bug似乎解决了

simulation_model.py

@@ -848,6 +848,8 @@ class SimulationModel(Model):
             )
 
     def _finalize_factory_errors(self, write_files: bool):
+        # Always reset cached factory error table so callers don't see stale data.
+        self.factory_error_df = None
         if not self.production_log:
             self.error = float("inf")
             return
@@ -889,17 +891,31 @@ class SimulationModel(Model):
             use_max=self.use_error_max,
         )
         cum_ratio_by_factory = stats["factory_cum_ratio"]
+        metric_fn = (lambda pct: pct.abs().max()) if self.use_error_max else (lambda pct: pct.abs().mean())
+        factory_metric = {fname: float(metric_fn(pct)) for fname, pct in cum_ratio_by_factory.items()}
+
         # Aggregate factory-level errors (mean across factories of max/mean per-month absolute ratio)
-        factory_errors = []
-        for ratio_series in cum_ratio_by_factory.values():
-            metric_val = ratio_series.abs().max() if self.use_error_max else ratio_series.abs().mean()
-            factory_errors.append(float(metric_val))
+        factory_errors = list(factory_metric.values())
         if not factory_errors:
             self.error = float("inf")
         else:
             # Aggregate across factories: max when use_error_max=True, otherwise mean.
             self.error = max(factory_errors) if self.use_error_max else sum(factory_errors) / len(factory_errors)
 
+        # Expose per-factory error (unitless ratio) even when files are not written (used by GA).
+        factory_df = pd.DataFrame(
+            {
+                "name": list(factory_metric.keys()),
+                "error_ratio": list(factory_metric.values()),
+            }
+        )
+        if not factory_df.empty:
+            factory_df = factory_df.sort_values(by="error_ratio")
+            factory_df["error_pct"] = factory_df["error_ratio"] * 100
+        else:
+            factory_df["error_pct"] = []
+        self.factory_error_df = factory_df.reset_index(drop=True)
+
         if not write_files:
             return
 
@@ -910,12 +926,12 @@ class SimulationModel(Model):
         factory_pivot.to_csv(factory_report_path, index=False, encoding="utf-8-sig")
 
         # Error summary based on cumulative deviation ratio (unitless, consistent with GA)
-        err_rows = []
         metric_col = "最大累计偏差" if self.use_error_max else "平均累计偏差"
-        for fname, ratio_series in cum_ratio_by_factory.items():
-            metric_val = ratio_series.abs().max() if self.use_error_max else ratio_series.abs().mean()
-            err_rows.append({"工厂名称": fname, metric_col: float(metric_val)})
-        error_df = pd.DataFrame(err_rows) if err_rows else pd.DataFrame(columns=["工厂名称", metric_col])
+        error_df = (
+            pd.DataFrame([{ "工厂名称": name, metric_col: val} for name, val in factory_metric.items()])
+            if factory_metric
+            else pd.DataFrame(columns=["工厂名称", metric_col])
+        )
         error_path = os.path.join(output_dir, f"factory_error_{timestamp}.csv")
         error_df.to_csv(error_path, index=False, encoding="utf-8-sig")
 
@@ -938,22 +954,11 @@ class SimulationModel(Model):
             plt.savefig(month_plot_path)
             plt.close()
 
-            factory_df = pd.DataFrame(
-                {
-                    "name": list(cum_ratio_by_factory.keys()),
-                    "error_ratio": [
-                        (pct.abs().max() if self.use_error_max else pct.abs().mean()) for pct in cum_ratio_by_factory.values()
-                    ],
-                }
-            )
-            factory_df = factory_df.sort_values(by="error_ratio")
-            factory_df["error_pct"] = factory_df["error_ratio"] * 100
-            self.factory_error_df = factory_df.reset_index(drop=True)
             plt.figure(figsize=(12, 6))
             ax = factory_df["error_pct"].reset_index(drop=True).plot(kind="bar")
             metric_label = "最大" if self.use_error_max else "平均"
-            plt.ylabel(f"{metric_label}????[%]")
-            plt.title(f"????{metric_label}?????????")
+            plt.ylabel(f"{metric_label}误差[%]")
+            plt.title(f"工厂{metric_label}累计偏差分布")
             plt.tight_layout()
             for idx, row in factory_df.reset_index(drop=True).iterrows():
                 val = row["error_pct"]