data monitoring is now required for utility-scale PV systems, where it is used to track performance and comply with regulatory reporting requirements, and increasingly used in commercial applications. Despite happening in fits and starts, the research and development of commercial PV monitoring systems is resulting in more creative and ingenious solutions that streamline system integration and installation. At the same time, the variety of solutions and providers is increasing.

These coincident trends offer an opportunity and a challenge. PV system integrators and installers are used to developing and implementing code-compliant electrical and mechanical designs. However, mastering the myriad variables that an IT component adds—with its new terminology, hardware, architectures and configurations—can prove challenging. In addition, while the cost for specific PV monitoring services is declining, the total price to monitor large PV systems may actually be increasing as more functionality is expected from data monitoring systems as a whole.

Working in PV project development and construction, we have seen firsthand the difficulties developers and integrators encounter when data monitoring systems are not specified until near project completion. To control the associated component and installation costs and optimize system reliability, data monitoring must be given a seat at the design table.

In this two-part article, we describe the value proposition of monitoring commercial PV systems. In Part One, we consider data monitoring options, components and selection criteria, as well as the pros and cons of various levels of monitoring granularity. We explain how monitoring networks transfer information from one place to another. In Part Two, we will examine site-specific considerations, potential design and hardware responses, and provide a cost analysis case study. Throughout, we reference best practices and common mistakes.

Value Proposition

Because monitoring systems provide a remote visual representation of PV system performance, they are valuable tools for system owners, investors, installers and operators. One way to ensure that PV systems are operating optimally is to physically go on-site to take instantaneous irradiance, cell temperature and inverter output power measurements. You can then calculate whether the actual system output power acceptably approximates the expected output power. (See “PV System Commissioning,” October/November, 2009, SolarPro magazine.) Alternatively, you can specify a monitoring system to continuously provide the desired level of performance assurance.

Over the past few years, a major shift has taken place in the way PV systems in North America are evaluated. The standard of evaluation used to be based on capacity (kW or MW), meaning either the installed dc nameplate-rated power (kWSTC) or the capacity at PVUSA test conditions (kWPTC). Now the emphasis has moved to ac energy production (kWh or MWh) or specific yield (kWh/kW or MWh/MW). According to Bill Reaugh, VP of project development at Draker Laboratories, a data monitoring hardware and services provider, “The PV industry in the US and Canada has moved from simply trying to install the most capacity possible to trying to get the best energy harvest, because we are catching up to Europe both in operating practice and incentive structure.”

The goal of plant production estimates is to calculate expected energy production or specific yield for a PV system as accurately as possible by modeling component performance across a range of operating and environmental conditions. However, system downtime is potentially the greatest loss factor for total system performance. Even if you have the highest efficiency modules mounted on the most sophisticated tracking system available, money is lost every minute that the sun is shining and the system is out of service or operating suboptimally, for whatever reason.

In the experience of Thomas Tansy, VP of business development at Fat Spaniel Technologies, a monitoring company recently acquired by inverter manufacturer Power-One, PV plant monitoring is a given in utility- and industrial-scale applications. “In this context, data monitoring has the potential to generate the highest return on investment and is required 100% of the time,” he explains.

Data monitoring systems that are well designed, installed and maintained can ensure that a PV asset achieves the highest return on investment by minimizing operations and maintenance (O&M) costs and system downtime. Not only is the risk of underperformance or nonperformance unacceptable to owners and investors, but federal laws also define and specify data monitoring requirements for utility-scale PV systems that fall under the jurisdiction of the Federal Energy Regulatory Commission and the North American Electric Reliability Corporation.

“No traditional power plant would be built without extensive monitoring for determining that it is operating properly at all times and figuring out what is wrong when it is not,” notes Chuck Wright, principal at PowerDash, a monitoring services provider. “Likewise, renewable energy will not be a serious component of the world energy supply unless monitoring is an integral component. It is just a cost of doing business properly.”

Nevertheless, Power-One’s Tansy observes, “At the residential scale, data monitoring is often considered optional because performance-based incentives are generally not involved.” While it is certainly possible to build a business case for offering data monitoring as a standard feature on residential PV systems (see Summarizing the added value gained from data monitoring, Blair Kendall, director of business development at Southern Energy Management, a North Carolina–based PV integrator, says: “High-quality, accurate and accessible solar PV monitoring for commercial systems serves two primary objectives. The first is to provide certainty to CFOs and investors that they are getting what they pay for. The second is to facilitate effective O&M on the system to ensure maximum system uptime and production. Both of these goals are really about mitigating investment risk that further facilitates greater investment in commercial PV systems.”

