Hybrid Power System for the Range Extension of Security Robots: Prototyping Phase

Abstract

This paper describes our best practices related to hybrid power system (HPS) development with a focus on the prototyping phase. Based on the main development goals of our security robot, 24 h continuous operation on a single charge as a top priority, the HPS specifications were developed in the previous phase. For long-duration missions, batteries are hybridized with hydrogen fuel cells. By hybridization, the practical issues of fuel cells can be addressed such as lack of durability and low power density. With the developed specifications of the HPS, its components were acquired and installed to build a prototype. Using an electronic load coupled with a charge-discharge system controller, the constructed prototype was tested, discovering the maximum output power (850 W) that the fuel cell can sustain for 24 h. To further increase the energy density of the HPS, its structure was converted to a plug-in hybrid. With the developed HPS simulator, the converted HPS was simulated, predicting an extended hours of operation (2.07 h) based on the larger battery (7S12P) over the widest SOC window (90%). The plug-in HPS prototype was integrated into the security robot. On a dedicated chassis dynamometer, the integrated prototype was tested, demonstrating its capability to continuously operate the security robot for 24 h.

1. Introduction

As the demand for mobile robots continues to grow, they are expected to perform long-duration missions [1,2,3,4,5,6,7,8]. Physical security is one of the missions that are also difficult, dangerous, and tedious. Commercial security robots have been introduced as supplements to help human guards [9,10,11,12,13,14]. For long-duration missions, power systems are of particular importance in a mobile robot. The energy demand will depend on a task given to the mobile robot, which usually lasts for a few days or even a week rather than just a few hours. The power demand will vary significantly during a task, often to levels that are instantaneously higher than that power systems can constantly supply. Power systems for use in mobile robots are constrained in their weight, volume, noise, and emissions. Conventional power sources are key limiting factors for mobile robots under many important missions. Internal combustion engines have high energy densities but produce toxic exhaust, loud noise, and strong thermal signatures, making them inadequate for confidential operations. Despite recent advances in their performance, rechargeable batteries still have relatively low energy densities and thus need to be stopped and recharged every few hours, making them ineffective for continuous operations. Therefore, there is a significant demand for an alternative power source that can supply as much energy as necessary for security missions that are long-lasting, clean, and quiet.

Fuel cells are electrochemical devices that convert chemical energy into electrical energy. In terms of energy conversion, fuel cells are outwardly similar to batteries or internal combustion engines that are used to supply electricity. However, unlike internal combustion engines, fuel cells can generate electricity directly from electrochemical reactions without multiple energy conversions that involve heat and mechanical motions. In contrast to batteries, fuel cells can generate as much electricity as needed as long as they are supplied with fuels. This is the key advantage of fuel cells over batteries, to be specific, sealed batteries, enabling them to work for longer duration even with much shorter refueling times.

While fuel cells show great promise in theory, they have a few critical issues that must be addressed for practical use. These include difficulties in fuel storage [15,16,17], lack of durability [18,19,20], and low power density [21]. Of these, the difficult fuel storage and low power density are inherent problems. Hydrogen is costly to store in bulk primarily because of its high energy content but low density. Fuel cells are limited in how rapidly they can increase their fuel supply rate, especially oxygen to the cathode, as the power demand grows. From a systematic point of view, the low durability is a problem in use, which can be mitigated by running fuel cells in a more controlled manner. Among the various factors known to influence the durability of fuel cells, voltage oscillations and high voltages during operation are of particular interest. The voltage oscillations are relevant to a situation where mobile robots operating in unstructured environments are subjected to large variations in load demand. Without proper control, the voltage oscillations can increase the risk of fuel cell degradation, resulting in a short life span and premature failure. In this regard, the poor durability can be overcome by power system hybridization [1]. By hybridizing with batteries, fuel cells can be controlled to run within a narrower operating area, which extends their lifetime. By regulating the operating area, higher efficiency can be obtained. Besides, the low power density can also be remedied to some extent.

This paper is intended as a sequel to our previous work. In our original paper [22], we described our best practices related to hybrid power system (HPS) development with a focus on the specification development phase. We first comprehensively compared several HPS structures and selected the best configuration for our security robot. Based on the selected structure, we then determined the component specifications that could help us achieve our major development goals: 24 h continuous operation on a single charge. We subsequently designed a simulator to verify the validity of the determined specifications. The determined specifications of the HPS components are listed in

Table 1. Procured component specifications.

