pub trait SortD{ fn sort_d(&mut self); }
impl<T: Ord> SortD for Vec<T>{ fn sort_d(&mut self) {
    self.sort_by(|u, v| v.cmp(&u));
} }
pub trait Mx{fn max(&self, rhs: Self)->Self;}
impl Mx for f64{ fn max(&self, rhs: Self)->Self{if *self < rhs{ rhs } else { *self } }}
pub trait Mi{ fn min(&self, rhs: Self)->Self; }
impl Mi for f64{ fn min(&self, rhs: Self)->Self{ if *self < rhs{ rhs } else { *self } } }
pub fn chmax<T: PartialOrd+Clone>(a: &mut T, b: &T){ if *a < *b{ *a = b.clone(); } }
pub fn chmin<T: PartialOrd+Clone>(a: &mut T, b: &T){ if *a > *b{ *a = b.clone(); } }
pub fn gcd(mut a: i64, mut b: i64)->i64{ if b==0{return a;}(a,b)=(a.abs(),b.abs());while b!=0{ let c = a;a = b;b = c%b; }a }
pub fn factorial_i64(n: usize)->(Vec<i64>, Vec<i64>){ let mut res = vec![1; n+1];let mut inv = vec![1; n+1];for i in 0..n{ res[i+1] = (res[i]*(i+1)as i64)%MOD; }inv[n] = mod_inverse(res[n], MOD);for i in (0..n).rev(){ inv[i] = inv[i+1]*(i+1) as i64%MOD; }(res, inv) }
pub fn floor(a:i64, b:i64)->i64{let res=(a%b+b)%b;(a-res)/b}
pub fn extended_gcd(a:i64,b:i64)->(i64,i64,i64)
{if b==0{(a,1,0)}else{let(g,x,y)=extended_gcd(b,a%b);(g,y,x-floor(a,b)*y)}}
pub fn mod_inverse(a:i64,m:i64)->i64{let(_,x,_) =extended_gcd(a,m);(x%m+m)%m}
pub fn comb(a: i64, b: i64, f: &Vec<(i64, i64)>)->i64{
    if a<b{return 0;}else if b==0 || a==b{ return 1; }
    else{let x=f[a as usize].0;
        let y=f[(a-b) as usize].1;let z=f[b as usize].1;return((x*y)%MOD)*z%MOD;}}
pub fn factorial(x: i64)->Vec<(i64, i64)>{
    let mut f=vec![(1i64,1i64),(1, 1)];let mut z = 1i64;
    let mut inv = vec![0; x as usize+10];inv[1] = 1;
    for i in 2..x+1{z=(z*i)%MOD;
        let w=(MOD-inv[(MOD%i)as usize]*(MOD/i)%MOD)%MOD;
        inv[i as usize] = w;
        f.push((z, (f[i as usize-1].1*w)%MOD));}return f;}
pub fn fast_mod_pow(x: i64,p: usize, m: i64)->i64{
    let mut res=1;let mut t=x;let mut z=p;while z > 0{
        if z%2==1{res = (res*t)%m;}t = (t*t)%m;z /= 2; }res}
#[allow(unused_imports)]
use std::{
    convert::{Infallible, TryFrom, TryInto as _}, fmt::{self, Debug, Display, Formatter,},
    fs::{File}, hash::{Hash, Hasher}, iter::{Product, Sum}, marker::PhantomData,
    ops::{Add, AddAssign, Sub, SubAssign, Div, DivAssign, Mul, MulAssign, Neg, RangeBounds},
    str::FromStr, sync::{atomic::{self, AtomicU32, AtomicU64}, Once},
    collections::{*, btree_set::Range, btree_map::Range as BTreeRange}, mem::{swap},
    cmp::{self, Reverse, Ordering, Eq, PartialEq, PartialOrd},
    thread::LocalKey, f64::consts::PI, time::Instant, cell::RefCell,
    io::{self, stdin, Read, read_to_string, BufWriter, BufReader, stdout, Write},
};
#[allow(unused_imports)]
use proconio::{input, input_interactive, marker::{*}};
#[allow(unused_imports)]
//use rand::{thread_rng, Rng, seq::SliceRandom, prelude::*};
#[allow(unused_imports)]
//use itertools::Itertools;
#[allow(unused_imports)]
//use ordered_float::OrderedFloat;
#[allow(unused_imports)]
//use ac_library::{*, ModInt998244353 as mint};
#[allow(dead_code)]
//type MI = StaticModInt<Mod998244353>;pub fn factorial_mint(n: usize)->(Vec<MI>, Vec<MI>){ let mut res = vec![mint::new(1); n+1];let mut inv = vec![mint::new(1); n+1];for i in 0..n{res[i+1] = res[i]*(i+1);}inv[n] = mint::new(1)/res[n];for i in (0..n).rev(){inv[i] = inv[i+1]*(i+1);}(res, inv)}