Value to owners and financial backers. According to Adrian De Luca, VP of sales and marketing at Locus Energy, a provider of software solutions to the distributed renewable energy market, the value of data monitoring for system owners and investors is twofold. “First, monitoring systems enable accurate and timely customer billing,” he notes. “Second, they maximize system uptime and therefore the return on investment.”

For owners and backers of commercial-scale PV systems in particular, the ability to manage a portfolio of distributed PV plants in a unified manner may be as important as the ability to track individual plant performance. Many monitoring solutions providers offer multiplant, portfolio-level management. The caveat, of course, is that data for every site must be centralized with a single vendor.

Because they can optimize system performance and return on investment, data monitoring solutions also reduce financial risk. Financing large PV systems is often contingent upon having a performance guarantee contract in place as a risk mitigation mechanism for investors. (See “PV Performance Guarantees” June/July, 2011 [Part One] and August/September, 2011 [Part Two], SolarPro magazine.) Data monitoring is central to every performance guarantee—it gives the guarantee its teeth and makes it enforceable.

Without accurate data monitoring, actual system performance in the field cannot reliably be compared to what was guaranteed. Therefore, performance guarantee terms need to outline minimum data monitoring requirements commensurate with the performance risk. Uncertainty in data collection may make it difficult, if not impossible, to collect damage payments.

While performance guarantees may not be in place for the majority of small- to medium-sized commercial PV systems, the basic premise holds. Effective data monitoring not only helps to identify system performance problems, but it also helps to resolve them.

“Constantly measuring power production against expected performance benchmarks allows the owner to determine if and when remedial action is necessary,” explains Mark Lane, director of product management at ArgusON, a provider of site monitoring and services. “Assisting system integrators or O&M providers to resolve problems and get solar power systems back online as quickly as possible is part of the value that data monitoring adds for owners and financiers,” he concludes.

Public kiosks, wall-mounted displays and open-access web-based interfaces—common monitoring system features for educational institutions and brand-conscious businesses— also have value for system owners. These features add a visual or interactive component to an otherwise invisible electrical process. If a company or institution wants to showcase the fact that it has made an investment in solar for PR purposes, it can do so via a kiosk or wall-mounted display in a building lobby, online via its website, or both.

Educational displays are especially important when local or federal government entities are involved, either as customers or underwriters. “Many projects funded by the recent Federal ARRA program actually require public education displays,” notes Reaugh of Draker Laboratories.

Value to installers and O&M providers. Although there are few moving parts, PV systems do require maintenance, and monitoring systems can allow integrators to efficiently allocate resources and quickly identify essential tasks.

“The role of the PV integrator no longer stops at the commissioning of the system,” notes David Boynton of Southern Energy Management. “With 25-year equipment warranties, the expectation is that the system will be up and running for decades,” he continues. “Performance guarantees, operations and maintenance contracts and limited installation warranties are now commonly included in a project’s scope of work to ensure that the system continues to perform as expected.”

While monitoring can bring some new challenges during installation, most integrators attest to the long-term benefits in terms of time and money savings for O&M services. According to J.R. Whitley, Southern Energy Management’s O&M manager, “Accurate and reliable monitoring is quite possibly the most important tool in the operations and maintenance tool kit, allowing immediate notification of system issues that without monitoring would not be discovered until a scheduled maintenance trip or an angry call from a system owner.”

Commercial monitoring systems are Internet-based, providing installers remote access through web portals. This enables centralized operations to manage maintenance and service activities for systems that are physically spread out. These portals can provide quick verification of system performance based on actual environmental conditions and also allow for in-depth analysis of actual system performance versus predicted performance.

Alerts signaling low system performance or equipment alarms appear on the portal to direct attention to potential issues. These alerts and alarms can also be sent directly to the service department or project manager via email, text message or both, as specified by the in-house monitoring system administrator.

The efficiency of O&M activities can also be increased with effective system monitoring. For example, the additional level of detail that inverter-direct, string-level or module-level monitoring affords can enable a system integrator or O&M provider to remotely troubleshoot the type of failure or to identify the general vicinity or even the exact location of a failed component.

“Maintenance costs are reduced if maintenance crews no longer have to spend significant time troubleshooting problems,” notes Jeff Krisa, senior VP of sales and marketing at Tigo Energy, a provider of module-level hardware and software solutions for monitoring and optimizing PV plant performance. “Armed with the right information, crews can go straight to the source of the problem,” he continues, “and they can bring exactly what they need to fix it.”

Monitoring systems are also useful for reducing unnecessary truck rolls due to false alarms. Whitley observes: “Everyone has gotten that phone call: ‘Why didn’t my system produce as much energy this June as it did June of last year?’ Having analytic tools at your fingertips allows you to deliver a logical explanation, backed up with tables and graphs of timestamped data. This is preferable to the alternate response, ‘I think it has been cloudier this June than it was last year,’ which never leaves anyone satisfied.”

