No es el más elegante, lo admito, pero debería darte lo que quieres.
No estoy seguro de cómo quiere manejar esos valores NULL que tienen un promedio de hora de NULL. En el siguiente ejemplo, estos se actualizarán a -1.
create table myTable
(myDate datetime not null,
P_f decimal(10,5) default null
);
insert into myTable(myDate,P_f) values ('2001-01-01 20:20:00',1.88);
insert into myTable(myDate,P_f) values ('2001-01-01 20:25:00',NULL);
insert into myTable(myDate,P_f) values ('2001-01-01 20:30:00',NULL);
insert into myTable(myDate,P_f) values ('2001-01-01 20:35:00',1.71);
insert into myTable(myDate,P_f) values ('2001-01-01 20:40:00',NULL);
insert into myTable(myDate,P_f) values ('2001-01-01 20:45:00',NULL);
insert into myTable(myDate,P_f) values ('2001-01-01 20:50:00',NULL);
insert into myTable(myDate,P_f) values ('2001-01-01 20:55:00',1.835);
insert into myTable(myDate,P_f) values ('2001-01-01 21:00:00',1.918);
insert into myTable(myDate,P_f) values ('2001-01-01 21:05:00',1.968);
insert into myTable(myDate,P_f) values ('2001-01-01 21:10:00',2.004);
insert into myTable(myDate,P_f) values ('2001-01-01 21:15:00',1.924);
insert into myTable(myDate,P_f) values ('2001-01-01 21:20:00',1.8625);
insert into myTable(myDate,P_f) values ('2001-01-01 21:25:00',1.94);
insert into myTable(myDate,P_f) values ('2001-01-01 21:30:00',2.0375);
insert into myTable(myDate,P_f) values ('2001-01-01 21:35:00',1.912);
insert into myTable(myDate,P_f) values ('2001-01-02 20:40:00',NULL);
-- Insert copy of null value P_f rows into myTable with 1 hour average about myDate
insert into myTable
(myDate,P_f)
select t.myDate,ifnull((select avg(P_f) from myTable t1 where t1.myDate between t.myDate - interval 1 hour and t.myDate +interval 1 hour),-1) as hourAvg
from myTable t
where t.P_f is null;
-- delete rows where P_f is null
delete from myTable
where P_f is null;