#[allow(dead_code)]
const INF: i64 = 1<<60;
#[allow(dead_code)]
const MOD: i64 = 998244353;
#[allow(dead_code)]
const D: [(usize, usize); 4] = [(1, 0), (0, 1), (!0, 0), (0, !0)];
#[allow(dead_code)]
const D2: [(usize, usize); 8] = [(1, 0), (1, 1), (0, 1), (!0, 1), (!0, 0), (!0, !0), (0, !0), (1, !0)];

#[derive(Default)]
pub(crate) struct SimpleQueue<T> {
    payload: Vec<T>,
    pos: usize,
}

impl<T> SimpleQueue<T> {
    pub(crate) fn reserve(&mut self, n: usize) {
        if n > self.payload.len() {
            self.payload.reserve(n - self.payload.len());
        }
    }

    pub(crate) fn size(&self) -> usize {
        self.payload.len() - self.pos
    }

    pub(crate) fn empty(&self) -> bool {
        self.pos == self.payload.len()
    }

    pub(crate) fn push(&mut self, t: T) {
        self.payload.push(t);
    }

    // Do we need mutable version?
    pub(crate) fn front(&self) -> Option<&T> {
        if self.pos < self.payload.len() {
            Some(&self.payload[self.pos])
        } else {
            None
        }
    }

    pub(crate) fn clear(&mut self) {
        self.payload.clear();
        self.pos = 0;
    }

    pub(crate) fn pop(&mut self) -> Option<&T> {
        if self.pos < self.payload.len() {
            self.pos += 1;
            Some(&self.payload[self.pos - 1])
        } else {
            None
        }
    }
}

use std::{
    ops::{
        BitAnd, BitAndAssign, BitOr, BitOrAssign, BitXor, BitXorAssign,
         Not, Rem, RemAssign, Shl, ShlAssign, Shr, ShrAssign,
    },
};

// Skipped:
//
// - `is_signed_int_t<T>`   (probably won't be used directly in `modint.rs`)
// - `is_unsigned_int_t<T>` (probably won't be used directly in `modint.rs`)
// - `to_unsigned_t<T>`     (not used in `fenwicktree.rs`)

/// Corresponds to `std::is_integral` in C++.
// We will remove unnecessary bounds later.
//
// Maybe we should rename this to `PrimitiveInteger` or something, as it probably won't be used in the
// same way as the original ACL.
pub trait Integral:
'static
+ Send
+ Sync
+ Copy
+ Ord
+ Not<Output = Self>
+ Add<Output = Self>
+ Sub<Output = Self>
+ Mul<Output = Self>
+ Div<Output = Self>
+ Rem<Output = Self>
+ AddAssign
+ SubAssign
+ MulAssign
+ DivAssign
+ RemAssign
+ Sum
+ Product
+ BitOr<Output = Self>
+ BitAnd<Output = Self>
+ BitXor<Output = Self>
+ BitOrAssign
+ BitAndAssign
+ BitXorAssign
+ Shl<Output = Self>
+ Shr<Output = Self>
+ ShlAssign
+ ShrAssign
+ fmt::Display
+ fmt::Debug
+ fmt::Binary
+ fmt::Octal
+ Zero
+ One
+ BoundedBelow
+ BoundedAbove
{
}

/// Class that has additive identity element
pub trait Zero {
    /// The additive identity element
    fn zero() -> Self;
}