In addition to the many specific uses that system integrators and O&M providers have for PV data monitoring over the service life of an operational PV system, there is a potential value to the installer at the point of sale. In his SolarPro article, “Making the Case for Residential PV System Monitoring,” Brian Farhi, VP of business development at SolarNexus, a provider of solar business management software, observes that “[PV system monitoring] presents a great opportunity to differentiate yourself from your competitors, which can give you a competitive advantage in a tough market.”

Networking Explained

By Bill Reaugh, VP of project development, Draker Laboratories

A TCP/IP (transmission control protocol/Internet protocol) network is created in several layers, or programming abstractions: the link layer, the Internet layer, the transport layer and the application layer. However, TCP/IP is only one type of network that can be used to transfer information from one place to another. Open data protocol (ODP) is another type of network architecture; Modbus and CANbus are others. Each network type has varying numbers of layers, but all are built up in a similar way.

Each layer of abstraction allows the layers around it to function without having to know the specific programming needs of the others. Information is passed through each layer, from the source to the destination, based on the needs of the network.

Physical layer. A data monitoring system’s physical layer is the network hardware. This includes the cables, jacks, connectors, computers and other physical devices that are connected together. While this is not technically a layer of programming abstraction, it is required for the other layers to exist. For example, CAT 5 cable, RJ-45 connectors and jacks and Ethernet cards are all physical components in a TCP/IP network.

Link layer. This is the basic structure used to connect one device to another. The link layer is where individual devices are addressed, generally by a media access control (MAC) address in a local area network (LAN) or wide area network (WAN). Note that MAC addresses are specific to the hardware and generally permanent. This is also the level where things like virtual private network (VPN ) connections are created.

Internet layer. Message routing takes place at the Internet layer. Devices are assigned an IP (Internet protocol) address that in a manner of speaking tells physical layer devices, like routers and switches, who they are. These IP addresses act as a proxy for the MAC addresses used in the link layer.

IP addresses can be dynamic, meaning that the device or MAC address currently using an IP address may be different today than it was yesterday or even 5 minutes ago. Devices called dynamic name servers (DNS) keep track of the MAC and IP address associations as they change and also allow for domain names to be used in lieu of IP addresses. For example, you probably do not recognize IP address—but because of DNS it has a recognizable domain name:

Transport layer. The protocol used to send IP data packets is assigned at the transport layer. The most common is TCP (transmission control protocol), but UDP (user datagram protocol) is also widespread. TCP and UDP provide the structure and error checking required for a packetized data transmission system.

In a packetized system, a message is broken into small fragments. Each fragment is then sent across the network in a “best effort” system, meaning each finds its own way from source to destination. TCP and UDP rebuild the message from all of the bits, make sure they have all arrived and are in the correct order and request that missing ones be resent.

Application layer. The main user interaction with the system occurs at the application layer. HTTP (hypertext transfer protocol), FTP (file transfer protocol), SMTP (simple mail transport protocol) and many other protocols exist at this level.

HTTP, of course, is the backbone of what is commonly referred to as the Internet. It can be secured, encrypted or open, depending on the users’ and programmers’ desires. Applications at this level can be written in a variety of languages, but the most common include HTML (hypertext markup language), XML (extensible markup language) and JavaScript.

Basic Monitoring Concepts and Components

PV system designers, integrators and installers are rarely IT experts. However, it is important that they understand basic IT concepts and can identify the major components in a data monitoring system. PV system installers are often expected to install and supply power to data monitoring system components, as well as to source and install the necessary conduit, cable and connectors.

To the extent that data monitoring is part of the scope of work, it needs to be included in the plan set—and not just as a separate, vendor-supplied, single-line item. Ideally, PV system designers call out the conduits and receptacles that the data monitoring system requires within the electrical plan set for the PV system itself. After all, this is what the installers are working from.

To ensure successful execution in the field, the best practice is for PV system integrators to coordinate or partner with a monitoring system provider during the project planning stage. The concepts and components detailed in the following pages assist PV system designers with this process. Note that all communications systems should comply with Chapter 8 of the NEC.

Physical layer. The physical layer of a data monitoring system includes all of the hardware: sensors, meters, conduit, cable, loggers, wireless transmitters and receivers, combiner boxes, inverters and so on. The physical layer is what is generally represented on the plan sets and is the basis for connecting all of the necessary components of the monitoring system together. (The relationship between the physical layer of a data monitoring system and the programming layers is described in the sidebar above.)

Bus driver. A bus driver is a method of data transmission that defines how voltages or currents on a serial communication bus should be interpreted. Examples of bus driver standards are RS-485, RS-232, and RS-422.

