多阶段随机优化中非预期性约束的索引错误排查求助

问题场景 #

构建多阶段随机优化模型时，添加非预期性约束出现索引错误，错误提示：

KeyError: "Index '(1, 1, 1, 1, 1)' is not valid for indexed component 'Prod'"

模型集合定义 #

graph = {   1 : [2],
            2 : [3, 4],
            3 : [4, 5, 6],
            4 : [7, 8, 9],
            5 : [],
            6 : [],
            7 : [],
            8 : [],
            9 : [10],
            10 : [],
            11 : [12],
            12 : [4, 8]
        }

connections = [(s, t) for s in graph.keys() for t in graph[s]]


model.N = pyo.Set(initialize=graph.keys())     ## 所有节点集合

model.P = pyo.Set(within=model.N, initialize= [1, 11])     ## 生产节点集合

model.connection = pyo.Set(model.N, within=model.N, initialize=graph)     ## 节点关联集合，按节点索引

model.E = pyo.Set(within=model.N * model.N, initialize=connections)     ## 所有边（弧）集合

model.T = pyo.Set(initialize=range(1,8))     ## 时间周期集合

model.S = pyo.Set(initialize= [1, 2, 3, 4])

model.Prod = pyo.Var(model.P, model.T, model.S, within=NonNegativeReals)

非预期性约束场景集合Omega #

Omega = {
    1: {
        1: [1, 2, 3, 4],
        2: [1, 2],
        3: [1, 2],
        4: [1, 2],
        5: [1],
        6: [1],
        7: [1],
    },
    2: {
        1: [1, 2, 3, 4],
        2: [1, 2],
        3: [1, 2],
        4: [1, 2],
        5: [2],
        6: [2],
        7: [2],
    },
    3: {
        1: [1, 2, 3, 4],
        2: [3, 4],
        3: [3, 4],
        4: [3, 4],
        5: [3],
        6: [3],
        7: [3],
    },
    4: {
        1: [1, 2, 3, 4],
        2: [3, 4],
        3: [3, 4],
        4: [3, 4],
        5: [4],
        6: [4],
        7: [4],
    }
}

出错的约束代码 #

model.AA1 = ConstraintList()
for P in model.Prod:
    for T in model.T:
        for S in model.S:
            for Omega in Omega[S][T]:
                model.AA1.add(model.Prod[P, T, S] == model.Prod[P, T, Omega])
model.pprint()

对比可正常运行的相似逻辑代码 #

W = Set(initialize=range(1, 4))
S = Set(initialize=range(1, 5))
TH = Set(initialize=range(1, 3))

model_parcel.z = Var(W, TH, S, within=NonNegativeIntegers)

Omega = {
    1: {1: [1, 2, 3, 4], 2: [1, 2], 3: [1]},
    2: {1: [1, 2, 3, 4], 2: [1, 2], 3: [2]},
    3: {1: [1, 2, 3, 4], 2: [3, 4], 3: [3]},
    4: {1: [1, 2, 3, 4], 2: [3, 4], 3: [4]},
}

model_parcel.non_anticipativity_z = ConstraintList()
for w in W:
    for t in TH:
        for s in S:
            for omega in Omega[s][t]:
                model_parcel.non_anticipativity_z.add(model_parcel.z[w, t, s] == model_parcel.z[w, t, omega])

错误原因 #

核心错误在于遍历对象错误：

• model.Prod是一个三维决策变量（索引为model.P, model.T, model.S），直接遍历它时，每次拿到的P是一个三元组（比如(1,1,1)），而非单个生产节点ID。
• 后续嵌套循环中又加入T和S，导致访问model.Prod[P, T, S]时，实际传入的是((1,1,1), 1, 1)，也就是五维索引，完全超出了Prod的三维定义范围，因此触发索引错误。

而可正常运行的代码中，遍历的是单个维度的集合W，而非变量本身，所以不会出现索引维度溢出的问题。

另外，循环变量Omega与外部字典Omega重名，会导致后续无法访问原字典，虽然不是本次报错的直接原因，但属于代码隐患。

修复方案 #

将遍历model.Prod改为遍历生产节点集合model.P，同时修正循环变量名避免重名：

model.AA1 = ConstraintList()
for P in model.P:  # 遍历生产节点集合，而非三维变量
    for T in model.T:
        for S in model.S:
            for omega in Omega[S][T]:  # 改用小写omega避免与字典名重名
                model.AA1.add(model.Prod[P, T, S] == model.Prod[P, T, omega])
model.pprint()

内容的提问来源于stack exchange，提问作者Cvakapoor