Fuel Cell	Model	H-1000W	Horizon Fuel Cell Technologies
	Type	PEM	Proton exchange membrane
	Configuration	48S	Series
	Voltage [V]	45.5~28.8
	Power [W]	1000	Nominal
Fuel Storage	Model	SCBA-9.0L	Inocom
	Type	Composite	Full-wrapped carbon fiber
	Configuration	4P	Parallel
	Mass [kg]	0.71	Compress at 310 bar
Battery	Model	18650-25R	Samsung SDI
	Type	NMC	Nickel manganese cobalt
	Configuration	7S8P	Series and Parallel
	Voltage [V]	29.4~17.5
	Capacity [Ah]	20	Nominal
DC/DC Converter	Model	SD-1000L-24	Mean Well
	Type	Non-isolated
	Voltage [V]	19~72 to 24	Input to Output
	Power [W]	960	Rated

. We finally summarized our lessons learned in this phase and planned the next phase for the construction of functional prototypes.

This paper presents our best practices for developing an HPS for applications in security robots with a focus on the prototyping phase, which follows the specification development phase. This phase is aimed at designing, producing, and testing functional prototypes, preferably in the form of the final product. The remainder of this paper is as follows. In Section 2, we build a HPS prototype and test it using a DC electronic load coupled with a charge-discharge system controller. In Section 3, we convert its structure to a plug-in hybrid with the aim of extending the operation hours. In Section 4, we integrate the converted HPS prototype into the security robot and test it on a dedicated chassis dynamometer. Following the bench tests, we finally test the integrated plug-in HPS prototype on the road to demonstrate its capability to continuously operate the security robot for 24 h. Finally, we conclude this paper in Section 5.

2. Unit Testing

2.1. Acquisition and Installation

The components for the HPS were acquired based on the developed specifications and installed to construct a prototype. Fuel cells in a stack with 1 kW nominal power were selected (Horizon Fuel Cell Technologies, H-1000W). The fuel cell stack is a series connection of 48 proton exchange membrane cells. From an open-circuit voltage of 45.5 V, the voltage range reaches 28.8 V, from which the peak power comes out with 35 A. Lithium-ion battery (LIB) cells with 2.5 Ah nominal discharge capacity and 3.6 V nominal voltage were selected (Samsung SDI, 18650-25R). To match the rated voltage of the load (24 V), seven cells were connected in series, producing a 7S configuration. To withstand the load demand higher than the power supply from the fuel cell, the series-connected cells were reconnected in parallel, producing a battery pack based on a 7S8P configuration. In response to the fuel cell power, a DC/DC converter with 960 W rated power was selected (Mean Well, SD-1000L-24). The converter features an input voltage range of 19 V to 72 V and an output voltage of 24 V that is manually adjustable from 23 V to 30 V. Despite manual adjustments with a control knob, fine-tunable output voltage of the converter is useful in controlling the power flow because this allows the conditions under which the fuel cell starts sharing the power demand with the battery to be adjusted. The procured components are listed in Table 1 with their brief specifications.

The battery and DC/DC converter required no installation because their parts are completely enclosed in a single unit. Contrary to them, the fuel cell needed to be installed because most of its parts, which include a controller, switches, connectors, valves, and tubes, were provided as external parts. Only a blower was built into the air-cooled fuel cell. To enclose these separate parts within a single unit, an all-in-one-box was developed in-house. In addition, the fuel cell enclosure helps withstand temperature extremes by integrating an air conditioner that consists of chillers and heaters.

2.2. Unit Test Bench

Rather than using large-scale and high-cost equipment specialized for batteries or fuel cells, our unit test bench consists mostly of general-purpose equipment. Static and dynamic loads were simulated using a DC electronic load (Kikusui, PLZ1004W) coupled with a DC power supply (Kikusui, PWR800L). The electronic load discharges a battery like in a situation where the security robot speeds up. The power supply charges a battery as in a situation where the security robot slows down, although regenerative braking might be insignificant in such a low-speed vehicle. Using the electronic load and power supply, a load transient from charge to discharge and vice versa is possible, but a significant latency period should be provided. To make the transition seamless, the electronic load and power supply were integrated with a charge-discharge system controller (Kikusui, PFX2512) that can guarantee a latency times less than 50 ms, which is required to emulate drive cycles for electric vehicles.

