Keywords

1 Introduction

Photovoltaic energy has just reached a new dimension. According to the research of SolarPower Europe, in May 2022, the total global installed solar capacity has passed the 1 TW threshold [1]. That is 500 times increase compared to 2002, when the capacity was only 2 GW. It took 16 years, until 2018, to reach the 500 GW level, and only three years later, the amount has doubled to 1,000 GW or 1 TW. The forecast for the rest of 2022 shows an increase of 200 GW in one year. Solar power has become the third most significant renewable energy source by installed capacity, after hydro and wind power [1].

At the same time, however, solar still meets only a small share of around 4% of the global electricity demand, while non-renewable sources provide over 70%. In 2022, global solar additional installation capacity expects to increase by 36% to 228.5 GW.

The world will see solid demand for solar in the next four years, growing from 255.8 GW of additional capacity in 2023 to 347 GW in 2026. It will likely add 314.2 GW in 2025 [1].

Most commercial solar panels have efficiencies from 15% to 20% [2]. Therefore there is a demand to improve the efficiency of energy conversion. Although emerging perovskite solar cells have a high conversion efficiency (over 25%), there is still a long way for perovskite solar cells to get commercialized since perovskite crystals are easily decomposed in a humid environment. At the moment, the lack of stability is a significant disadvantage.

This paper reports on a project to diagnose and optimize the use of solar panel installations by developing a prototype sensor platform fixed to the panels and delivering data to a cloud-based service for analytics. The project was a collaborative effort between ICPE (Romania), EPRA (Turkey), and the University of South-Eastern Norway in 2021 and 2022. Photovoltaic panel output depends on several factors. The idea of Dr. Solar is to optimize the placement of the solar panels based on an analysis of such elements.

The essential characteristics of solar panels include the following:

  • Structural characteristics (roof-top mounted, ground mounted, wall mounted)

  • Type (monocrystalline, polycrystalline, thin film)

  • Geographical position (latitude, longitude, altitude)

  • Magnetic orientation, inclination, and base slop (roll)

Operational characteristics.

  • Solar radiation level

  • Solar panel back side temperature

  • Ambient temperature

  • Ambient relative humidity

The installation of the photovoltaic panels should consider some factors to make the most use of solar radiation: the material types, the geographical position, the elevation, inclination/angle, and solar radiation conditions. The correct measurement and adjustment of these factors ensure that photovoltaic panels produce maximum energy by being exposed to the greatest intensity of solar radiation during operation. Photovoltaic cells employ solar radiation in their operation; hence the cells are affected by the weather or condition of the atmosphere to which they are exposed. Here, the temperature is an essential factor which can be evaluated from two standpoints, i.e., ambient temperature and solar cell surface temperature. The electrical efficiency of solar cells depends on the ambient temperature. Solar cells are made of semiconductor materials like the most used crystalline silicon, and semiconductors are sensitive to temperature. Solar panels represent a negative temperature coefficient, and their performance declines when the temperature increases. The main reason is that an increment in the temperature decreases the band gap of a semiconductor which decreases the open circuit voltage of solar cells and the output power [3].

Regarding the cell surface temperature, the surface temperature can also be affected by the ambient temperature. However, the challenging point is the occurrence of hotspot areas. Hot spots are areas of high temperature that affect a solar cell by consuming energy instead of generating it [4]. As the solar cells are connected in series, a weak cell or group of cells will affect the energy production of all the cells on the same string.

The other factor is humidity. Solar radiation is electromagnetic waves, and when the light consisting of photon strikes the water layer, which is denser, refraction appears and decreases the light's intensity. Because the efficiency depends upon the value of the maximum power point of the solar cell, the effect of humidity deviates from the maximum power point, decreasing solar cell efficiency [5].

Humidity readily affects the efficiency of the solar cells and creates a minimal layer of water on their surface. It also decreases the total power output produced. Panjwani et al. [6] conducted more than four experiments with water absorbents to improve efficiency and make the system more efficient.

