Sticking to monoalphabetic ciphers, Vigenere can be combined with a secret random-like substitution of the plaintext or/and ciphertext alphabet, making it significantly more resistant.
If Vigenere encryption is
$$x_j\mapsto (x_j+K[j\bmod k])\bmod{26}$$
where array $K$ is the key or length $k$, I'm discussing
$$x_j\mapsto (S[x_j]+K[j\bmod k])\bmod{26}$$
$$x_j\mapsto S'[(x_j+K[j\bmod k])\bmod{26}]$$
$$x_j\mapsto S'[(S[x_j]+K[j\bmod k])\bmod{26}]$$
where arrays $S$ or/and $S'$ are secret 26-letter permutations, which extends the key. Using permutations is necessary so that decryption is possible (this is not required for $K$).
Another option is to use a (possibly public) substitution $S$, and two (or more) rounds of Vigenere cipher with different keys $K$ and $K'$ (preferably: of coprime length):
$$x_j\mapsto (S[(x_j+K[j\bmod k])\bmod{26}]+K'[j\bmod k'])\bmod{26}$$
None of this is secure; in particular, any variation on this line suffers that with the same key, identical letters at the same position in the ciphertext occurs iff there are identical letters at the corresponding position in the plaintext, which is a serious weakness.