如何在CPLEX Studio的OPL中实现WFP优先级约束？

我正在处理世界粮食计划署（WFP）的优化问题：一组分属不同区域的国家，每个国家对应R、L2或L3（严重程度递增）的紧急等级。已完成大部分约束建模，现需实现以下核心约束：仅当所有紧邻的更高紧急等级国家的粮食需求至少满足60%时，才能向更低紧急等级国家分发食品包。需要在OPL语言中实现该约束以在CPLEX Studio中求解，现有代码如下：

{string} Countries = ...;
{string} Regions = ...;
int Emergency_levels[Regions][Countries] = ...;
int Needs[Regions][Countries] = ...;
int MAX_BASKETS = ...;
int Container_Cost_Transport[Regions] = ...;
int Container_Capacity = ...;
float LG[Regions][Countries] = ...;
float LA[Regions][Countries] = ...;
float DG[Regions][Countries] = ...;
float DA[Regions][Countries] = ...;
float CG[Regions][Countries] = ...; 
float CA[Regions][Countries] = ...;
int MAX_BUDGET = ...;
float Threshold = ...;

// Decision variables
dvar int Region_Containers[Regions];//number of containers to send to region
dvar int GX[Regions][Countries]; // number of baskets transported by ground from region i to country j
dvar int AX[Regions][Countries]; // number of baskets transported by air from region i to country j
dvar int Threshold_Respected[]

maximize sum(i in Regions, j in Countries) ((GX[i][j]*(1-LG[i][j])*DG[i][j]) + (AX[i][j]*(1-LA[i][j])*DA[i][j]));

subject to{
  //Check that total number of baskets is respected
  sum(i in Regions, j in Countries) (GX[i][j] + AX[i][j]) <= MAX_BASKETS;
  //cant send more baskets than the needs
  forall(i in Regions, j in Countries)
    GX[i][j] + AX[i][j] <= Needs[i][j];
  
  //Ensures full container loads
  forall(i in Regions)
    sum(j in Countries) (GX[i][j] + AX[i][j]) == Region_Containers[i]*150;
  //Ensure budget is respect
  (   sum(i in Regions) Container_Cost_Transport[i]*Region_Containers[i] + sum(i in Regions,j in Countries) (GX[i][j]*CG[i][j] + AX[i][j]*CA[i][j])   )  <= MAX_BUDGET;
  // TODO: Constraint for emergency level threshold
}

实现步骤说明 #

1. 补充邻接关系定义：需明确国家间的“紧邻”业务/地理关系，添加集合存储每个国家的邻国列表。
2. 紧急等级数值化：将R、L2、L3转换为可比较的数值（如R=1，L2=2，L3=3），方便判断等级高低。
3. 线性化约束逻辑：用大M法处理“仅当...才...”的逻辑，避免除法运算，将需求满足率要求转换为线性不等式。

修改后的完整OPL代码 #

{string} Countries = ...;
{string} Regions = ...;
// 补充：定义每个国家的紧邻国家列表，需根据实际业务填充
{string} AdjacentCountries[Countries] = ...;

// 紧急等级数值化：R=1，L2=2，L3=3（严重程度递增）
int Emergency_levels[Regions][Countries] = ...;
int Needs[Regions][Countries] = ...;
int MAX_BASKETS = ...;
int Container_Cost_Transport[Regions] = ...;
int Container_Capacity = ...;
float LG[Regions][Countries] = ...;
float LA[Regions][Countries] = ...;
float DG[Regions][Countries] = ...;
float DA[Regions][Countries] = ...;
float CG[Regions][Countries] = ...; 
float CA[Regions][Countries] = ...;
int MAX_BUDGET = ...;
// 定义需求满足阈值：60%
float Threshold = 0.6;
// 大M常数：取单个国家最大需求值，确保约束逻辑生效
int M = max(i in Regions, j in Countries) Needs[i][j];

// Decision variables
dvar int Region_Containers[Regions];//number of containers to send to region
dvar int GX[Regions][Countries]; // number of baskets transported by ground from region i to country j
dvar int AX[Regions][Countries]; // number of baskets transported by air from region i to country j
// 0-1变量：标记是否向国家j分发了食品包，简化约束逻辑
dvar boolean IsDistributed[Regions][Countries];

maximize sum(i in Regions, j in Countries) ((GX[i][j]*(1-LG[i][j])*DG[i][j]) + (AX[i][j]*(1-LA[i][j])*DA[i][j]));

subject to{
  //Check that total number of baskets is respected
  sum(i in Regions, j in Countries) (GX[i][j] + AX[i][j]) <= MAX_BASKETS;
  //cant send more baskets than the needs
  forall(i in Regions, j in Countries)
    GX[i][j] + AX[i][j] <= Needs[i][j];
  
  //Ensures full container loads
  forall(i in Regions)
    sum(j in Countries) (GX[i][j] + AX[i][j]) == Region_Containers[i]*150;
  //Ensure budget is respected
  (sum(i in Regions) Container_Cost_Transport[i]*Region_Containers[i] + 
   sum(i in Regions,j in Countries) (GX[i][j]*CG[i][j] + AX[i][j]*CA[i][j])) <= MAX_BUDGET;

  // 关联分发数量与IsDistributed变量
  forall(i in Regions, j in Countries) {
    // 如果分发了食品包，IsDistributed必须为1
    GX[i][j] + AX[i][j] <= M * IsDistributed[i][j];
    // 如果IsDistributed为1，至少分发1个食品包（适配整数变量特性）
    GX[i][j] + AX[i][j] >= 1 * IsDistributed[i][j];
  }

  // 核心约束：仅当所有紧邻更高等级国家需求满足≥60%时，才能向低等级国家分发
  forall(i in Regions, j in Countries) {
    forall(k in AdjacentCountries[j]) {
      // 仅当j的紧急等级低于k，且k有需求时应用约束
      if (Emergency_levels[i][j] < Emergency_levels[i][k] && Needs[i][k] > 0) {
        // 若向j分发，则k的需求必须满足至少60%；否则约束自动失效
        GX[i][k] + AX[i][k] >= Threshold * Needs[i][k] - M * (1 - IsDistributed[i][j]);
      }
    }
  }
}

关键约束解释 #

• 邻接关系：AdjacentCountries需根据实际地理或业务规则填充，例如AdjacentCountries["Kenya"] = {"Uganda", "Tanzania"}。
• 大M法逻辑：当IsDistributed[i][j]=1（向j分发）时，右侧惩罚项为0，强制GX[i][k]+AX[i][k]≥0.6*Needs[i][k]；当IsDistributed[i][j]=0时，约束自动放松，不限制对k的分发情况。
• 等级数值化：确保Emergency_levels的数值与紧急程度正相关，通过Emergency_levels[i][j] < Emergency_levels[i][k]快速判断j为更低等级国家。

内容的提问来源于stack exchange，提问作者rui