Our unit test bench also comprises multi-purpose and modular data acquisition systems (National Instruments, cRIO-9035), to monitor, save, and post-process test data. The charge-discharge system controller is also capable of measuring the voltage, current, and temperature over a few channels. The fuel cell can take such measurements internally, and the battery can do so. However, such individual measurements were not sufficient because system-level, rather than component-level, data acquisition was required. The use of the data acquisition systems allowed us to meet our requirements, which include multiple point measurements, exact time synchronization between multiple measurements, and the provision of an interface to external equipment such as a chassis dynamometer.

In the power network of the HPS, the voltages were measured at the output of the fuel cell (high-side) and converter (low-side). Because of the characteristics of the one-stage series structure, the low-side voltage is equipotential with the DC bus voltage and battery voltage. As the high and low-side voltages span from 21 V to 45 V, voltage input module with a 60 V input range (National Instruments, 9229) were used for both the high and low-sides. The currents were also measured at both the high and low-sides. At the low-side, the current was additionally measured at the output of the battery to trace the power flow. The high and low-side currents range up to 50 A, and thus the hall-effect current transducers (LEM, DHAB S/145) were used for both sides. The voltage produced by the current transducer was acquired using another voltage input module with a 5 V input range (National Instruments, 9239). This was also used to measure the voltage produced by a hydrogen pressure transducer (Keller, PA-23). The hydrogen pressure was used to estimate the hydrogen consumption and fuel cell efficiency. The temperatures were also measured. Because the temperature within the fuel cell was already given by its controller, the temperatures of air flowing in and out of the fuel cell were measured using a temperature input module (National Instruments, 9212). The apparatus in our unit test bench are listed in

Table 2. Apparatus in unit test bench.

Property	Function	Model	Installation Position in Power Network
Power	DC electronic load	Kikusui, PLZ1004W	Low-side
	DC power supply	Kikusui, PWR800L
	Charge-discharge system controller	Kikusui, PFX2512
Signal	Controller	NI, cRIO-9035
	Voltage measurement	NI, 9229	High-side Low-side
	Current measurement	LEM, DHAB S/145 NI, 9239	High-side Low-side (2 ea)
	H2 pressure measurement	Keller, PA-23 NI, 9239
	Air temperature measurement	NI, 9212	Inlet Outlet

.

2.3. Unit Test Results

The constructed HPS prototype was bench-tested under static and dynamic loads. The static test was aimed at discovering the maximum output power that the fuel cell can sustain for 24 h. This goal time was set in consideration of the main development objective of the security robot, that is, to achieve 24 h continuous operation on a single charge.

For prolonged and stable operation, the voltage, current, and temperature of the fuel cell must be maintained within the safe operating area (SOA). A typical SOA is defined mainly in terms of the under-voltage (UV), over-current (OC), and over-temperature (OT) conditions. In the selected fuel cell, an UV fault occurs if the average voltage per cell is less than 0.5 V, thus, 24 V per stack. An OC fault occurs if the current exceeds 30 A. An OT fault occurs if the internal temperature exceeds 65 °C or if the ambient temperature is higher than 45 °C. If one of the above fault conditions is met, the operating point of the fuel cell falls outside of the SOA, the controller will then immediately disconnect the load and sound an alarm. The alarm can be interpreted by counting the number of bleeps or by reading the text in the status display window incorporated into the controller. The controller will attempt to reconnect the load every ten seconds. Once the fault condition that caused the initial disconnection has passed, and the operating point of the fuel cell consequently returns to the SOA, the load can be connected again.

