The building model used in the case study is 10 m width, 40 m length and 40 m height (=10 stories) [50] . It was assumed that 40 households lived in the building described in the reference [51] . The BIPV system assumed to be installed on the top of the building consists of solar panels and a FC using H₂ produced by water electrolysis with surplus power of PV.

The power generated by PV system is calculated by using the following equation [52] :

where E_PV is hourly electric power of PV system (kWh), H is amount of solar radiation (kWh/m²), K is power generation loss factor (−), P is system capacity of PV (kW), 1 is solar radiation under standard state (AM1.5, solar radiation: 1 kWh/m², module temperature: 25 degree Celsius) (kW/m²).

In this study, the high performance PV P250a Plus produced by Panasonic whose module conversion efficiency and maximum power per module are 19.5% and 250 W [53] , respectively is adopted for PV system. The size of PV module is 1580 mm × 812 mm × 35 mm. P is calculated by installing this PV module on a roof of the building model, which is 75 kW_p (=300 solar panels).K is calculated by using the following equation:

where K_p is power conversion efficiency of power conditioner (−), K_m is correction factor decided by module temperature (−), K_i is power generation loss by interconnecting and dirty of module surface (−). In this study, K_p and K_i are set at 0.945 and 0.95, respectively. K_p is assumed by referring to the performance of commercial power conditioning device VBPC259B3 manufactured by Panasonic [54] . K_m is calculated by the following equation:

where T_m is PV module temperature (degree Celsius), T_s is temperature under standard test condition (=25 degree Celsius), C is temperature correction factor which is 0.35 [55] (%/degree Celsius). T_m is calculated by using the following equation [56] :

where T_a is ambient air temperature (degree Celsius), U_m is wind velocity over module of PV (m/s).

The meteorological data, such as solar radiation, the ambient air temperature, and wind velocity of some cities in Japan are from the data base METPV-11 during the period from 1999 to 2009 [57] and the irradiation data of the project “PV300” during the period from August, 2012 to July, 2014 [58] . An example of meteorological data of METPV-11 for Tsu in Japan is shown in Figure 1. In this figure, a red circle, a green triangle and a flag symbol indicate the amount of horizontal solar radiation, the air temperature and a wind speed including direction, respectively. The solar radiation data for several tilt angles can be obtained from METPV-11 by changing configuration. Table 1 also lists an example data of PV300, which shows the data of air temperature and amount of horizontal solar radiation at 10 sec intervals on 1^st August, 2013 for Tsu in Japan.

It is important to consider the impact of shadow on power generation performance when PV system is installed. The longest shadow length which is obtained at AM 9:00 on the winter solstice is calculated by the following equation [52] :

where L is the longest shadow length of solar panel to north direction (mm), α is solar altitude (degree), β is solar azimuth angle from north and south direction (degree), h is height of solar panel (mm).

In other words, the solar panels need to keep L distance apart in N-S direction to avoid the shadowing each other. The number of solar panels, which can be installed on the roof of building without shadowing other panels, was estimated with various tilt angles.

In this case study, it is assumed that the surplus power generated by PV system over the electricity demand of households [49] living in the building model would be used for water electrolysis. The Type-S electrolyzer manufactured by IHT [59] whose H₂ production rate, power consumption and electrolysis efficiency are 760 Nm³/h, 4.45 kWh/Nm³ and 79.5%, is used in this case study. The amount of H₂ could be produced by the surplus power generated from PV system is calculated by the following equation:

whereV_H2 is amount of H₂ produced (Nm³), E_s is surplus power generated by PV system (kWh), P_e is power consumption (kWh/Nm³), η_e is electrolysis efficiency (−).

It is assumed that the H₂ produced by the electolyzer would be used to generate power through a polymer electrolyte fuel cell (PEFC) system. H₂ is converted into electricity by FC following the below equation:

where η_f is power generation efficiency of latest PEFC stationary system based on lower heating value (=0.39) [60] , Q is lower heating value of H₂ (=242) (kJ/mol).

In this study, 4 monthly values of the self-sufficiency rate of the proposed combination system consisting of PV and FC was investigated, which are representative of four seasons for Tsu city in Japan. The self-sufficiency rate is defined as the power supplied (from the combined PV and FC system to electricity demand of the building), in this study. The hourly time change in the self-sufficiency rate in the day when the daily mean amount of horizontal solar radiation per month was obtained was estimated.

At first, the optimum tilt angle for installment of solar panel on the roof is investigated using the hourly meteorological data base [57] for several cities in Japan. After that, the instant time change in power generation characteristics of PV system under the optimum tilt angle (a_t) condition is investigated using the meteorological data base of PV300 including air temperature, solar intensity and wind speed at 10 sec intervals [58] . Compared the power generation characteristics of PV system with the electricity demand data of households in Japan [49] , the surplus and deficit between power supply (from PV) and the electricity demand in the building were estimated. The power generation characteristics of FC system using H₂ produced by water electrolysis with the surplus power of PV system is investigated to bridge the gap between the power supply and the electricity demand.

To clarify the optimum a_t of solar panel installed on the roof of the building model universally in Japan, this study investigates the daily power energy of PV system averaged per month for five cities in Japan such as Tsu (north latitude: 34 degrees; east longitude: 136 degrees), Fukushima (north latitude: 37 degrees; east longitude: 140 degrees), Tokyo (north latitude: 35 degrees; east longitude: 139 degrees), Sapporo (north latitude: 43 degrees; east longitude: 141 degrees) and Naha (north latitude: 26 degrees; east longitude: 127 degrees) using the hourly meteorological data METPV-11 including air temperature, solar intensity and wind speedat 1 hour interval [57] . Tsu is located in the center of Japan. Fukushima is the city which is necessary to be rebuilt overcoming the big earthquake damage. Tokyo is the capital in Japan. Sapporo and Naha are located in the farthest north area and the farthest south area in Japan.

