SARAHAI-ENERGY

1. Market Intelligence
Purpose: Provide a summarized view of recent market data, forecasts, and anomaly scores.

30-Day Average vs. Current Price (Bar + Line):
A combined bar/line chart shows the 30-day average prices (bar) and the latest current price (line) for each energy type and subtype. This quickly reveals which commodities are currently over- or under-priced versus their recent average.

Current Price Matrix:
A small interactive table showing each Energy Type + Subtype, along with the most recent price. This is your at-a-glance view of where today’s prices stand.

Recent Market Data (Last 50):
A time-series line chart that shows up to 50 of the most recent market data points. Different colored lines represent different (energy_type, subtype) pairs, such as “Power-Hydro,” “Coal-Anthracite,” “Gas-HenryHub,” etc.

Latest Forecast & Anomaly Scores (Last 20):
A combined plot overlaying forecast prices and anomaly scores. The left y-axis usually tracks forecasted price, while the right y-axis tracks anomaly scores. Peaks in anomaly score can indicate unusual market behavior.

System Status & Explanations:
A text box summarizing key statuses:

Whether a GPU is in use.

If multi-agent RL is active.

Any concurrency or HPC load notes.

Use Case: Quickly assess real-time market conditions, whether any anomalies were detected, and how the forecast is shaping up. If anomalies are persistently high, it might suggest further investigation or caution in trading strategies.

2. Futures Pricing Explorer
Purpose: Drill down into intraday vs. day-ahead futures pricing, along with anomaly listings and advanced line/bar charts.

Futures Pricing (~100 latest entries):
A time-series chart showing raw data for each commodity. This is valuable for short-term (intraday) vantage points.

Recent Anomalies (Last 10):
A scrolling list of anomaly events with timestamps, scores, and details (JSON). Quickly see which commodity or metric triggered the anomaly.

Intraday Forecast (Extended Line + Bars):
This chart merges:

Historical price data as a line.

Forecasted intraday data for the next 24 hours as stacked bars.

Each forecast bar often has a label with the predicted price. It’s a visual “continuation” of the historical line into the future.

Day-Ahead Forecast (Extended Line + Bars):
Similar concept for day-ahead, extending ~48 hours out. The system can overlay these longer-term forecasts on top of the historical line chart to illustrate potential future price movement.

Use Case: For short-term or next-day trading decisions. Easily see how the system projects future prices and check for potential anomalies or divergences.

3. Power Transactions
Purpose: Focus on cross-market intraday and day-ahead “simulated” trades to visualize potential profit & anomaly interaction.

Cross-Market Intraday Trades (PnL + Anomaly):
A line chart that accumulates the simulated “PnL” (profit or loss) from a naive cross-market strategy. Overlaid on the second y-axis is the anomaly score. This helps illustrate whether anomalous market conditions coincide with unexpectedly high or low trade performance.

Cross-Market Day-Ahead Trades (PnL + Anomaly):
Same concept but for “day-ahead” simulated trades, typically with a 48-hour horizon.

Generate ODS Trades Report:
A button at the bottom triggers a function that creates an .ods (OpenDocument Spreadsheet) file summarizing recent simulated trades. The file typically includes columns for buy_market, sell_market, buy_price, sell_price, quantity, profit, horizon, etc.

Use Case: Great for seeing how a simple cross-market approach might yield profits when one commodity is underpriced (buy) relative to another (sell). Monitoring anomaly spikes may guide you on whether to trust or hedge these trades.

4. GIS & Data Layers
Purpose: Manage external data keys (EIA, OWM) and view region-based weather/demand data.

OWM & EIA Key Inputs:
If you have valid OpenWeatherMap or EIA API keys, you can enter them here, then click the “Set” button. This allows SAIEv11 to fetch real data instead of using random or fallback simulated values.

Demand vs. Capacity (By Region):
A dropdown lets you select an ISO/RTO region (CAISO, ERCOT, MISO, PJM, etc.). The corresponding chart shows the last 50 data points for demand and capacity.

