AI 与供应链自动化

AI 正从"辅助决策"走向"自主决策"——供应链是 AI 落地最成熟的领域之一。

AI 供应链应用全景

智能需求预测

"""
AI 需求预测（特征工程示例）
"""
from dataclasses import dataclass
@dataclass
class DemandFeatures:
"""需求预测特征"""
historical_sales: list[float]   # 历史销量
price: float
promotion: bool
holiday: bool
weather_temp: float
competitor_price: float
day_of_week: int
class AIForecaster:
"""AI 预测器（简化演示）"""
@staticmethod
def feature_engineering(f: DemandFeatures) -> dict:
"""特征工程"""
recent = f.historical_sales[-7:]
features = {
"7日均值": sum(recent) / len(recent),
"7日趋势": recent[-1] - recent[0],
"价格": f.price,
"促销": 1 if f.promotion else 0,
"节假日": 1 if f.holiday else 0,
"气温": f.weather_temp,
"价格竞争力": f.competitor_price / f.price,
"星期": f.day_of_week,
}
return features
@staticmethod
def predict(features: dict) -> dict:
"""预测（规则模拟）"""
base = features["7日均值"]
adjustment = 1.0
if features["促销"]:
adjustment *= 1.35
if features["节假日"]:
adjustment *= 1.25
if features["价格竞争力"] > 1.1:
adjustment *= 1.10
if features["7日趋势"] > 0:
adjustment *= 1.05
predicted = base * adjustment
return {
"预测销量": round(predicted),
"基准销量": round(base),
"调整系数": f"{adjustment:.2f}",
"置信区间": f"[{round(predicted*0.85)}, {round(predicted*1.15)}]",
}
# 演示
features = DemandFeatures(
historical_sales=[120, 135, 128, 142, 155, 148, 160],
price=99,
promotion=True,
holiday=False,
weather_temp=22,
competitor_price=109,
day_of_week=5,
)
forecaster = AIForecaster()
feat = forecaster.feature_engineering(features)
print("=== 特征工程 ===")
for k, v in feat.items():
print(f"  {k}: {v}")
pred = forecaster.predict(feat)
print("\n=== 预测结果 ===")
for k, v in pred.items():
print(f"  {k}: {v}")

自动补货系统

"""
自动补货引擎
"""
from dataclasses import dataclass
@dataclass
class AutoReplenishment:
"""自动补货"""
@staticmethod
def calculate(
current_stock: int,
daily_demand: float,
lead_time: int,
safety_days: int,
min_order: int,
) -> dict:
"""计算是否需要补货"""
safety_stock = daily_demand * safety_days
reorder_point = daily_demand * lead_time + safety_stock
days_of_stock = current_stock / daily_demand if daily_demand else 999
need_reorder = current_stock <= reorder_point
if need_reorder:
# 补到 (lead_time + safety_days + review_period) 天的量
target = daily_demand * (lead_time + safety_days + 7)
order_qty = max(target - current_stock, min_order)
else:
order_qty = 0
return {
"当前库存": current_stock,
"日均需求": round(daily_demand),
"可售天数": round(days_of_stock, 1),
"补货点": round(reorder_point),
"需要补货": "是" if need_reorder else "否",
"建议订量": round(order_qty),
}
# 演示
items = [
{"sku": "A001", "stock": 150, "demand": 30, "lt": 7, "safety": 3, "moq": 100},
{"sku": "A002", "stock": 500, "demand": 20, "lt": 14, "safety": 5, "moq": 200},
{"sku": "A003", "stock": 50, "demand": 15, "lt": 10, "safety": 3, "moq": 50},
]
engine = AutoReplenishment()
print("=== 自动补货检查 ===")
for item in items:
result = engine.calculate(
item["stock"], item["demand"], item["lt"],
item["safety"], item["moq"]
)
status = "🔴" if result["需要补货"] == "是" else "🟢"
print(f"\n  {status} {item['sku']}:")
for k, v in result.items():
print(f"    {k}: {v}")

AI 供应链技术成熟度

行动清单

• [ ] 选取 3 个高波动 SKU，用特征工程方式（加入促销/节假日/竞品价格）测试 AI 预测准确率提升
• [ ] 为自动补货系统设定参数（日均需求、前置期、安全天数、最小订货量），并跑一周试验
• [ ] 设置补货异常告警：自动补货触发量超过平时 3 倍时，转人工审核再确认
• [ ] 评估仓储机器人投资 ROI（见第 2 节详细计算），识别回收期 ≤ 3 年的优先场景
• [ ] 调研 AI 供应链 SaaS 工具（如 Lokad、Relex、o9），对比自建模型的成本
• [ ] 建立 AI 预测 vs 人工预测的双轨对比机制，用 MAPE 数据说话

下一节：02-仓储机器人与自动化 — 仓储自动化技术选型与 ROI 计算。