Figures 2-5 show the relationship between daily power energy of PV system averaged per month and a_t of solar panel installed on the roof of the building model. In these figures, January, April, July and October are selected as the months which are representative of four seasons in Japan, respectively. In addition, Figure 6 shows the relationship between annual power energy of PV system and a_t of solar panel installed on the roof of the building model for five cities. From these figures, it is revealed that the optimum a_t is 0 degree irrespective of city as well as season. Though the optimum a_t of solar panel installed on the ground is approximately 20 - 30 degrees in Japan generally [57] , the optimum a_t of solar panel installed on the roof of the building is 0 degree due to the limitation of the rood area of the building.

Table 2 lists the number of installable solar panel on the roof of the building model with consideration of shadow impact. In this table, the relationship between a_t and number of installable solar panel is shown. N_s and N mean the number of installable solar panel with consideration of shadow and that without consideration of shadow, respectively, when N is 300. From this table, N_s decreases with increase in a_t. Since the roof area of the building is limited, N_s decreases with increase in a_t due to shadow extension. According to the review by this study, there is no report concluding that 0 degree is the optimum a_t for installing solar panel on the roof of the building. Since the effect of shadow and restricted area of roof of building on the power generation performance of PV system is not considered at the same time in the past studies, it is believed that this new finding is obtained. It is thought that the dust impact on the power generation of PV system is afraid when a solar panel is installed at a_t of 0 degree for the roof of the building actually. However, this study suggests cleaning up the dust by water for the case of a_t of 0 degree.

Figures 7-10 show hourly power generated from proposed BIPV system and energy demand of 40 households assumed living in the building model in Tsu city in Japan. Hourly power generated by PV system was estimated for the standard day in the month with the mean amount of horizontal solar radiation of the month using the meteorological data base of PV300 [58] . In these figures, January, April, July and October are selected as the months which are representative of four seasons, respectively. The hourly self-sufficiency rate of PV system (R_PV) is also shown in order to reveal the power supply surplus or deficit in a particular hour.

From Figures 7-10, it is obvious that the generation of PV system increases in the morning up to the noon and decreases after noon to the evening due to solar trajectory, while the electricity demand keeps almost constant during the day and increases from the evening until the midnight. Due to the mismatch between the power generation and electricity demand, R_PV is over 100% during the daytime and is 0% during the night-time. Moreover, R_PV during the daytime in April and July are higher than those in January and October. Since the amount of horizontal solar radiation in spring and summer are higher than those in autumn and winter in Japan, the R_PV are higher in April and July than those in January and October.

It was assumed that the surplus electricity generated was used for water electrolysis to produce H₂ which is fed into the FC system. The self-sufficiency rate of the FC system (R_FC) is defined as the ratio of the power generated by FC system to the electricity demand which is not covered by PV system. The average daily R_FC in representative months is calculated and it is given in Table 2. In this table, E_FC is the power energy of FC system using H₂ generated by water electrolysis with the surplus power of PV system, and E_D is the electricity demand which is not covered by PV system. In addition, E_FC, E_D and R_FC were calculated as a daily value by summing up the hourly values for the day when the mean amount of horizontal solar radiation was obtained per month in January, April, July and October.

From Table 3, it is known that R_FC is in the range of 15% - 38%. Since the power energy is high in January and the power energy is low in October, R_FC for these two months are lower. However, it is implied that the power generation of FC system using H₂ generated by the surplus power of PV system is effective for bridging the gap between the power supply of PV system and electricity demand of households.

Table 4 shows the change in R_FC in a week which from three days ago to three days after the day when the mean amount of horizontal solar radiation was obtained per month in Tsu city using the meteorological data base of PV300 [58] . It is seen from Table 3 that R_FC changes following the daily meteorological condition. For example, in a cloudy or rainy day when R_FC is 0%, indicating that the surplus power of PV was not obtained and R_FC can be as high as 49% in a good sunny day.

As we know that the electricity demand/consumption depends on the number of households in the building, Table 5 and Table 6 give the self-sufficiency rates in a week calculated by the meteorological data base of PV300 [58] and the electricity demand data of households in Japan [49] under the assumptions that there were 20 and 12 households in the building with the same roof areas (i.e.

same amount of solar panels installed), respectively.

According to Table 5, it is seen that the R_FC increases compared to Table 4, but still falls below 100% for each day investigated.

Table 6 shows R_FC in a week in the case of 12 households who live in the building with 12 m height (=3 stories), when the mean amount of horizontal solar radiation was obtained per month in Tsu city using the meteorological data base of PV300 [58] . It can be seen, in this case, the R_FC over 100% are obtained in some days in April and high R_FC near 100% are obtained in some days in July and October. The surplus H₂ in the days when R_FC is over 100% can be stored and used for the days when R_FC is less than 100%. Although Tables 3-6 show the assessment result of one city (Tsu) only, it is expected that the proposed combined PV and FC system has potential to achieve the 100% self sufficient supply for buildings in Japan, especially in good solar resource areas.

In addition, the proposed systems in different buildings with different energy demands may be connected in future, so the surplus electricity generated from PV and/or H₂ can be transferred between buildings. Moreover, H₂ as an energy storage medium has longer term storage efficiency than secondary battery.