Using the electronic load, the load was increased from 500 W to 900 W in steps of 100 W. We observed that the power supply from the fuel cell increases with increasing temperature and definitely with lowering efficiency (see Figure 1a). A part of the fuel cell power is consumed by its auxiliaries, mostly, the fuel cell blower and controller. Therefore, the fuel cell must supply approximately 150 W more power than the given load. For instance, the fuel cell needs to deliver nearly 1050 W to meet the maximum load demand of 900 W (see Figure 1b). To maintain a constant output power, the heat generated by the fuel cell should be counterbalanced by the heat dissipated through its thermal management system. At the output power of 1050 W, which slightly exceeded the nominal power of the fuel cell, the fuel cell appeared to reach thermal equilibrium at 56 °C. However, the fuel cell abruptly shut down after only 2.24 h with an OT alarm and could not recover until the load demand was reduced (see Figure 1c). The operating points of the fuel cell at the moment it stopped were measured to be 25.1 V and 35.8 A by our data acquisition system and measured to be 23.9 V, 38.7 A, and 57 °C by the fuel cell controller. The temperature within the fuel cell, which triggered an OT alarm, could only be monitored internally. Therefore, according to the user’s manual, either an UV (23.9 V < 24 V) or OC (38.7 A >> 30 A) alarm should have gone off, whichever alarm condition occurred first. However, in reality, the fuel cell went down with an UT (57 °C << 65 °C) alarm, despite the temperature margin of 8 °C. Such discrepancies between the fault conditions were often observed throughout the test. Nonetheless, we found that the fuel cell could continuously meet the load demand up to 700 W, which required the power supply to be as high as 850 W. This safety margin is needed because of the uncertain conditions under which the fuel cell will be run, for example, when its performance decays over time or when the ambient conditions are hot and dry.

The dynamic test was aimed at identifying the transient behavior of the entire system and its components. Two situations that are likely to occur in the HPS are sudden changes in the power supply or the load demand. With regard to the first situation, variations in the power supply were produced by a short-circuit unit (SCU) that is exclusive for the selected fuel cell. As its name suggests, the SCU short-circuits the fuel cell for 100 ms every 10 s to condition the fuel cell and thus enhancing its durability. Although the SCU might be beneficial to the fuel cell, it can be damaging to the entire system because it causes the fuel cell voltage to drop every 10 s. Such concerns proved to be unnecessary because we found that despite variations in the fuel cell voltage (high-side), the DC bus voltage (low-side) was almost unaffected (see Figure 2a). This was mostly because the battery in the HPS readily made up for the power loss from the fuel cell when the SCU was activated. In addition to the regular short-circuiting by the SCU, irregular voltage drops in the fuel cell were observed because of the infrequent measurements at intervals of 1 s. With regard to the second situation, sudden changes in the power demand may occur more often for security robots operating in unstructured environments. Because of the limited knowledge on the actual operating conditions of the security robot, its operation mode was defined using a modified version of a well-known drive cycle, specifically, a charge-depleting cycle life test profile for an electric vehicle battery [23]. A drive cycle in the form of power versus time, which is usually called a power profile, is more appropriate for unit tests based on an electronic load than a drive cycle in the form of speed with respect to time. The modified drive cycle lasts for a total of 12 min, with the security robot surveilling and patrolling at the same time for the first half of the cycle, and only surveilling for the second half. During the first six minutes of simultaneous surveilling and patrolling, we observed a clear difference in the dynamic characteristics between the two power sources, which is in agreement with previous simulation results (see Figure 2b). As the load increased, the battery started supplying power to the load before the fuel cell to ensure the responsiveness of the entire system. As the load fell, the fuel cell supplied the remaining power to the battery and promptly recovered its state-of-charge (SOC).

3. Conversion to Plug-In Hybrid Structure

The goal time was set boldly along with the target distance of 70 km, but it was uncertain that what the security robot will do for 24 h and how much energy it will need. We started by assuming that the operation mode of the security robot is divided equally between surveillance and surveillance while patrolling. This assumption is reflected in the modified drive cycle, which is closer to a modal cycle. A modal cycle differs from a transient cycle in that it contains protracted periods of constant speed or power. Without a confirmation of the real-world representativeness of the modified drive cycle, we remained concerned that our goals will not be achieved. To address this, the HPS was converted into a so-called plug-in HPS with the aim of extending operating hours.

3.1. Retrofitting

The plug-in HPS is similar in nature to the powertrain in modern plug-in hybrid electric vehicles (PHEVs). More batteries are added, and they can be charged as much as necessary by plugging them into a wall outlet. Compared to hybrid electric vehicles, the ability to charge larger batteries in advance enables PHEVs to drive for a longer range purely on electric power. PHEVs usually operate in the charge-depleting (CD) mode at start-up, namely a battery-only mode or an all-electric mode. Then, they switch to the charge-sustaining (CS) mode, if the battery reaches the lower limit of its SOC, exhausting the vehicle’s all-electric range.