The next section provides a conceptual description of the Dr. Solar system. The following sections describe data collection, data analysis, and results. Finally, the last section contains the conclusion and ideas for further work.

2 Conceptual Description

Dr. Solar is a system designed to assess photovoltaic systems by collecting relevant measurement parameters followed by an analysis in the cloud using dedicated algorithms that give indications for optimizing energy production. The system consists of the following main parts:

  • Dr. Solar box containing various sensors

  • The inverter that collects and provides the data concerning energy production

  • A cloud-based platform that provides analytics service based on data obtained from the Dr. Solar box sensors and the inverter

The results from the cloud-based service are presented to the user as a report which contains, along with the presentation of the results regarding the configuration of the installed photovoltaic system, indications related to possible adjustments of the system to increase energy production. The inverter production-related problems are also included in the report. Figure 1 shows the main parts that compose the Dr. Solar solution.

Fig. 1.
figure 1

The main components of Dr. Solar

Full size image

The inverter transforms the DC low voltage gained from the solar panels into 220–240 V AC used in households. The prototype of the Dr. Solar device was used with a Huawei inverter SUN2000L-3KTL model to demonstrate the functionality.

The inverter collects and submits data about its input and output parameters, such as input and grid currents, input and grid voltages, active and reactive power, and energy production.

The Dr. Solar box also collects and uploads data into the same cloud-based platform, where analytics service is implemented using a GSM-based mobile router. A router is necessary since the solar panels are usually not placed in areas covered by Wi-Fi connection. The Dr. Solar box is set to collect data every 5 min for three consecutive days, ideally when radiation exceeds 600 W/m2. Both the Dr. Solar device and inverter provide data at five minutes intervals that are polled by the cloud-based analytics service for the time interval between 10:00 am to 2:00 pm.

3 Collecting Data

The Dr. Solar box is a sensor-based module designed to collect data from sensors and send them to the cloud platform to be processed. The Dr. Solar box is attached to the solar panel and stays there for a few days. The device can then be moved to a new location to collect and submit new data to be processed by the cloud service. The cost of the box does not justify a permanent mounting in only one location, except for large installations. During a few days, the Dr. Solar box will collect the data used for analytics. The sharing of the Dr. Solar box between many photovoltaic systems users justifies the costs of producing the sensors units/boxes.

Fig. 2.
figure 2

Dr. Solar Box

Full size image

Figure 2 shows the different building blocks of the Dr. Solar box. The Dr. Solar box includes several sensors relevant to photovoltaic systems investigation, as follows:

  • Solar radiation

  • Magnetic compass, inclination, and roll (base slope)

  • Photovoltaic panel temperature (measures the back-side surface temperature of the solar panel)

  • GPS tracker

  • Ambient temperature and relative humidity

Figure 3 shows the prototype of the Dr. Solar device. The radiation sensor is on the top front view right side. The wire on the left side is for the surface temperature sensor.

Fig. 3.
figure 3

The Dr. Solar box

Full size image

Figure 4 shows how the Dr. Solar device prototype mounted on the solar panel.

Fig. 4.
figure 4

The Dr. Solar box fixed to the solar panel

Full size image

Table 1 briefly describes the different components of the Dr. Solar box. The components are also shown in Fig. 5

Table 1. Dr. Solar box sensors
Full size table

The controller is a data logger that also has a protocol stack. The data logger can be used as a web client and server. It is the datalogger that submits information to the cloud-based service.

Fig. 5.
figure 5

Dr. Solar sensors: (a) Solar radiation sensor, (b) Ambient temperature and relative humidity sensor, (c) Photovoltaic panel temperature sensor, (d) Magnetic compass and inclinometer, (e) GPS tracker

Full size image

4 Analyzing the Data

The analytics process is visualized in Fig. 6. The left side shows the Dr. Solar device and the inverter. The cloud-based service is in the middle, and the web-based user-interface is on the right.