Protocols. A protocol is a program language that electronic devices use to transmit data to one another. One common example is the Modbus protocol, published by Modicon in 1979. Modbus is one of the most widely used protocols in the US because it is an “open” protocol. An open protocol is one that manufacturers can build their equipment to use without paying royalties or license fees to the publisher. TCP/IP (transmission control protocol/Internet protocol) and DNP3 (distributed network protocol 3.0) are also widespread. Proprietary protocols come in a variety of flavors from various equipment manufacturers.

Ideally all devices in your monitoring system network use the same bus drivers and communication protocols. This is not always necessary, however: some dataloggers are capable of speaking more than one protocol and may even be able to do so simultaneously. When incompatibilities arise—generally this occurs with proprietary protocols—translation devices, or protocol converters, can be deployed.

“Putting in a little bit of time on the front end will save it tenfold in the field,” advises Whitley of Southern Energy Management. “Monitoring systems have so many options and possible configurations, and the installation manuals often leave something to be desired,” he warns. “If all the components are not sourced from one supplier, confirm that they are all using compatible protocols and have worked together on other sites. Otherwise, you run the risk of wasting time later chasing phantom alarms.”

Network. A monitoring network can be connected in a variety of ways—point to point, daisy chain or peer to peer— depending on the bus driver and protocols used. The simplest network consists of two devices connected directly to each other. The devices could be connected by a two-pair twisted cable, such as Belden 9842 cable, and use the RS-232 bus driver to transmit data via the Modbus protocol from one device to the other.

Using a similar type of cable, Belden 9841 or 3601A, it is possible to connect several devices in a daisy chain. In this case, an RS-485 bus driver can be used with the Modbus protocol to collect data from all the devices. Data is collected at a single point, usually a dedicated special-purpose datalogger.

Finally, a third arrangement is a peer-based network using TCP/IP and CAT 5 or CAT 6 cables. This system is installed much like a local area network for computers. However, each device would be an energy meter, inverter, combiner box, datalogger and so on.

Wireless devices could replace any or all of the cables in these sample networks and generally run their own bus drivers and protocols to simulate or replace those used on hardwired connections.

In conversation, Ethernet and Internet are sometimes used interchangeably, like PV panel and module, but they are not the same thing. An Ethernet network is any network built with TCP infrastructure in mind. The Internet is the collection of computers, servers and sites that we call the web. The ubiquitous “www” refers to an earlier nomenclature: World Wide Web.

While it is possible to create an Ethernet network that is separate from the Internet, the Internet cannot exist without Ethernet networks. An example of just such a separate network is a Modbus TCP network in a utility-scale PV plant. This network uses Ethernet devices—cables, connectors and switches—but uses Modbus at the application layer instead of HTTP or other protocols.

Communication cables. Generally speaking, communication cables use one or more twisted pairs of stranded (7x30 or 7x32) small-gauge wires, usually 18 to 24 AWG. Using twistedpair cable minimizes radiated and conducted electromagnetic interference (EMI).

Different bus driver standards allow for longer or shorter transmission distances. RS-485, the most commonly used bus driver, is specified to transmit data for up to 4,000 feet. This distance is achieved in part by the standard itself, which uses positive and negative voltages for 1s and 0s, as well as the EMI resistance offered by the cable type. RS-232 specifies positive voltages for 1s and no voltage for 0s. A bus driver that uses RS-232 is inherently more sensitive to line noise and interference. Even with the same EMI resistance, this specification allows connection only up to 50 feet.

CAT 5, CAT 5e and CAT 6 cables—which consist of four twisted pairs of wires and terminate with RJ-45 connectors— were created to carry larger amounts of data. While additional carrying capacity is needed as networks become more complex, CAT 5 or CAT 5e cables are sufficient for the transfer rate needed for most PV monitoring applications. The bus driver behind TCP/IP-based communication provides for a transmission distance of only 300 feet between networked devices. Devices like routers, switches and hubs may be used as repeater stations to increase the distance between devices.

Like other types of electrical cable, the wires in a data monitoring system are used to carry electricity. In this case, voltage and current signals are used to communicate data within the monitoring network. Longer cable runs reduce the force or amplitude of data monitoring signals in a process similar to voltage drop in power conductors. This gradual loss of intensity is known as attenuation.

The reason that distance limits are defined for each bus driver has to do with the ability of devices to differentiate between 1s and 0s as the signal is dissipated across the cable. Most communication protocols have error detection routines that identify data corrupted by signal attenuation, but this only forces the master device to request the same data multiple times. In most situations, it is best to keep cable runs as short as possible and to minimize the amount of EMI-producing equipment in the vicinity.

On a final note, be aware that cable manufacturers may have multiple cables with the same basic specifications—for instance, one twisted pair, 18 AWG, shielded, with drain—that are differentiated by insulation properties like oil resistance, burial rating, UV resistance and so forth. As when choosing power conductors, make sure that you are selecting communication cables with the proper environmental resistance properties for your application. Installing indoor-rated cable in a buried conduit results in a malfunctioning monitoring system in short order.

