Describe the workload
Identify the input dataset, forecast target and horizon.
Explore configuration examples for the eomer API. Draft examples remain labelled while the interface develops.
Request demoworkflow:
source: SAMPLE_DATASET
forecast:
horizon: 7 days
review: required
output: SAMPLE_DESTINATIONIllustration only · not executableIdentify the input dataset, forecast target and horizon.
Compare illustrative code with the equivalent configuration in the Situation Room.
Agree the current endpoints and deployment requirements before production use.
eomer API
For data scientists who want to customise inputs, test assumptions, and integrate results into their own tools. Explore four Python notebooks, each with ten applications and three sample datasets.
Open API documentationPython / Energy
Based on the supplied eomer quickstart. Select inputs to update the code and sample data, then update the preview. All data and results are synthetic; this page does not execute Python or call the API. The examples do not indicate live market coverage.
Import the forecast and backtest helpers from the quickstart package. In your own notebook, read the API key from an environment variable and keep it outside shared code.
import os
import pandas as pd
from eomer_api import predict_df, backtest
from eomer_plots import plot_forecast
API_URL = os.environ["EOMER_API_URL"]
HEADERS = {"Authorization": f'Bearer {os.environ["EOMER_API_KEY"]}'}Select the country, application, sample dataset, and horizon above. Table 1 shows the last three rows of the synthetic history; target represents price in EUR/MWh. In your own notebook, supply a CSV with these columns.
# Germany / Day-ahead prices
# Sample: Winter demand (synthetic)
CONTEXT_PATH = "energy_day-ahead_sample_1.csv"
PREDICTION_LENGTH = 96
context_df = pd.read_csv(
CONTEXT_PATH,
parse_dates=["timestamp"],
)
context_df.tail(3)| item_id | timestamp (UTC) · 15 min | target (EUR/MWh) |
|---|---|---|
| germany_day-ahead | 2026-01-11 15:15 | 57.24 |
| germany_day-ahead | 2026-01-11 15:30 | 54.81 |
| germany_day-ahead | 2026-01-11 15:45 | 52.22 |
Pass the history and horizon to predict_df. The median describes the central forecast; P10 and P90 describe the requested uncertainty interval. The button below updates the synthetic preview for your selected inputs.
pred_df = predict_df(
context_df,
api_url=API_URL,
headers=HEADERS,
prediction_length=PREDICTION_LENGTH,
quantile_levels=[0.1, 0.5, 0.9],
freq="h",
)
pred_df.head()Figure 1 shows the synthetic median and interval for Germany, Day-ahead prices, over 96 steps. Use the interval to review uncertainty alongside the operating constraints.
plot_forecast(
context_df,
pred_df,
item_id=context_df["item_id"].iloc[0],
y_label="EUR/MWh",
lead_in_steps=3 * 96,
).show()Hover to inspect values. Drag to zoom; use the toolbar to pan, reset the axes, or export a PNG. Select a legend entry to show or hide a series.