Fig. 6.
figure 6

The analytics process visualized

Full size image

Dr. Solar is the implementation of an idea of a cloud-based analysis and diagnosis platform for photovoltaic prosumers, shortened as (PVADIP-C). The proposed platform provides a cost-effective solution for small on-grid photovoltaic systems. The relevant onsite data are collected and sent to a dedicated cloud platform to be processed, analyzed, and diagnosed. Hence, there is no need to implement expensive control and optimization engines at small-scale photovoltaic locations. Two main functionalities are envisioned for the PVADIP-C. The outline of the proposed method is depicted in Fig. 7. As can be seen, the first functionality is an operational decision-making engine that optimizes a photovoltaic-based prosumer's operation. The second engine is the asset management engine which identifies the possible failures in the photovoltaic-based prosumer.

Fig. 7.
figure 7

Outline of the proposed methodology for PVADIP-C

Full size image

4.1 Operational Decision Support

This section deals with a small-scale photovoltaic system proposed operational decision support mechanism. Figure 8 depicts a schematic of a load point which includes a connection to the upstream grid, small-scale photovoltaic system, fixed loads, controllable loads (washer & dryer, dishwasher, and pump), storage device, and plug-in electric vehicle.

Fig. 8.
figure 8

Schematic of a load point with a small-scale photovoltaic system, flexible loads, and storage devices

Full size image

For optimal operation, the objective is minimum cost while satisfying technical constraints pertaining to nodal power balance, power transactions with the upstream network, and permissible operation range of elements.

The objective function can be formulated as follows:

$$ Minimize\;OF = \sum_{t \in I_T } {\left( {\lambda_t^{Buy} P_t^{\,Buy} - \lambda_t^{Sell} P_t^{Sell} } \right)} $$
(1)

  • t, ITI ndex and set of time

  • Buy, Sell Superscript for buying from and selling to the market

  • P Active power

  • λ Price per kilowatt

The primary constraint is to maintain nodal power balance, which is formulated as follows:

$$ P_{t}^{Gen} - P_{t}^{Load} = 0\quad \quad \forall \,t \in I_T $$
(2)
$$ P_t^{Gen} = P_t^{Buy} + P_t^{PV} + P_t^{Bat - } + P_t^{EV - } \quad \forall t \in I_T $$
(3)
$$ P_t^{Load} = P_t^{Sell} + P_t^{Fix} + P_t^{Flex} + P_t^{Bat + } + P_t^{EV + } \quad \forall \,t \in I_T $$
(4)

PV, Bat, EV, Fix, and Flex are superscripts for solar generation, plug-in electrical vehicles, fixed load at home, and flexible load at home, respectively. Here, + and − superscripts show the charging\discharging stage of battery storage and plug-in electric vehicle, respectively.

The technical constraints for small-scale photovoltaic operation are represented by (5)–(23). The power transacted between the load point and upstream network should be within a pre-defined range modeled by (5) and (6).

$$ 0 \le P_t^{Buy} \le \alpha_t^M P_t^{Buy,Max} $$
(5)
$$ 0 \le P_t^{Sell} \le \left( {1 - \alpha_t^M } \right)P_t^{Sell,Max} $$
(6)

M is the symbol for market-related quantities, Min and Max are symbols for lower and upper limits, respectively. In (5) and (6), the binary variable is used to avoid enabling both selling and buying options simultaneously. The upper and lower limits of photovoltaic generation are modeled by (7)

$$ P_t^{PV,Min} \le P_t^{PV} \le P_t^{PV,Max} $$
(7)

The state-of-charge (SOC) of the storage device at each period is calculated by (8) [7]. Constraint (10) is limitations on SOC, and (11)–(12) is limitations on charge/discharge power associated with storage.