Datalogger. A datalogger, also called a data acquisition unit, is an electronic digital processing device that resides in the physical layer. This device records data over time and has internal memory for data storage. Some dataloggers can also function as analog-to-digital signal converters or Internet gateways, or have other capabilities.

In a network, a “master” device initiates communication between the “slave” devices and itself. In a PV monitoring system, the master device is usually the datalogger. Slave devices, which all have unique network addresses, may include inverters, weather station equipment, energy meters of revenue grade or lower accuracy, building load or net-energy meters and so on.

Weather sensors. Weather sensors measure the environmental conditions in which the PV system is operating. Examples of weather sensors include pyranometers to measure sunlight intensity (irradiance) and sun hours (insolation); thermometers, thermistors, thermocouples or other devices to measure cell or ambient temperature; anemometers to measure wind speed and vanes to measure wind direction; barometric pressure sensors; precipitation meters to measure rainfall; and many others.

While revenue-grade plant metering is essential for billing and reporting purposes, its value is limited within the context of O&M. Plant operators need additional information in order to determine if a PV asset is performing as expected and to optimize scheduled and unscheduled maintenance activities. Weather sensors provide this additional information.

According to Blake Gleason, director of engineering at Sun Light & Power, an integrator based in Berkeley, California, “Plane-of-array irradiance and module cell temperature are all that you really need to check an array’s performance ratio.” A pyranometer or reference cell mounted in the same plane as the PV array typically provides irradiance data. A thermistor or thermocouple mounted on the back of one or more PV modules typically provides cell temperature data.

As characterized by Matt Taylor and David Williams in part one of their SolarPro article on PV performance guarantees (June/July, 2011), the performance ratio for an operational PV plant “separates out the uncertainty and variability of irradiance and is intended to normalize out weather factors to produce a consistent measure of system performance.” It is an index of PV plant performance, usually expressed as a percentage rather than in units, that represents the ratio of actual metered PV output power as compared with the ideal irradiance- and temperature-corrected output power.

In the context of unscheduled O&M activities, the value of monitoring a PV plant’s performance ratio is that this index can provide an early indication of installation or commissioning problems. Tracking this index can also be used to optimize scheduled maintenance activities, such as array cleaning. This is because the cumulative effects of dirt and dust buildup show as a steadily declining performance ratio. A stepwise or gradually progressing decline in performance ratio can indicate other issues, such as blown string or subarray combiner box fuses, stolen modules or inverter failures.

“While it might seem unnecessary,” notes Southern Energy Management’s Whitley, “I recommend getting a full weather station package, with sensors for ambient air temperature, wind speed and direction, global horizontal irradiance and precipitation. The worst kind of data is the data you wish you had when you need it.”

Power-One’s Tansy concurs, explaining: “Precipitation data is useful for predicting when to wash panels. Wind speed and direction are useful for discriminating the effects of ambient temperature on performance versus some other heat-induced system defect.” According to Tansy, skimping on environmental monitoring is one of the most common mistakes made when data monitoring solutions are implemented: “Installers routinely omit environmental monitoring from their plant installations and then scratch their heads when system owners ask ‘Why didn’t my plant produce last Tuesday morning?’” he says. “The answer is often ‘it was raining’ or ‘the ice hadn’t melted from the panels’ or something similar.”

When you are designing PV systems for large acreages or sites with variable terrains, consider incorporating multiple weather stations for more specific performance ratio calculations. In many cases, projects scaled for utility-power production require multiple weather stations for this very reason. Tansy recommends that designers “segment large plants into subsegments and use a cascading design for plant data collection devices.”

It is also important to follow the manufacturer’s instructions when installing weather monitoring hardware. For example, the correct placement of irradiance and temperature sensors is critical.

“Many weather sensors provide output signals on the order of microvolts or millivolts,” notes Draker Laboratories’ Reaugh. “As a result, the sensitivity of the datalogger or analog-to-digital converter is of prime interest when reading the output of these sensors,” he continues. “An analog-to-digital converter that cannot distinguish measurements smaller than 1 millivolt is useless when connected to a sensor that outputs in microvolts.”

For a similar reason it is advisable to keep the cables connecting the weather sensors to the datalogger as short as possible to avoid the loss of signal strength.

Energy meters. In addition to solar irradiance and cell temperature, a third measurement is essential to array performance ratio calculations: delivered ac energy, which is recorded by the energy meter.

All revenue-grade energy meters conform to American National Standards Institute (ANSI) Standard C12.20 and are required to provide energy readings to within ± 0.25%. Energy meters collect data from current transducers (CTs) and may also have voltage transducers for high-voltage applications above 600 Vac. The energy meter processes the current and voltage information in a variety of ways and communicates it to the datalogger.