WeatherData & Demand Adjustment:
A table listing the last 10 entries of WeatherData (latitude, longitude, event, severity). Below it, a line chart illustrates how weather might adjust demand from a base level to an adjusted level.

North America Weather Map:
A map visualization (Scattergeo) plotting lat/long points of the last recorded weather events or severity markers. Good for spotting weather anomalies that might affect grid demand.

Use Case: Perfect for a grid operator or an energy trader who needs to see how real-time weather conditions might influence localized demand/capacity. Combine weather-based insights with the overall price forecast to refine your trades or generation dispatch.

5. Financial & Benchmarking
Purpose: Aggregate simulated trade performance, cumulative PnL, and generate an ODS report summarizing system analytics.

Simulated Trades (Last 20):
A line chart of buy/sell prices for the last 20 simulated trades, letting you see how close or far each trade’s “buy vs. sell” spread is.

Simulated Trading PnL (Cumulative):
A running sum of all profits from the simulated cross-market trades. This can highlight both short-term streaks and longer-term performance.

Profit by Energy Type (All Trades):
A bar chart grouping total profits allocated to each energy type. This helps spot which markets have historically provided the best alpha.

Generate ODS Report:
Creates a comprehensive .ods (OpenDocument Spreadsheet) with:

Forecast & Actual data

Anomalies

Shape strategy info (if available)

Intraday & Day-Ahead forecasts

Trading strategy table with recommended BUY/SELL signals

Use Case: This tab is the “financial view,” showing how trades performed historically and letting you quickly generate a shareable offline report. Perfect for management review or deeper analysis outside the web app.

6. Predictive AI
Purpose: Dive deeper into anomaly scores, HPC GPU usage, and the behind-the-scenes AI metrics.

Intraday & Next-Day Futures Pricing Predictions + PoL Anomaly Scores:
A time-series chart focusing on the latest 50 MarketData points, highlighting the PoL anomaly detection scores (price vs. load). The higher the anomaly score, the more unusual the data point is relative to the learned pattern-of-life.

Ongoing HPC GPU Simulation & Monitoring:
If ENABLE_GPU_STRESS is turned on in your environment variables, a background thread performs matrix multiplications on the GPU to test concurrency. Meanwhile, this chart displays GPU usage over time—Load(%) and memory usage.

Latest GPU Usage (Bar):
A bar chart showing the latest snapshot of GPU usage:

GPU Load(%)

MemUsed(MB)

MemTotal(MB)

Use Case: For advanced users, HPC or AI engineers who want to confirm the system’s concurrency and GPU load are within expected ranges. They can see if anomaly detection or RL training is saturating the GPU and adjust system resources accordingly.

Additional Notes
Threads & Background Services

SAIEv11 spawns multiple background threads to fetch real-time data from ISOs, weather APIs, and EIA.

A “unified data collector” periodically sends data to /api/add-data to keep the LSTM forecasting and PoL anomaly detection live.

Database

By default, SAIEv11 uses a local SQLite database named saiev11.db.

This file is created or updated in the same folder as SAIE.exe unless you override it via environment variables.

Environment Variables

Several environment variables can customize intervals, GPU usage, database file name, and more (e.g., DATA_COLLECTOR_INTERVAL, EIA_FETCH_INTERVAL, WEATHER_FETCH_INTERVAL, etc.).

Typically, you’ll only need to set these if you want to change default behaviors (like disabling GPU stress or pointing to a different DB file).

Generated Reports

Clicking “Generate ODS Report” or “Generate ODS Trades Report” will produce .ods files in your working directory. You can open these with tools like LibreOffice, OpenOffice, or Microsoft Excel (with an ODF plugin).

GPU vs. CPU Fallback

If SAIEv11 detects no CUDA-enabled GPU, it logs a warning and uses CPU fallback for all PyTorch operations.

This allows you to run it on any Windows 11 machine, though performance may be slower for training and HPC tasks.