Likewise, the HPS prototype was retrofitted by upsizing the existing battery and adding an on-board charger (see Figure 3). The storable energy was increased by connecting more batteries in parallel, yielding a 7S12P configuration. The upper limit of the usable SOC range was broadened by plugging them into a power outlet through the charger. The lower limit of the usable SOC range was extended by resetting the output voltage of the existing DC/DC converter. In theory, the wider the usable SOC range is, the more extended hours of operation the security robot serves. Before testing the plug-in HPS prototype, a computer simulation was performed to predict the impact of these modifications on the operation hours.

3.2. Simulation Conditions

The HPS simulator that was built in the specification development phase [22] was modified to reflect an increase in battery size. The affected parameters in the battery block are the rated capacity (40 Ah) and the initial SOC (60~100%). As shown in

Table 3. Battery model parameters.

Name [Units]	HPS	Plug-in HPS	Remarks
Nominal voltage [V]	23.1		3.3 × 7S
Rated capacity [Ah]	20	40	2.5 × 12P
Initial state-of-charge [%]	50	60~100
Battery response time	30
Max. capacity [Ah]	20	40
Cut-off voltage [V]	17.5		2.5 × 7S
Fully charged voltage [V]	29.4		4.2 × 7S
Nominal discharge current [A]	20	40	1C
Internal resistance [ohm]	0.01
Capacity at nominal voltage [Ah]	16.8	36.2	2.1 × 8P
Exponential zone [V, Ah]	26.18, 5.84	26.18, 8.76	3.74 × 7S 0.73 × 12P

, the change in the rated capacity also led to changes in its subordinate parameters, such as the maximum capacity (40 Ah), nominal discharge current (40 A), and capacity at nominal voltage (36.2 Ah).

From the aspect of power sources, the plug-in HPS differs from the HPS on two factors. One is about larger batteries, and the other is about a wider SOC. With the simulator, we quantified the contribution of each of these aspects to an overall increase in operating hours. An increase in the battery size was mentioned earlier. Changes in the usable SOC range are as follows. With the HPS, the initial SOC was fixed at 50%; however, in the plug-in HPS, it can vary up to 100% by plug-in charging as much as necessary. Several upper limits of the SOC from 60% to 100% were therefore set in steps of 10%. The final SOC can also be changed by charge-depleting as deeply as possible. Accordingly, a few lower limits of the SOC were set from 40% to 10% in intervals of 10%. As in the unit test, a modified drive cycle was applied in the simulation.

3.3. Simulation Results

Under the most favorable conditions, we predicted that the plug-in HPS will run for 2.07 longer hours than the HPS. By discharging the larger battery (7S12P) over the widest SOC window (90%), the CD mode lasting for 2.07 h can be added before the transition to the CS mode. The CD mode was assumed to comprise surveillance during patrolling consuming 450 W. If it was replaced with only surveillance consuming 250 W, the CD mode can be extended to 3.73 h.

Nearly half of the total increase in operating hours are accounted for by the wider SOC. In the simulation, the plug-in HPS was allowed to discharge down to 10% starting from 100% SOC. This is in contrast to the HPS, which has almost no net discharge capacity. The gain from the wider SOC thus amounts to approximately 18 Ah (466.2 Wh). However, although a wider battery SOC range increases the operating time of the security robot, it is well known that high SOCs of over 90% are detrimental to the battery health. In addition, such high SOCs are not favored as the upper limit of SOC, considering the low quality of the components used to retrofit the HPS prototype. For instance, the charger was not customized to the battery. The charger voltage stops at approximately 28 V, which corresponds to a SOC of not higher than 81%. More importantly, only rudimentary monitoring capabilities are provided in the battery, in which the battery voltage can only be monitored across the entire pack and not across each series group of cells. Therefore, operation at high SOCs is prone to overvoltage faults, regardless of how well the SOC difference between the modules is balanced. The lower limit of the SOC is typically less than 20%, which is not low enough to affect the battery health. However, low SOCs are prone to operational instability. This is particularly serious in overload conditions in which a higher power demand is imposed on the plug-in HPS than can be delivered by the fuel cell alone. If such higher power demands occur frequently and/or last for a long time, the battery might need to bear the power demand without any time to recharge. The plug-in HPS might then fail to meet the voltage requirements as a result. In the simulation, a SOC of as low as 10% was favored as the lower SOC limit partly because of the relatively moderate power demand in the given drive cycle. However, the validity of this assumption in practice is questionable and needs to be confirmed through actual tests.