In its most basic form, an energy meter provides a continuous record of the kilowatt-hours that pass through the wires it is monitoring. More advanced meters can also provide additional information about the electrical system—power factor, reactive power, power quality, peak demand and so forth—based on whatever is of interest to the customer, the utility or the entity selling energy to the utility.

Most incentive programs now operating in North America require the use of revenue-grade meters to collect a record of the energy produced and a monitoring services provider to formally vet and report that data. All revenue-grade meters are calibrated against defined standards. The manufacturer or testing facility should be able to provide a calibration certificate from the National Institute of Standards and Technology (NIST), ANSI or an equivalent institution.

Some inverter manufacturers—such as Advanced Energy, Solectria Renewables and others—are now offering revenue-grade metering capabilities within their central inverters. Alternately, a revenue-grade meter can be part of the hardware package supplied by a third-party monitoring provider. Revenue-grade metering also offers the system owner and financers the ability to verify utility-provided metering and billing against a separate, independent dataset.

Note that many energy meters need an external power supply. Others draw power from the voltage connection they are reading. Consult the meter manufacturer’s documentation regarding power supply requirements, terminations and fusing.

Internet gateway. All monitoring services providers use a web-based interface to display, analyze and report data collected at the project site. To get the data from the project site to the Internet cloud, a gateway of some sort is required.

Collected data can be transmitted wirelessly via a cell phone or satellite modem or via a hardwired connection, such as CAT 5, cable, DSL, T-1 or fiber optic lines. Many utility-scale sites have no access to hardwired Internet connections, so cell modem reporting is generally the only option. This requires additional upfront hardware cost for the modem, plus ongoing monthly fees for cell service. However, the simplicity of the connection provided by a cell modem often offsets the time and trouble caused by negotiating with customers and their IT teams to get the necessary permissions established and maintained over the 25-year life of the project.

The frequency of data transmission varies by manufacturer— from about once per minute to once every 15 minutes. If Internet connectivity is temporarily unavailable, some dataloggers may store data in internal memory until reconnected. Dataloggers generally need a separate power source, which might output 120, 240, 277 or 480 Vac depending on the power supply configuration. Consult your vendor’s installation requirements to determine the needed power supply, terminations and fusing.

Web portal. A web portal functions as an access point for information on the Internet. While all currently available commercial and utility-scale PV data acquisition systems have some sort of web portal, their functionality and presentation format varies widely.

Multiple portal interfaces are common. For example, there may be an open-access customer portal in addition to a password-protected O&M portal. Commercial PV systems that serve educational or PR purposes may have yet another web portal.

Multiple portals are essential for utility-scale PV systems, which should have a portal for O&M that is separate from the owner portal. From the O&M portal, it is typically possible to view error alerts and manage work orders for all of the systems being monitored.

Degrees of Granularity

PV systems can be thought of as a network of subsystems, much like a river. Just as you could not determine the volume of water in a tributary feeding into the Mississippi River by measuring the total volume of water pouring into the Gulf of Mexico, you cannot determine individual PV component performance by looking at only the ac output at the end of the system.

In a PV system, power is combined at module, string, combiner box, subarray and inverter levels. It is possible to collect power data at each of these points. Higher levels of data granularity provide a more complete look at total system performance. The further a PV system is divided, the more available data there is for benchmark comparisons and troubleshooting.

Unlike a river, the electrical design for PV power systems is generally characterized by system symmetry—by the repetition of similar, if not identical, subsystems. This principal of symmetry is useful when determining the optimal degree of monitoring for a PV system.

To the extent that a PV system design is electrically symmetrical, it may be possible to reduce the amount of granularity needed to adequately monitor the system. For troubleshooting or O&M purposes, manually or automatically comparing the power output at equivalent collection points in a PV system can provide a level of functionality that is similar to installing a more complex and granular monitoring system.

For example, if a PV system designer plans to use 10 sourcecircuit combiner boxes to aggregate power into a single central inverter, each with an identical number of string inputs, then all 10 PV output-power circuit conductors should essentially carry the same amount of power when averaged across a period of minutes. Therefore, if the output of these combiner boxes is monitored, then a combiner box with lower-than-expected power output can be spotted at a glance by comparing the power curves taken at each combiner.

Assuming a symmetrical electrical design, source-circuit combiner box-level monitoring can provide string-level insight into system performance at a fraction of the cost and complexity of string-level monitoring. While monitoring at the combiner box level would not identify exactly which source circuit is not contributing, it can identify that a string has failed and, in this example, target the troubleshooting activities required to fix the problem to just 10% of the array.