$$ SoC_t^{Bat} = SoC_{(t - 1)}^{Bat} + \frac{{\eta^{Bat} \Delta t}}{{E^{Bat,Max} }}\left( {P_{(t - 1)}^{Bat + } - (\eta^{Bat} )^{ - 2} P_{(t - 1)}^{Bat - } } \right) $$
(8)
$$ SoC^{Bat,Min} \le SoC_t^{Bat} \le SoC^{Bat,Max} $$
(9)
$$ 0 \le P_t^{Bat + } \le \alpha_t^{Bat} P^{Bat + ,\,Max} $$
(10)
$$ 0 \le P_t^{\,Bat - } \le \eta^{Bat} \left( {1 - \alpha_t^{Bat} } \right)P^{Bat - ,\,Max} $$
(11)

η is the conversion efficiency coefficient, α is a binary decision variable, and E is the energy capacity for the storage unit. The EV charge/discharge constraints are modeled by (12)-(13) associated with storage. In (14), the SOC of parking lots in various time intervals is computed. Constraints on the target SOC of parking lots are stated by (15)-(17), respectively.

$$ 0 \le P_t^{EV + } \le \alpha_t^{EV} P^{EV + ,Max} $$
(12)
$$ 0 \le P_t^{EV - } \le \eta^{EV} \left( {1 - \alpha_t^{EV} } \right)P^{EV - ,Max} $$
(13)
$$ SoC_t^{EV} = SoC_{(t - 1)}^{EV} + \frac{{\eta^{EV} \Delta t}}{{E^{EV,Max} }}\left( {P_{(t - 1)}^{EV + } - (\eta^{EV} )^{ - 2} P_{(t - 1)}^{EV - } } \right) $$
(14)
$$ SoC^{EV,Min} \le SoC_t^{EV} \le SoC^{EV,Max} $$
(15)
$$ 0 \le P_t^{EV + } \le \alpha_t^{EV} P^{EV + ,Max} \left( {1 - SoC_t^{EV} /1 - SoC_t^{Sat - EV} } \right) $$
(16)
$$ 0 \le P_t^{EV - } \le \left( {1 - \alpha_t^{EV} } \right)P^{EV - ,Max} \left( {1 - SoC_t^{EV} /1 - SoC_t^{Sat - EV} } \right) $$
(17)

Equations (18)–(23) model controllable loads as follows:

$$ P_t^{Flex} = \alpha_t^{Flex} P^{Flex,Dep} $$
(18)
$$ \sum_{t \in I_T } {\alpha_t^{Flex} } = T^{Flex} $$
(19)
$$ \alpha_t^{Flex} - \alpha_{t - 1}^{Flex} = \beta_t^{Flex} - \zeta_t^{Flex} \quad \forall t \in \left[ {2,24} \right] $$
(20)
$$ \sum_{i = t - T^{Flex} + 1}^t {\beta_i^{Flex} } \le \alpha_t^{Flex} \quad \forall t \in \left[ {T^{Flex} + 1,24} \right] $$
(21)
$$ \sum_{i = t - DT^{Flex} + 1}^t {\zeta_i^{Flex} } \le 1 - \alpha_t^{Flex} \;\forall t \in \left[ {DT^{Flex} + 1,24} \right] $$
(22)
$$ DT^{Flex} = 24 - T^{Flex} $$
(23)

Dep is superscript for deployment, T is the time required for the complete operation of the equipment and are auxiliary binary variables. Equation (18) expresses that the power consumed by the flexible load at time t, i.e., is equal to its nominal power consumption while deployment if it is in operation, i.e., otherwise, is 0. In addition, the operation time of a flexible load should be equal to the required time for its complete operation, modeled by (19). Here, continuous operation of flexible load should also be considered. In other words, when the flexible load is committed, it should be continuously in operation until the time required for its complete operation (interruption is not allowed). To this end, the suit of constraints represented by (20)–(23) is added.

The outputs of the optimization engine are:

  • Optimal amount of photovoltaic-based generation to be consumed at each hour.