The remaining half of the increase in operating hours can be accounted for by the larger battery. By upsizing the battery configuration from 7S8P to 7S12P, the battery size was increased from 20 Ah (518 Wh) to 40 Ah (1036 Wh). Accordingly, the gain from the wider SOC was doubled to nearly 36 Ah (932.4 Wh). We concluded that the total energy gain of 624.6 Wh from these two contributions is the maximum gain that can be expected by converting to the plug-in hybrid structure. However, as mentioned above, much of the gain from the wider SOC will be reduced in practice by various practical constraints.

4. Vehicle Testing

4.1. Integration Test Bench

For unit testing, the electronic load was employed to emulate the loads. As previously assumed, the majority of the loads placed on the security robot were split in two, for surveillance and for patrolling. The load required for patrolling comes from the four electric motors on the security robot, whereas the load required for surveillance comes from the sensors and computers. In the plug-in HPS prototype integrated into the security robot, the load necessary for patrolling was directly provided by a chassis dynamometer through the tires and wheels of the security robot, instead of the electronic load. The security robot was placed on a rolling road, but not on an actual road crowded with objects to be perceived using the sensors and computers. As before, the load required for surveillance was substituted with the electronic load.

Chassis dynamometers are one of the widely used equipment for vehicle testing purposes. One or more fixed rollers are used to simulate different road loads within a controlled environment. However, to the best of our knowledge, there is no commercially available dynamometer that caters to our security robot. Dynamometers for automobiles are too lengthy to be installed in our security robot, in which the wheelbase is only 0.8 m long. Dynamometers for motorcycles are sufficiently short, but a four-wheel drive is required for our security robot. A dynamometer dedicated to our security robot was therefore developed in-house. Unlike conventional dynamometers, the set of two rollers per wheel is not fixed. The rollers can move up to 0.5 m back and forth, accommodating vehicles with different wheelbases.

Typical dynamometers are capable of auto-regulating dynamic loads that vary as a function of the vehicle speed, which is usually represented by three relevant parameters. Unfortunately, this capability is still under development, and hence a static load which can be set manually was exerted instead. However, we discovered that the settable range of loads was shifted upwards with respect to what was initially assumed, which means that the security robot was subjected to a much higher load demand while it was spinning its rollers. Despite these limitations in accurate load regulation, the integration tests under conditions similar to real-world environments were still worth performing, especially at the late prototyping phase.

4.2. Integration Test Results

The integrated plug-in HPS prototype was bench-tested on the dynamometer during development (see Figure 4). Owing to the lack of capabilities, the integration tests were mostly carried out at a typical speed (5 kph).

In accordance with the previous tests, the first test was aimed at verifying an increase in operating hours thanks to larger batteries and a wider SOC in the plug-in hybrid structure. In the simulation, the widest SOC window (90%) was assumed, but it is not feasible in reality. The upper limit of the SOC was fixed at 81% because of the limit imposed by the low quality of the charger in the plug-in HPS prototype. The lower limit of the SOC was first set at 15%, and it would be optimized in the following test. In consequence, the CD mode lasting for 1.27 h was added before the transition to the CS mode (see Figure 5a). Note that the battery bears the load alone during the CD mode that is followed by the CS mode in which the fuel cell shares the load with the battery. The added hours in the actual test is nearly 40% shorter than that in the simulation. The greater part of this shortage is attributable to the narrower SOC window (90% to 66%), but it cannot cover all. The remaining part is related to inaccurate and unstable load regulation in the dynamometer. While spinning the rollers at a constant speed (5 kph), the load demand abruptly dropped at 5.67 h. This was witnessed by a change in the power supply at low-side (see Figure 5b), which was reduced from 390 W to 240 W approximately. This can also be manifested by a variation in the rate of change in the hydrogen pressure (see Figure 5c). By accident, the reduced load (240 W) is close to the estimated power (200 W) required to move the security robot at a typical speed. Details on the estimation process can be found in our original paper [22]. In this circumstance, we observed that the integrated plug-in HPS prototype continuously drove the security robot on the dynamometer at 5 kph over 24 h.