/// Class that has multiplicative identity element
pub trait One {
    /// The multiplicative identity element
    fn one() -> Self;
}

pub trait BoundedBelow {
    fn min_value() -> Self;
}

pub trait BoundedAbove {
    fn max_value() -> Self;
}

macro_rules! impl_integral {
    ($($ty:ty),*) => {
        $(
            impl Zero for $ty {
                #[inline]
                fn zero() -> Self {
                    0
                }
            }

            impl One for $ty {
                #[inline]
                fn one() -> Self {
                    1
                }
            }

            impl BoundedBelow for $ty {
                #[inline]
                fn min_value() -> Self {
                    Self::MIN
                }
            }

            impl BoundedAbove for $ty {
                #[inline]
                fn max_value() -> Self {
                    Self::MAX
                }
            }

            impl Integral for $ty {}
        )*
    };
}

impl_integral!(i8, i16, i32, i64, i128, isize, u8, u16, u32, u64, u128, usize);
use std::cmp::min;
use std::iter;

impl<Cap> MfGraph<Cap>
where
    Cap: Integral,
{
    pub fn new(n: usize) -> MfGraph<Cap> {
        MfGraph {
            _n: n,
            pos: Vec::new(),
            g: iter::repeat_with(Vec::new).take(n).collect(),
        }
    }

    pub fn add_edge(&mut self, from: usize, to: usize, cap: Cap) -> usize {
        assert!(from < self._n);
        assert!(to < self._n);
        assert!(Cap::zero() <= cap);
        let m = self.pos.len();
        self.pos.push((from, self.g[from].len()));
        let rev = self.g[to].len() + usize::from(from == to);
        self.g[from].push(_Edge { to, rev, cap });
        let rev = self.g[from].len() - 1;
        self.g[to].push(_Edge {
            to: from,
            rev,
            cap: Cap::zero(),
        });
        m
    }
}

#[derive(Clone, Debug, PartialEq, Eq)]
pub struct Edge<Cap: Integral> {
    pub from: usize,
    pub to: usize,
    pub cap: Cap,
    pub flow: Cap,
}

impl<Cap> MfGraph<Cap>
where
    Cap: Integral,
{
    pub fn get_edge(&self, i: usize) -> Edge<Cap> {
        let m = self.pos.len();
        assert!(i < m);
        let _e = &self.g[self.pos[i].0][self.pos[i].1];
        let _re = &self.g[_e.to][_e.rev];
        Edge {
            from: self.pos[i].0,
            to: _e.to,
            cap: _e.cap + _re.cap,
            flow: _re.cap,
        }
    }
    pub fn edges(&self) -> Vec<Edge<Cap>> {
        let m = self.pos.len();
        (0..m).map(|i| self.get_edge(i)).collect()
    }
    pub fn change_edge(&mut self, i: usize, new_cap: Cap, new_flow: Cap) {
        let m = self.pos.len();
        assert!(i < m);
        assert!(Cap::zero() <= new_flow && new_flow <= new_cap);
        let (to, rev) = {
            let _e = &mut self.g[self.pos[i].0][self.pos[i].1];
            _e.cap = new_cap - new_flow;
            (_e.to, _e.rev)
        };
        let _re = &mut self.g[to][rev];
        _re.cap = new_flow;
    }

    /// `s != t` must hold, otherwise it panics.
    pub fn flow(&mut self, s: usize, t: usize) -> Cap {
        self.flow_with_capacity(s, t, Cap::max_value())
    }
    /// # Parameters
    /// * `s != t` must hold, otherwise it panics.
    /// * `flow_limit >= 0`
    pub fn flow_with_capacity(&mut self, s: usize, t: usize, flow_limit: Cap) -> Cap {
        let n_ = self._n;
        assert!(s < n_);
        assert!(t < n_);
        // By the definition of max flow in appendix.html, this function should return 0
        // when the same vertices are provided.  On the other hand, it is reasonable to
        // return infinity-like value too, which is what the original implementation
        // (and this implementation without the following assertion) does.
        // Since either return value is confusing, we'd rather deny the parameters
        // of the two same vertices.
        // For more details, see https://github.com/rust-lang-ja/ac-library-rs/pull/24#discussion_r485343451
        // and https://github.com/atcoder/ac-library/issues/5 .
        assert_ne!(s, t);
        // Additional constraint
        assert!(Cap::zero() <= flow_limit);