  • Optimal amount of power to be traded with the upstream network.

  • Optimal charging\discharging pattern for battery storage and electric vehicle.

  • Optimal commitment of controllable loads.

  • Hourly cost of community operation.

The required inputs for the proposed model are:

  • Hourly energy price ($/kWh).

  • Hourly fix load (kW).

  • Hourly generation of the photovoltaic generation (kW).

  • Installed capacity for renewable generation, battery storage, and electric vehicle.

  • Permissible operating range of devices.

From the inputs mentioned above, the hourly fix load and the photovoltaic generation can be acquired from the Dr. Solar box. The user or the operator can enter additional information. With this data in place, the devised model offers the optimal operation for a load point equipped with a small-scale photovoltaic system. Note that Dr. Solar’s optimal decision support service can be offered when Dr. Solar is installed for a considerable period at the customer’s place, for instance, a month.

4.2 Asset Management

This section deals with devising an asset management approach for small-scale photovoltaic systems. The main objective is to detect any failure or deficiency that results in the efficient operation of a small-scale photovoltaic system. The outline of the proposed approach for asset management of small-scale photovoltaic systems is depicted in Fig. 9.

Fig. 9.
figure 9

Outline of the proposed approach for asset management

Full size image

The main objective is to estimate the small-scale photovoltaic system’s output power and ensure that all system elements are intact. The estimated output power is then compared with the measured power, and a failure occurrence is reported in case of substantial difference. Here, the proposed methodology calculates the DC side parameters for accurate decision-making. The detailed calculation within the asset management toolbox in Fig. 9 is depicted in Fig. 10.

Fig. 10.
figure 10

The detailed calculation within the asset management toolbox

Full size image

In Fig. 10, three types of inputs are required. The first category is the measurement which is solar irradiation and ambient temperature. The second and third groups are the data on solar cells and the inverter, accessible through the data sheets. Once the input data are acquired, the next step is calculating the power at the DC side of the inverter. The DC power is calculated as [8]:

$$ I_{mpp} = I_{mpp,ref} \frac{G}{{G_{ref} }} $$
(24)
$$ V_{mpp} = V_{mpp,ref} \ln \left( {e + c\left( {G - G_{ref} } \right)} \right) $$
(25)
$$ P_{DC} = V_{mpp} I_{mpp} $$
(26)

Impp, ref, and Vmmp, ref is the current and voltage associated with maximum power point. G and Gref are solar irradiation and associated reference value. PDC is the output DC power, excluding the effect of ambient temperature. The DC power can be affected by the ambient temperature, which is considered. The cell temperature should be calculated as follows [9]:

$$ T_c = T_{air} + \left( {T_{NOCT} - T_{ref} } \right)\frac{G}{{G_{ref} }} $$
(27)

Tair is the ambient temperature, TNOCT is the normal operating cell temperature, and Tref is the reference temperature. Once the cell temperature is calculated, the DC power, including the ambient temperature effect, can be calculated as follows:

$$ I_{mpp}^{Temp} = I_{mpp,ref} \left( {1 + a\left( {T_c - T_{ref} } \right)} \right)\frac{G}{{G_{ref} }} $$
(28)
$$ \begin{gathered} V_{mpp}^{Temp} = V_{mpp,\,ref} \left( {1 + b\left( {T_c - T_{ref} } \right)} \right) \hfill \\ \quad \;\, \times \ln \left( {e + c\left( {G - G_{ref} } \right)} \right) \hfill \\ \end{gathered} $$
(29)
$$ P_{DC} = V_{mpp} I_{mpp} $$
(30)

The superscript Temp stands for ambient temperature-affected quantities.