The second test was aimed at determining the lower limit of the SOC. Note that by optimizing the usable SOC range, the operating hours of the security robot can be maximized while maintaining the operational stability of the plug-in HPS. The lower limits of the SOC were set at 15%, 46%, and 22% in sequence. Based on the characteristics of the one-stage series structure, these limits were realized by adjusting the output voltage of the converter to 23.8 V, 24.8 V, and 24.3 V, respectively. A lower final SOC makes the operating time of the security robot longer but also makes the plug-in HPS more vulnerable to sudden load changes. Hence, as performed in the first test, the final SOC was first set at 15%, and at the beginning, the vehicle speed was set to 5 kph, which is approximately equivalent to a motor speed of 1500 rpm. A sudden change in the load demand was imposed by stepping the motor speed up by 500 rpm. At 24.188 h, we observed that the load increased to nearly 860 W by revving up from 2500 rpm to 3000 rpm (see Figure 6a). At this moment, the fuel cell shut down with an OC alarm, and the battery bore the load alone. The fuel cell controller subsequently attempted to reconnect the load every ten seconds but failed because of the same fault. About 4.2 min later, the fuel cell recovered when the motor speed was revved down to 1500 rpm (see Figure 6b). The operating points of the fuel cell when it stopped were measured by our data acquisition system to be 28.5 V and 30.1 A (see Figure 6c,d). Unlike the previous unit test, an OC (30.1 A > 30 A) alarm went off, as specified in the user’s manual. However, this fault condition was questionable because if it was true, the maximum output power (1 kW) would not have been extracted. According to the data sheet, the maximum operating points were set as 35 V and 28.8 A. It is apparent that the fault alarm did not work as expected. This requires further investigation, in cooperation with the manufacturer if possible. The remaining two candidates of the lower SOC limit, 46% and 22%, were tested in the same manner. As a result, the final SOC that could maximize the operating hours of the security robot while uncompromising the operational stability of the plug-in HPS was 22%, which was therefore used as the lower limit of the SOC in the following integration tests.

As mentioned previously, the dynamometer under development is incapable of strict load regulation, as witnessed in the previous tests. The power required to move the security robot at a constant speed (5 kph) was measured to be about 400 W, which is almost twice as high as the estimate (200 W). The third test was therefore aimed at determining the amount of load exerted by our dynamometer. The integrated plug-in HPS prototype was road-tested with all other conditions remaining the same. We noticed that the power required to travel on the road was usually less than that required to spin the rollers (see Figure 7). The traveled ground was paved but not flat. This resulted in variations in the power supply, which had the average value of 190 W.

5. Conclusions

The process of prototyping an HPS for use in security robots was described. Based on the developed specifications for the HPS, the following components were acquired: fuel cells with 1 kW nominal power (Horizon Fuel Cell Technologies, H-1000W), batteries with 40 Ah nominal capacity (Samsung SDI, 18650-25R), a DC/DC converter with 960 W rated power (Mean Well, SD-1000L-24), and composite cylinders storing up to 0.71 kg of hydrogen compressed at 310 bar (Inocom, SCBA-9.0L). Unlike the other components, the fuel cell required installation as most of its parts were provided externally. Therefore, a fuel cell enclosure was developed to incorporate the separate parts, such as the controller, switches, connectors, valves, and tubes. Chillers and heaters were also installed in the enclosure to condition the air flowing into the fuel cell. An electronic load coupled to a charge-discharge system controller was used to emulate both the mechanical load required for patrolling and the electrical load required for surveillance. The HPS prototype was tested and the maximum output power that the fuel cell could sustain for 24 h was determined to be 850 W. To further increase the energy density of the HPS, the structure was converted to a plug-in hybrid featuring larger batteries and a wider SOC. Prior to testing the converted prototype, the impact of the plug-in hybrid structure was simulated. The results show that the operation time was extended by 2.07 h with a larger battery (7S12P) over the widest SOC window (90%). In consideration of the simulation results, the plug-in HPS prototype was integrated into the security robot. To emulate the mechanical load required for patrolling and apply it directly to the security robot through its tires and wheels, a chassis dynamometer is required. However, typical off-the-shelf dynamometers are not suitable because our security robot is based on a four-wheel drive and has a short wheelbase (0.8 m). Therefore, a chassis dynamometer dedicated to our security robot was developed. The integrated prototype was tested, and its capability to continuously operate the security robot for 24 h was demonstrated.