This　article　is　the　second　part　of　a　two-part　study　of　the　modeling　and　optimizing　of　anode-supported　solid　oxide　fuel　cells　(SOFC)　with　gradient　anode.　Based　on　the　3D　numerical　model　established　and　validated　in　part　I　of　this　two-part　article,　two　kinds　of　continuous　gradient　anodes　at　the　micro-scale　level　(continuous　gradient　porosity　and　continuous　gradient　particle　size)　were　introduced　and　described　in　this　article.　Two　widely　used　neural　networks　(NNs),　i.e.　back　propagation　neural　network　(BPNN)　and　radial　basis　neural　network　(RBNN),　were　tested　to　find　the　proper　approximation　between　various　microstructure　parameters　distribution　and　cell　electrical　performance.　For　the　optimization　of　both　continuous　gradient　anodes,　BPNN　shows　a　better　approximation　of　the　objective　function　in　comparison　with　RBNN.　Based　on　the　BPNN,　the　genetic　algorithm　(GA)　was　used　to　find　the　optimal　microstructure　parameters　distribution　for　the　highest　electrical　performance　of　SOFC.　The　optimal　porosity　distribution　was　found　to　be　expressed　as　Ε(z)　=　(0.57–0.70)z29.73/73429.73　+　0.7　(0　µm　≤　z　≤　734　µm).　The　optimal　particle　diameter　distribution　was　expressed　as　dp(z)　=　(0.10–0.60)z65.12/73465.12　+　0.6　(0　µm　≤　z　≤　734　µm).　Compared　with　the　experimental　maximum　output　power　density　(Pmax)　of　0.31　W·cm−2　at　800　°C　and　125　mL·min−1　in　Part　I,　Pmax　of　the　optimal　porosity　and　particle　diameter　distribution　is　increased　by　17.10%　and　37.42%,　respectively.　Compared　with　the　optimal　porosity　distribution,　the　optimal　particle　size　distribution　is　more　advantageous　to　improve　the　cell　electrical　performance.　These　results　can　provide　a　guidance　of　microstructure　parameters　distribution　during　the　experimental　fabrication　for　cell　performance　enhancement.　©　2019,　©　2019　Taylor　&　Francis　Group,　LLC.