| Timestamp (UTC) | P10 | P50 | P90 |
|---|---|---|---|
| 2026-01-11 16:00 | 38.52 | 52.02 | 65.52 |
| 2026-01-11 16:15 | 40.15 | 53.65 | 67.15 |
| 2026-01-11 16:30 | 41.69 | 55.19 | 68.69 |
| 2026-01-11 16:45 | 42.41 | 55.91 | 69.41 |
| 2026-01-11 17:00 | 41.96 | 55.46 | 68.96 |
| 2026-01-11 17:15 | 40.47 | 53.97 | 67.47 |
| 2026-01-11 17:30 | 38.49 | 51.99 | 65.49 |
| 2026-01-11 17:45 | 36.82 | 50.32 | 63.82 |
| 2026-01-11 18:00 | 36.18 | 49.68 | 63.18 |
| 2026-01-11 18:15 | 36.92 | 50.42 | 63.92 |
| 2026-01-11 18:30 | 38.91 | 52.41 | 65.91 |
| 2026-01-11 18:45 | 41.56 | 55.06 | 68.56 |
| 2026-01-11 19:00 | 44.10 | 57.60 | 71.10 |
| 2026-01-11 19:15 | 45.79 | 59.29 | 72.79 |
| 2026-01-11 19:30 | 46.28 | 59.78 | 73.28 |
| 2026-01-11 19:45 | 45.71 | 59.21 | 72.71 |
| 2026-01-11 20:00 | 44.65 | 58.15 | 71.65 |
| 2026-01-11 20:15 | 43.89 | 57.39 | 70.89 |
| 2026-01-11 20:30 | 44.13 | 57.63 | 71.13 |
| 2026-01-11 20:45 | 45.72 | 59.22 | 72.72 |
| 2026-01-11 21:00 | 48.50 | 62.00 | 75.50 |
| 2026-01-11 21:15 | 51.88 | 65.38 | 78.88 |
| 2026-01-11 21:30 | 55.06 | 68.56 | 82.06 |
| 2026-01-11 21:45 | 57.34 | 70.84 | 84.34 |
| 2026-01-11 22:00 | 58.35 | 71.85 | 85.35 |
| 2026-01-11 22:15 | 58.26 | 71.76 | 85.26 |
| 2026-01-11 22:30 | 57.63 | 71.13 | 84.63 |
| 2026-01-11 22:45 | 57.26 | 70.76 | 84.26 |
| 2026-01-11 23:00 | 57.84 | 71.34 | 84.84 |
| 2026-01-11 23:15 | 59.71 | 73.21 | 86.71 |
| 2026-01-11 23:30 | 62.69 | 76.19 | 89.69 |
| 2026-01-11 23:45 | 66.19 | 79.69 | 93.19 |
| 2026-01-12 00:00 | 69.40 | 82.90 | 96.40 |
| 2026-01-12 00:15 | 71.62 | 85.12 | 98.62 |
| 2026-01-12 00:30 | 72.51 | 86.01 | 99.51 |
| 2026-01-12 00:45 | 72.24 | 85.74 | 99.24 |
| 2026-01-12 01:00 | 71.39 | 84.89 | 98.39 |
| 2026-01-12 01:15 | 70.76 | 84.26 | 97.76 |
| 2026-01-12 01:30 | 71.03 | 84.53 | 98.03 |
| 2026-01-12 01:45 | 72.53 | 86.03 | 99.53 |
| 2026-01-12 02:00 | 75.07 | 88.57 | 102.07 |
| 2026-01-12 02:15 | 78.06 | 91.56 | 105.06 |
| 2026-01-12 02:30 | 80.69 | 94.19 | 107.69 |
| 2026-01-12 02:45 | 82.27 | 95.77 | 109.27 |
| 2026-01-12 03:00 | 82.47 | 95.97 | 109.47 |
| 2026-01-12 03:15 | 81.47 | 94.97 | 108.47 |
| 2026-01-12 03:30 | 79.88 | 93.38 | 106.88 |
| 2026-01-12 03:45 | 78.49 | 91.99 | 105.49 |
| 2026-01-12 04:00 | 77.99 | 91.49 | 104.99 |
| 2026-01-12 04:15 | 78.69 | 92.19 | 105.69 |
| 2026-01-12 04:30 | 80.40 | 93.90 | 107.40 |
| 2026-01-12 04:45 | 82.52 | 96.02 | 109.52 |
| 2026-01-12 05:00 | 84.25 | 97.75 | 111.25 |
| 2026-01-12 05:15 | 84.89 | 98.39 | 111.89 |
| 2026-01-12 05:30 | 84.15 | 97.65 | 111.15 |
| 2026-01-12 05:45 | 82.23 | 95.73 | 109.23 |
| 2026-01-12 06:00 | 79.73 | 93.23 | 106.73 |
| 2026-01-12 06:15 | 77.46 | 90.96 | 104.46 |
| 2026-01-12 06:30 | 76.11 | 89.61 | 103.11 |
| 2026-01-12 06:45 | 75.97 | 89.47 | 102.97 |
| 2026-01-12 07:00 | 76.85 | 90.35 | 103.85 |
| 2026-01-12 07:15 | 78.14 | 91.64 | 105.14 |