Similarly, string-level monitoring can be used to provide module-level insight. In fact, monitoring pairs of ungrounded current-carrying source-circuit conductors may effectively achieve this level of granularity at a relatively reduced cost. The challenge for system integrators is determining at what point additional granularity no longer justifies additional costs.

“Deciding between different levels of monitoring generally breaks down into a cost versus granularity of data argument,” explains Locus Energy’s De Luca. “String-level monitoring enables operators to spot a given string that may be performing below spec, but installing source-circuit combiner boxes with string-level monitoring capabilities is often prohibitively expensive.”

There are, of course, always exceptions. As the value of the data being collected increases, higher degrees of granularity may be justified. In some markets the combined value of solar renewable energy credits (SRECs) and the rate of the energy being offset may warrant an investment in a very granular monitoring system. For example, if every PV-generated kilowatthour is valued at $0.50, then there is a significant incentive to optimize plant performance, as will be shown in a case study appearing in Part Two of this article.

According to Power-One’s Tansy, “Performance monitoring at higher levels of granularity enables the operator to fine-tune plant performance at a more granular level.” In addition to enabling detailed performance variance comparisons, it may also speed root-cause resolution. The faster and more accurately a problem can be identified, the faster it can be resolved.

Because the cost of data monitoring is proportional to the number of plant metrics collected and the granularity of the data, system owners want to optimize data monitoring, not just maximize it. Determining what constitutes excess data collected at too high a cost and what level of data collection provides the best return on investment is an installation- and project-specific exercise. Just keep in mind that the data you fail to collect today may turn out to be priceless tomorrow.

Freeman Corbin, director of sales at DECK Monitoring, says: “O&M service structure is the most important consideration for determining the appropriate monitoring level.” After all, the costs associated with future O&M services are typically going to be paid by the integrator.

If a company has a large crew that performs regular on-site system checks and the cost of this service is built into the system operating costs, then inverter-level monitoring might be sufficient. Similarly, if the O&M model pays the integrator to send someone into the field to check every string manually, then there is little incentive to install a granular monitoring system.

However, as integrators install more systems and their service territory grows, centralizing O&M by using a more sophisticated monitoring system might prove to be a better option. This can allow one person at the company to monitor an entire fleet of PV systems from a single web portal. More granular monitoring data can then be used to send troubleshooting crews out with a focused agenda or maintenance schedule, which may save time and money over the life of the system.

Beware of putting systems on autopilot, though. According to Bill Brooks, principal engineer at Brooks Engineering: “Too often, system operators think that with string-level monitoring they can sit back and wait for things to go wrong. This is a flawed approach. Data monitoring systems should never take the place of regularly scheduled preventative maintenance.”


The higher the level of granularity in your monitoring system, the more complex the monitoring system becomes. If you are collecting inverter-level data and information from a single energy meter, the number of communication circuits and data collectors is relatively small and thus simple to design and implement. However, when you are collecting large amounts of string-level data, the network becomes necessarily more complex, more devices are involved and the conduit and cable schedules get larger as well.

Inverter level. Total system or inverter-level monitoring has the advantage of being the least expensive option and the least complex to install. Typically the inverter or inverters are daisy-chained with cable appropriate for an RS-485 based Modbus network to the datalogger, revenue-grade meter and weather-station hardware.

This level of monitoring provides data on total system performance only. If there is a single inverter, it is difficult to benchmark production, because there is nothing to compare inverter output against. If issues arise in a system that is monitored at the inverter level, the entire array or subarray must be examined. Another disadvantage to this monitoring scheme is that performance problems can easily go unnoticed and persist until the next scheduled maintenance interval, if not longer.

When more than one inverter is used, at least some level of performance variance analysis is possible. With multiple inverters in the system, the inverters’ output power can be compared for consistency. This can reduce troubleshooting time by enabling a process of comparison and elimination. When many smaller-capacity inverters—single-phase string inverters, for example—are individually monitored and compared, the net effect is equivalent to combiner box outputlevel monitoring and may even provide string-level insight into system performance.

Combiner box output level. The next level of complexity and granularity is combiner box output-circuit monitoring, which is also referred to as zone or subarray monitoring. This level of monitoring can be accomplished in several different ways, depending upon the system configuration and the locations where measurements are taken.

The simplest way to achieve zone monitoring for commercial-scale PV systems is generally to specify subarray monitoring at a central inverter’s fused input combiner. In some cases, the manufacturer offers this option.

“We currently offer system-, inverter- or subarray-level monitoring,” notes Michael Zuercher-Martinson, CTO at inverter manufacturer Solectria Renewables. “Subarray monitoring comes factory-installed within the inverter and gets us 80% of the string-level monitoring benefits for 20% of the equipment cost,” he continues. “No additional wiring or IT setup or configuration is required.”

