You're tasked with designing and implementing an automated trading system based on Reinforcement Learning (RL). The goal is to manage a portfolio of 20-30 stocks in a highly liquid equity market, but one with significant microstructure frictions such as transaction costs, market impact, and order book depth limitations. This system needs to make intraday, high-frequency trading decisions, aiming to maximize the portfolio's risk-adjusted returns (e.g., Sharpe Ratio or Sortino Ratio), while strictly controlling Maximum Drawdown (MDD) and minimizing market impact. Please detail your system design, including the core RL components (state, action, reward), model selection, training methodology, how you'd address market microstructure challenges, your risk management strategies, and most importantly – how you would rigorously backtest and perform production validation to ensure its robustness and safety before deployment.

您被要求设计并实现一个基于强化学习（Reinforcement Learning, RL）的自动化交易系统，目标是在一个高流动性但具有显著微观结构摩擦（如交易成本、市场冲击、订单簿深度限制）的股票市场中，管理一个包含20-30只股票的投资组合。该系统需在日内高频层面做出交易决策，旨在最大化投资组合的风险调整收益（例如，夏普比率或Sortino比率），同时严格控制最大回撤（Maximum Drawdown, MDD）和实现最低的市场冲击。请详细阐述您的系统设计方案，包括RL的核心组件（状态、动作、奖励）、模型选择、训练方法、如何处理市场微观结构、风险管理策略以及最重要的——您将如何进行严格的回测和生产验证，以确保其鲁棒性和部署前的安全性。