| 2026-01-12 07:30 | 79.04 | 92.54 | 106.04 |
| 2026-01-12 07:45 | 78.88 | 92.38 | 105.88 |
| 2026-01-12 08:00 | 77.37 | 90.87 | 104.37 |
| 2026-01-12 08:15 | 74.71 | 88.21 | 101.71 |
| 2026-01-12 08:30 | 71.55 | 85.05 | 98.55 |
| 2026-01-12 08:45 | 68.68 | 82.18 | 95.68 |
| 2026-01-12 09:00 | 66.78 | 80.28 | 93.78 |
| 2026-01-12 09:15 | 66.14 | 79.64 | 93.14 |
| 2026-01-12 09:30 | 66.56 | 80.06 | 93.56 |
| 2026-01-12 09:45 | 67.42 | 80.92 | 94.42 |
| 2026-01-12 10:00 | 67.92 | 81.42 | 94.92 |
| 2026-01-12 10:15 | 67.40 | 80.90 | 94.40 |
| 2026-01-12 10:30 | 65.58 | 79.08 | 92.58 |
| 2026-01-12 10:45 | 62.69 | 76.19 | 89.69 |
| 2026-01-12 11:00 | 59.36 | 72.86 | 86.36 |
| 2026-01-12 11:15 | 56.40 | 69.90 | 83.40 |
| 2026-01-12 11:30 | 54.48 | 67.98 | 81.48 |
| 2026-01-12 11:45 | 53.89 | 67.39 | 80.89 |
| 2026-01-12 12:00 | 54.40 | 67.90 | 81.40 |
| 2026-01-12 12:15 | 55.37 | 68.87 | 82.37 |
| 2026-01-12 12:30 | 56.02 | 69.52 | 83.02 |
| 2026-01-12 12:45 | 55.69 | 69.19 | 82.69 |
| 2026-01-12 13:00 | 54.11 | 67.61 | 81.11 |
| 2026-01-12 13:15 | 51.51 | 65.01 | 78.51 |
| 2026-01-12 13:30 | 48.55 | 62.05 | 75.55 |
| 2026-01-12 13:45 | 46.02 | 59.52 | 73.02 |
| 2026-01-12 14:00 | 44.60 | 58.10 | 71.60 |
| 2026-01-12 14:15 | 44.53 | 58.03 | 71.53 |
| 2026-01-12 14:30 | 45.59 | 59.09 | 72.59 |
| 2026-01-12 14:45 | 47.13 | 60.63 | 74.13 |
| 2026-01-12 15:00 | 48.36 | 61.86 | 75.36 |
| 2026-01-12 15:15 | 48.62 | 62.12 | 75.62 |
| 2026-01-12 15:30 | 47.66 | 61.16 | 74.66 |
| 2026-01-12 15:45 | 45.72 | 59.22 | 72.72 |
The quickstart backtest evaluates three consecutive historical windows, each from its own cutoff. Table 2 calculates error from synthetic predictions and synthetic observations, with a seasonal baseline for comparison. Lower CRPS, MAE, and RMSE indicate lower error; these values do not measure production model performance.
scores, rows = backtest(
context_df,
api_url=API_URL,
headers=HEADERS,
prediction_length=96,
windows=3,
quantile_levels=[0.1, 0.5, 0.9],
)
pd.DataFrame(scores["per_window"])| Metric | Seasonal baseline | Demo prediction |
|---|---|---|
| Continuous ranked probability score (CRPS) | 4.25 | 3.08 |
| Mean absolute error (MAE) | 4.25 | 2.48 |
| Root mean squared error (RMSE) | 4.98 | 3.02 |
| P10–P90 interval coverage | Not provided | 100.0% |
Baseline: repeat the value from 24 hours earlier, repeated from the last complete period before each window. Interval coverage is the share of synthetic observations inside the illustrative interval. CRPS evaluates the full predictive distribution. This demo assumes a uniform distribution whose P10 and P90 match the plotted interval; the baseline is a point forecast, for which CRPS equals MAE.
Use the forecast to compare dispatch periods. Add asset limits, round-trip efficiency, and transaction costs before you estimate revenue.