temperture input handling built

README.md

@@ -53,6 +53,7 @@ Analyzes sensor data and returns ML predictions with optional AI diagnosis.
 {
   "vdc1": [600.0, 601.0, 602.0],
   "idc1": [10.0, 10.1, 10.2],
+  "pvt": [25.0, 25.1, 25.2],
   "api_key": "your_google_api_key_here",
   "asset_id": "PV_INVERTER_001"
 }
@@ -60,6 +61,7 @@ Analyzes sensor data and returns ML predictions with optional AI diagnosis.
 
 **Parameters:**
 - `vdc1`, `idc1`: Voltage and current sensor readings
+- `pvt`: PV temperature readings (required)
 - `api_key`: Optional Google API key for AI diagnosis
 - `asset_id`: Optional asset identifier (auto-generated if not provided)
 
@@ -86,11 +88,12 @@ Analyzes sensor data and returns ML predictions with optional AI diagnosis.
 
 1. **Input Validation**: Ensures voltage/current arrays match and contain sufficient data points
 2. **Data Preparation**: Pads input to 100 data points for consistent processing
-3. **Feature Engineering**: Creates 7 statistical features using rolling window analysis:
+3. **Feature Engineering**: Creates 10 statistical features using rolling window analysis:
    - Voltage mean/standard deviation
    - Power mean/standard deviation
    - Power delta and slope
    - Normalized efficiency
+   - Temperature mean/standard deviation/delta
 4. **ML Inference**: Processes features through anomaly detection and prediction models
 5. **Agent Analysis**: LLM analyzes ML results for human-readable diagnosis (if API key provided)
 
@@ -116,6 +119,7 @@ curl -X POST "http://localhost:7860/analyze" \
   -d '{
     "vdc1": [600.0, 601.0, 602.0],
     "idc1": [10.0, 10.1, 10.2],
+    "pvt": [25.0, 25.1, 25.2],
     "api_key": "your_api_key",
     "asset_id": "PV_INVERTER_001"
   }'

app.py

@@ -14,7 +14,7 @@ service = AnalysisService(config)
 async def analyze_sensor_data(data: SensorData):
     try:
         logger.info(f"Processing request with {len(data.vdc1)} voltage and {len(data.idc1)} current data points")
-        ml_output, agent_output = service.analyze(data.vdc1, data.idc1, data.api_key, data.asset_id)
+        ml_output, agent_output = service.analyze(data.vdc1, data.idc1, data.pvt, data.api_key, data.asset_id)
         return AnalysisResponse(ml_output=ml_output, agent_output=agent_output)
     except HTTPException:
         raise

ml/artifacts/lstm_autoencoder.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:57c4a04b17e72241d655281137a398ea72c33680f2b0a2b84f72ad5fc32a91b4
-size 26232
+oid sha256:9a766a985bdb1649847bbe1a8025a324fa4361cd0599275ae3400afb13905ecd
+size 30216

ml/artifacts/ml_config.json

@@ -6,13 +6,16 @@
     "pdc_std",
     "pdc_delta",
     "pdc_slope",
-    "efficiency_norm"
+    "efficiency_norm",
+    "temp_mean",
+    "temp_std",
+    "temp_delta"
   ],
   "window": 50,
   "seq_len": 30,
   "downsample": 10,
   "failure_horizon_days": 30,
-  "lstm_input_dim": 7,
+  "lstm_input_dim": 10,
   "lstm_hidden_dim": 32,
   "design_life_days": 10958
 }

ml/artifacts/scaler.json

@@ -1,20 +1,26 @@
 {
   "mean": [
-    142.25967157616776,
-    3.5676508560940654,
-    635.9378633904394,
-    16.520853955427857,
-    1.1263708972285581e-07,
-    1.2714375420941854e-07,
-    1.0000000011260433
+    277.14991332828424,
+    3.8061906228008,
+    1362.8301711669567,
+    29.156264912725195,
+    -0.0001488146566343459,
+    8.639857281699144e-05,
+    1.0000000000001135,
+    34.75638390717355,
+    0.6580879751931045,
+    0.00028594267085960774
   ],
   "scale": [
-    135.83560758590238,
-    9.960201153946878,
-    867.4447952608547,
-    72.25659454825748,
-    22.465312616553394,
-    3.96862260102056,
-    8.867862393270387e-07
+    17.600415431393053,
+    4.996958923838216,
+    796.0751430809097,
+    68.8226394869352,
+    22.78849103980007,
+    3.9935425420114656,
+    9.124015105734193e-10,
+    12.189764634031704,
+    0.6327830276832512,
+    0.17153519012956447
   ]
 }

