Energy Data Exploration

Use Zorora’s data analysis capabilities to explore production solar energy time series data with sandboxed pandas and matplotlib — directly from the REPL.

Scenario

You have a year of 30-minute resolution inverter data from a Namibian solar site, exported via Solarman. You need to:

• Assess data quality (gaps, nulls, resolution consistency)
• Identify generation patterns (daily profiles, seasonal trends)
• Compare inverter performance (INV 001 vs INV 002)
• Prepare insights for integration with the Ona Intelligence Layer

Step-by-Step Guide

Step 1: Load the Dataset

Use the /load command to ingest and profile a CSV file:

zorora
[1] ⚙ > /load docs/demo-data.csv

Zorora automatically profiles the dataset on load:

📊 Dataset Profile
──────────────────────────────────
Rows:        17,569
Columns:     3
Time range:  2023-01-01 → 2024-01-01 (366 days)
Resolution:  30 min
Gaps:        0

Column Summary:
  Timestamp              datetime64    0 nulls
  INV 001 (W)            float64       12 nulls
  INV 002 (W)            float64       8 nulls

Format detected: ODS-E compatible (Solarman export)

Step 2: Explore Structure

Inspect the DataFrame loaded into the session:

[2] ⚙ > /analyze df.info()
[3] ⚙ > /analyze result = df.head(10)

Check column names and dtypes to confirm the profile output. The Timestamp column is automatically parsed as datetime64 during loading.

Step 3: Data Quality Assessment

Check for nulls and gaps in the time series:

[4] ⚙ > /analyze result = df.isnull().sum()

INV 001 (W)    12
INV 002 (W)     8

Detect gaps in the expected 30-minute resolution:

[5] ⚙ > /analyze result = df['Timestamp'].diff().value_counts().head()

A healthy dataset shows the dominant interval as 0 days 00:30:00. Any other intervals indicate missing records or irregular exports.

Step 4: Statistical Analysis

Get descriptive statistics for the power columns:

[6] ⚙ > /analyze result = df.describe()

Compute daily generation totals:

[7] ⚙ > /analyze result = df.set_index('Timestamp').resample('D').sum().head(10)

Monthly aggregation for seasonal comparison:

[8] ⚙ > /analyze result = df.set_index('Timestamp').resample('M').sum()

Step 5: Visualization

Plot the full time series for both inverters:

[9] ⚙ > /analyze plt.figure(figsize=(14,5)); plt.plot(df['Timestamp'], df['INV 001 (W)'], label='INV 001', alpha=0.7); plt.plot(df['Timestamp'], df['INV 002 (W)'], label='INV 002', alpha=0.7); plt.legend(); plt.title('Inverter Power Output — Full Year'); plt.xlabel('Date'); plt.ylabel('Power (W)'); plt.savefig('__zorora_plot__.png')

Generate a daily average profile (hour-of-day pattern):

[10] ⚙ > /analyze df['hour'] = df['Timestamp'].dt.hour; result = df.groupby('hour')[['INV 001 (W)', 'INV 002 (W)']].mean()

Step 6: Comparative Analysis

Compare cumulative generation between the two inverters:

[11] ⚙ > /analyze inv1_total = df['INV 001 (W)'].sum() * 0.5; inv2_total = df['INV 002 (W)'].sum() * 0.5; result = f"INV 001: {inv1_total/1e6:.2f} MWh, INV 002: {inv2_total/1e6:.2f} MWh, Ratio: {inv1_total/inv2_total:.3f}"

Compute correlation between inverters (should be very high for co-located units):

[12] ⚙ > /analyze result = df['INV 001 (W)'].corr(df['INV 002 (W)'])

Example Output

Profile Summary (from /load)

📊 Dataset Profile
──────────────────────────────────
Rows:        17,569
Columns:     3
Time range:  2023-01-01 → 2024-01-01 (366 days)
Resolution:  30 min
Format:      ODS-E compatible

Analysis Result (from /analyze)

{
  "type": "scalar",
  "result": "0.9987",
  "plot_generated": false
}

Best Practices

Timestamp Handling

• Zorora auto-parses Timestamp columns to datetime64 on load
• Always verify with /analyze df.dtypes before time-based operations
• Use .set_index('Timestamp') for resampling workflows

Null-Aware Analysis

• Solar inverter data commonly has nulls during nighttime or communication drops
• Use .dropna() or .fillna(0) deliberately — don’t silently ignore
• Check null counts per column before aggregation

Resolution Validation

• The profile output shows detected resolution (e.g., 30 min)
• Verify with /analyze df['Timestamp'].diff().value_counts() to catch irregular intervals
• Gaps matter for energy calculations (kWh = kW × hours)

Working with Large Files

• The demo dataset (17,569 rows) loads in under a second
• For larger files (100k+ rows), prefer column-specific operations over full DataFrame scans
• Use .resample() to reduce data volume before plotting

Next Steps

See Also

• Getting Started — Installation and setup
• Guides: Slash Commands — Full /load and /analyze reference
• Technical Concepts: Architecture — How the data analysis sandbox works
• Academic Research — Deep research use case