As can be seen, the DC power in Fig. 10 is calculated for both with and without the effect of temperature. The main reason to calculate these two parameters is to identify an occurrence of overheating. The next step is calculating the AC power, which is realized using the DC power and inverter data outputs. The AC power is calculated as follows [10]:

$$ \begin{gathered} P_{ac} = \left( {\frac{{P_{aco} }}{A - B} - C\left( {A - B} \right)} \right)\left( {P_{DC}^{Temp} - B} \right) \hfill \\ \quad \; + C\left( {P_{DC}^{Temp} - B} \right)^2 \hfill \\ \end{gathered} $$
(31)
$$ A = P_{dco} \left( {1 + C_1 \left( {V_{DC} - V_{dco} } \right)} \right) $$
(32)
$$ B = P_{so} \left( {1 + C_2 \left( {V_{DC} - V_{dco} } \right)} \right) $$
(33)
$$ C = C_o \left( {1 + C_3 \left( {V_{DC} - V_{dco} } \right)} \right) $$
(34)

Paco is the maximum AC-power rating for an inverter at reference or nominal operating condition, Pdco is the DC-power level at which the ac-power rating is achieved at the reference operating condition, Vdco DC-voltage level at which the ac-power rating is achieved at the reference operating condition, Pso is DC-power required to start the inversion process, or self-consumption by inverter, strongly influences inverter efficiency at low power levels, and C0, C1, C2, and C3 are constants.

Fig. 11.
figure 11

Failure identification

Full size image

Once the powers at AC and DC sides are calculated, they will be evaluated using the approach depicted in Fig. 11 to identify the failure.

5 Results

Dr. Solar employs an efficient web platform as the user interface where the user or the operator can check the measured values and implement desired settings. This section summarizes the input and output pages that the user/operator would be faced while working with Dr. Solar’s solution.

5.1 Optimum Decision Support

Figure 12 depicts the input and output pages for the optimum decision support service of the Dr. Solar box. The input page acquires the daily fixed load and photovoltaic generation from Dr. Solar’s measurements. The electricity prices are based on historical data. For the selected date, the user\operator is asked to choose the preference on including the battery storage system (if available) and electric vehicle (EV) (If available) in the optimization process. In addition, the availability of any other flexible load is also asked for and considered in the optimization process.

Fig. 12.
figure 12

The user interface for optimal decision support page: input page

Full size image

The output page, i.e., Fig. 13, shows the optimum schedule offered by the optimum decision support algorithm and associated saving for the customer.

Fig. 13.
figure 13

The user interface for optimal decision support page: output page

Full size image

5.2 Asset Management

Figure 14 depicts the inputs required for performing the asset management algorithms. The required inputs are based on the datasheet of the photovoltaic power plant, which the operator can easily access. Then, the asset management evaluation period is asked. At least three days are required for decision-making with an acceptable level of accuracy. When the evaluation is finished, the output page, as depicted in Fig. 15, will be shown to the operator/user. Here, the main output is the message which shows the potential deficiency in the solar photovoltaic system. The list of possible output messages is as follows:

  • No failure is detected

  • Defect in solar module

    • Defective module or short circuit of by-pass diode

    • A defect in the blocking diode or one of the rows of modules is disconnected

  • Defect in inverter

Fig. 14.
figure 14

The user interface for the asset management page: input page

Full size image
Fig. 15.
figure 15

The user interface for the asset management page: output page

Full size image

6 Conclusion

This paper discusses the development of Dr. Solar, a sensor-based device for solar panel optimization and diagnostics. The device uses sensors to measure solar radiation, photovoltaic panel surface temperature, temperature, relative humidity, and position. The data collected by the Dr. Solar box is combined with data collected from the inverter and sent to a cloud service for processing. Using artificial intelligence algorithms, the data is analyzed, and the results are presented to the user through a web-based interface. The results will provide the user with information on faults or degradations, as well as advice on optimizing adjustments of the physical position (tilt and angle).

6.1 Future Work

A companion paper (to be submitted) describes business model considerations. Another work in progress discusses neighborhood sentiments toward solar panel adoption based on aesthetics. For this research, we use a 3D household model with the possibility of adjusting tilt and angle.