Autovacuum running out of memory

Whenever I turn on Autovacuum on my database, I get a ton of error
messages like this in my Postgres log:

Oct 16 06:43:47 [2897]: [1-1]  user=,db= ERROR:  out of memory
Oct 16 06:43:47 [2897]: [1-2]  user=,db= DETAIL:  Failed on request
of size 524287998.

It always fails on the same request. When I turn off autovacuum, they
go away. However, when I run VACUUM FULL manually, I don't get this
error.

My server has 2gb of ram, and my postgres settings are:

autovacuum = on                         # enable autovacuum subprocess?
                                         # 'on' requires
stats_start_collector
                                         # and stats_row_level to
also be on
#autovacuum_naptime = 1min              # time between autovacuum runs
#autovacuum_vacuum_threshold = 500      # min # of tuple updates before
                                         # vacuum
#autovacuum_analyze_threshold = 250     # min # of tuple updates before
                                         # analyze
#autovacuum_vacuum_scale_factor = 0.2   # fraction of rel size before
                                         # vacuum
#autovacuum_analyze_scale_factor = 0.1  # fraction of rel size before
                                         # analyze
#autovacuum_freeze_max_age = 200000000  # maximum XID age before
forced vacuum
                                         # (change requires restart)
#autovacuum_vacuum_cost_delay = -1      # default vacuum cost delay for
                                         # autovacuum, -1 means use
                                         # vacuum_cost_delay
autovacuum_vacuum_cost_limit = -1       # default vacuum cost limit for
                                         # autovacuum, -1 means use
                                         # vacuum_cost_limit

shared_buffers = 20000          # min 128kB or max_connections*16kB
                                         # (change requires restart)
#temp_buffers = 8MB                     # min 800kB
#max_prepared_transactions = 5          # can be 0 or more
                                         # (change requires restart)
# Note: increasing max_prepared_transactions costs ~600 bytes of
shared memory
# per transaction slot, plus lock space (see max_locks_per_transaction).
work_mem = 4096                         # min 64kB
maintenance_work_mem = 500MB            # min 1MB
#max_stack_depth = 2MB                  # min 100kB


Any ideas as to what might be going on?

Thanks
Jason