        let mut calc = FlowCalculator {
            graph: self,
            s,
            t,
            flow_limit,
            level: vec![0; n_],
            iter: vec![0; n_],
            que: SimpleQueue::default(),
        };

        let mut flow = Cap::zero();
        while flow < flow_limit {
            calc.bfs();
            if calc.level[t] == -1 {
                break;
            }
            calc.iter.iter_mut().for_each(|e| *e = 0);
            while flow < flow_limit {
                let f = calc.dfs(t, flow_limit - flow);
                if f == Cap::zero() {
                    break;
                }
                flow += f;
            }
        }
        flow
    }

    pub fn min_cut(&self, s: usize) -> Vec<bool> {
        let mut visited = vec![false; self._n];
        let mut que = SimpleQueue::default();
        que.push(s);
        while let Some(&p) = que.pop() {
            visited[p] = true;
            for e in &self.g[p] {
                if e.cap != Cap::zero() && !visited[e.to] {
                    visited[e.to] = true;
                    que.push(e.to);
                }
            }
        }
        visited
    }
}

struct FlowCalculator<'a, Cap> {
    graph: &'a mut MfGraph<Cap>,
    s: usize,
    t: usize,
    flow_limit: Cap,
    level: Vec<i32>,
    iter: Vec<usize>,
    que: SimpleQueue<usize>,
}

impl<Cap> FlowCalculator<'_, Cap>
where
    Cap: Integral,
{
    fn bfs(&mut self) {
        self.level.iter_mut().for_each(|e| *e = -1);
        self.level[self.s] = 0;
        self.que.clear();
        self.que.push(self.s);
        while let Some(&v) = self.que.pop() {
            for e in &self.graph.g[v] {
                if e.cap == Cap::zero() || self.level[e.to] >= 0 {
                    continue;
                }
                self.level[e.to] = self.level[v] + 1;
                if e.to == self.t {
                    return;
                }
                self.que.push(e.to);
            }
        }
    }
    fn dfs(&mut self, v: usize, up: Cap) -> Cap {
        if v == self.s {
            return up;
        }
        let mut res = Cap::zero();
        let level_v = self.level[v];
        for i in self.iter[v]..self.graph.g[v].len() {
            self.iter[v] = i;
            let &_Edge {
                to: e_to,
                rev: e_rev,
                ..
            } = &self.graph.g[v][i];
            if level_v <= self.level[e_to] || self.graph.g[e_to][e_rev].cap == Cap::zero() {
                continue;
            }
            let d = self.dfs(e_to, min(up - res, self.graph.g[e_to][e_rev].cap));
            if d <= Cap::zero() {
                continue;
            }
            self.graph.g[v][i].cap += d;
            self.graph.g[e_to][e_rev].cap -= d;
            res += d;
            if res == up {
                return res;
            }
        }
        self.iter[v] = self.graph.g[v].len();
        res
    }
}

#[derive(Clone, Debug, Default)]
pub struct MfGraph<Cap> {
    _n: usize,
    pos: Vec<(usize, usize)>,
    g: Vec<Vec<_Edge<Cap>>>,
}

#[derive(Clone, Debug)]
struct _Edge<Cap> {
    to: usize,
    rev: usize,
    cap: Cap,
}

const B: i64 = 1<<40;

//use proconio::fastout;
//#[fastout]
fn main() {
    input!{
        n: usize,
        p: [i64; n],
        m: usize,
        e: [(Usize1, Usize1); m],
        k: usize,
        s: [(Usize1, Usize1, i64); k],
    }
    let mut mf = MfGraph::new(n+k+2);
    let st = n+k;
    let end = n+k+1;
    let mut ans = 0;
    for (i, &v) in p.iter().enumerate(){
        if v >= 0{
            ans += v;
            mf.add_edge(st, i, v);
            mf.add_edge(i, end, 0);
        } else {
            mf.add_edge(st, i, 0);
            mf.add_edge(i, end, -v);
        }
    }
    for &(u, v) in &e{
        mf.add_edge(v, u, B);
    }
    for (idx, &(u, v, add)) in s.iter().enumerate(){
        let p = idx+n;
        mf.add_edge(st, p, add);
        mf.add_edge(p, u, B);
        mf.add_edge(p, v, B);
        ans += add;
    }
    println!("{}", ans-mf.flow(st, end));
}