ml/features.py

@@ -14,4 +14,7 @@ def build_features(df, window):
     )
     df["efficiency"] = df["pdc1"] / (df["vdc1"] * df["idc1"] + 1e-6)
     df["efficiency_norm"] = df["efficiency"] / df["efficiency"].rolling(window).mean()
+    df["temp_mean"] = df["pvt"].rolling(window).mean()
+    df["temp_std"] = df["pvt"].rolling(window).std()
+    df["temp_delta"] = df["pvt"].diff()
     return df

ml/inference.py

@@ -36,6 +36,8 @@ class MLEngine:
         self.window = config["window"]
         self.seq_len = config["seq_len"]
         self.design_life_days = config["design_life_days"]
+        self.lstm_input_dim = config.get("lstm_input_dim", len(self.feature_cols))
+        self.lstm_hidden_dim = config.get("lstm_hidden_dim", 32)
 
     def _load_scaler(self):
         """Load and reconstruct StandardScaler from JSON."""
@@ -68,8 +70,8 @@ class MLEngine:
     def _load_lstm_model(self):
         """Load LSTM autoencoder from safetensors."""
         self.lstm = LSTMAutoencoder(
-            input_dim=len(self.feature_cols),
-            hidden_dim=32
+            input_dim=self.lstm_input_dim,
+            hidden_dim=self.lstm_hidden_dim
         )
         state_dict = load_file(os.path.join(ARTIFACTS_DIR, "lstm_autoencoder.safetensors"))
         self.lstm.load_state_dict(state_dict)

src/models.py

@@ -3,6 +3,7 @@ from pydantic import BaseModel
 class SensorData(BaseModel):
     vdc1: list[float]
     idc1: list[float]
+    pvt: list[float]
     api_key: str = None  # Optional Google API key for LLM features
     asset_id: str = None  # Optional asset identifier
 

src/services.py

@@ -12,11 +12,11 @@ class AnalysisService:
         self.config = config
         self.ml_engine = MLEngine()
 
-    def analyze(self, vdc1: list, idc1: list, api_key: str, asset_id: str = None) -> tuple:
+    def analyze(self, vdc1: list, idc1: list, pvt: list, api_key: str, asset_id: str = None) -> tuple:
         """Analyze sensor data and return ML and agent outputs."""
         logger.info(f"Complete analysis start - processing {len(vdc1)} data points")
-        validate_sensor_data(vdc1, idc1)
-        raw_df = prepare_dataframe(vdc1, idc1)
+        validate_sensor_data(vdc1, idc1, pvt)
+        raw_df = prepare_dataframe(vdc1, idc1, pvt)
         ml_output = self.ml_engine.predict_from_raw(raw_df, asset_id)
         agent_output = self.get_agent_output(api_key, ml_output)
         

src/utils.py

@@ -1,16 +1,20 @@
 from fastapi import HTTPException
 import pandas as pd
 
-def validate_sensor_data(vdc1: list, idc1: list) -> None:
+def validate_sensor_data(vdc1: list, idc1: list, pvt: list) -> None:
     """Validate sensor data consistency. Raises HTTPException on error."""
     if len(vdc1) != len(idc1):
         raise HTTPException(status_code=400, detail="Voltage and current lists must have the same length")
+    if len(vdc1) != len(pvt):
+        raise HTTPException(status_code=400, detail="Voltage, current, and temperature lists must have the same length")
     if len(vdc1) < 3:
         raise HTTPException(status_code=400, detail="Need at least 3 data points")
 
-def prepare_dataframe(vdc1: list, idc1: list) -> pd.DataFrame:
+def prepare_dataframe(vdc1: list, idc1: list, pvt: list) -> pd.DataFrame:
     """Prepare sensor data for ML inference by padding to 100 points."""
-    return pd.DataFrame({
+    df = pd.DataFrame({
         "vdc1": (vdc1 * (100 // len(vdc1) + 1))[:100],
-        "idc1": (idc1 * (100 // len(idc1) + 1))[:100]
-    })
+        "idc1": (idc1 * (100 // len(idc1) + 1))[:100],
+        "pvt": (pvt * (100 // len(pvt) + 1))[:100]
+    })
+    return df