Alternatively, system designers can specify smart subarray combiners within the array field. These large combiner boxes are outfitted with CTs and bolt-in fuses that aggregate PV output conductors from source-circuit combiner boxes. Subarray combiners are useful on large commercial systems. For example, if a 500 kW central inverter with six 450 A fused inputs is monitored at the input combiner only, then there is very little granularity to zone monitoring. The loss of a single string is very difficult, if not impossible, to detect. However, if six smart subarray combiners are specified upstream from the inverter, then the granularity of the zones being monitored can be improved until string-level resolution is achieved.

While combiner box output-level monitoring is more expensive than inverter-level monitoring, it provides much more information about the performance of the system. It improves the resolution of the data used for comparative analyses, allowing for quick baseline system performance verification. If differences are detected in combiner output circuits, troubleshooting efforts can be targeted at a much smaller section of the array.

Planning becomes extremely important when monitoring at this level, especially if communication circuits extend into the array field. Additional components are needed, such as a smart combiner box with CT equipment. A communication line needs to be run to each combiner box location, unless a wireless solution is available. Each combiner box requires its own power supply. For best results, work directly with a monitoring solutions provider on the specific design details as the project is conceived and developed.

String level. If more granular data is desired, the resolution of the data monitoring system can be extended to the level of the source circuits or module strings. Further dividing the system in this manner allows owners and operators to pinpoint performance issues and reduce on-site troubleshooting time.

The equipment necessary for this level of monitoring is very similar to that needed for smart subarray combiner monitoring. However, additional CTs are needed within the monitoring system. Instead of CTs located at the subarray level on PV output circuits, string-level monitoring requires smart sourcecircuit combiner boxes, which typically have a CT for every one to two source circuits.

A string-level CT unit—like the Multi Circuit DC Monitor manufactured by Ovius—typically has eight noncontact Hall Effect sensors and provides a Modbus RS-485 output for monitoring eight module strings. The units are modular in the sense that multiple CT units can be incorporated into a single enclosure for 16-, 24- or 32-circuit combiner boxes, which are available from companies like AMtec Solar and SolarBOS. The Ovius CT units require a 24 Vdc power source, which typically means that a 120 Vac circuit needs to be run to each combiner box in order to power a 120 Vac/24 Vdc power supply.

While string-level monitoring can be effective for detecting faulty equipment and low-performing modules, it also adds a great deal of complexity to the PV data acquisition system. For this monitoring to be effective in the long term, a thorough review of the plant is necessary during system commissioning to map each string to its corresponding sensor. Despite the best intentions of power system and data acquisition system designers, changes in the field do happen. A well-documented set of as-built drawings is crucial for troubleshooting later with this level of data analysis.

Module level. Module-level monitoring is not currently offered by independent third-party monitoring solutions providers. Rather, it is a side benefit of installing module-level power electronics: ac modules, microinverters or dc-to-dc optimizers. At present, these three technologies are rarely seen in commercial PV applications due to the economy of scale that larger string or central inverters provide.

While this level of monitoring can obviously be very effective in detecting low-performing modules, the sheer amount of data generated in a large array necessitates a computer-driven model for detecting anomalies. As with string monitoring, for the data to be useful later for troubleshooting in the field, mapping the devices to the sensors is a crucial commissioning step—one that becomes more important with each higher level of monitoring system granularity.

“The best practice that we can recommend is to make sure that the asset owner or maintenance team has a good plan for making use of the data we are providing,” says Krisa of Tigo Energy. He explains: “Instrumenting a system well is just the first step. The system owners must also ensure that the organization and business processes are in place to use the data once they have it. We can help by providing the tools to manage fleets of systems and summary metrics that can mitigate the issue of having too much data. But our best customers also have the team and operational plan in place to make use of the tools that we provide.”

System Implementation

In Part Two of this article, we will consider commercial PV data monitoring system selection and specification criteria in more detail. Completing a site survey is an important early step. You need to decide what data must to be collected and what web portal views are necessary. You need to provide Internet access. The location of data acquisition system components drive conduit and circuit routing. Making these design decisions early is critical, so that installation crews can efficiently execute the scope of work associated with the data monitoring system.

Special thanks to Bill Reaugh at Draker Laboratories for providing expert technical review services and input during the preparation of this article.


Kyra Moore / Southern Energy Management / Durham, NC /

Rebekah Hren / 02 Energies / Durham, NC /


Third-Party Commercial PV Monitoring Providers:

AlsoEnergy / 866.303.5668 /

ArgusON / 866.459.4103 /

DECK Monitoring / 503.224.5546 /

Draker Laboratories / 866.486.2717 /

Fat Spaniel (Power-One) / 408.785.5200 /

Locus Energy / 877.562.8736 /

Solar-Log / 203.702.7189 /

Article Discussion

Related Articles