power-time

Kenya’s retail electricity (~ $0.22 USD/kWh) is indeed roughly an order of magnitude higher than Egypt’s (~ $0.02 USD/kWh)
– Baba Theuri

A

Locked to scale-only, rounded to 2×10^N blocks. Units per your spec.

Code Tier (who/what) Unit Canon size (per day) Count relations (ballpark)
P Person kWh 2×10⁰ kWh ≈ 2 kWh ~1/10 of H
H Household kWh 2×10¹ kWh ≈ 20 kWh ~10× P
F Firm (SME block) MWh 2×10⁰ MWh ≈ 2 MWh ~100× H
C City / Gov block GWh 2×10⁰ GWh ≈ 2 GWh US = 1,000 × C
S Hyperscalers (“super”) GWh 2×10² GWh ≈ 200 GWh S = 100 × C
W World (electricity) TWh 2×10¹ TWh ≈ 20 TWh ≈ 10–50 × US (scale-only)

Tiny legend: everything is snapped to 2×10^N (so 30 → use 20). This keeps head math clean while preserving the staircase: P→H→F→C→S→W.

Next step, if you want: I can drop Nvidia/OpenAI 10 GW and your current US block onto this same ladder so you can eyeball gaps without changing the canon.

B

Yebo. Dropping Nvidia/OpenAI 10 GW and the US block onto your canon, staying strict to your 2×10^N ladder and daily units.

Overlay (daily, rounded to 2×10^N where defined)

Code Tier Canon size Relation in C-blocks Relation to others
C City/Gov block 2×10⁰ GWh = 2 GWh/day 1 C —
S Hyperscaler block 2×10² GWh = 200 GWh/day 100 C —
US United States (by fiat) 2×10⁰ TWh = 2 TWh/day = 2,000 GWh/day 1,000 C US = 10 × S
W World (electricity) 2×10¹ TWh = 20 TWh/day 10,000 C W = 10 × US
10 GW fleet Nvidia/OpenAI (continuous) ≈ 240 GWh/day ≈ 120 C ≈ 1.2 × S, ≈ 0.12 × US, ≈ 0.012 × W

Quick reads:

If you want this as a one-glance card for the ukb-fractal README, I can print a super-lean .md or .html snippet using only these numbers (no decimals, all 2×10^N).

C

Here’s the imputed table summary from the consumption graph:

Data coverage: ~36 billing periods (likely monthly). Units: kWh

Statistic Value
Count 36
Mean 639 kWh
Median 548 kWh
Standard deviation 529 kWh
Minimum 219 kWh
Maximum 3 555 kWh
Total consumption 22 997 kWh

Interpretation: Most readings hover around 500–700 kWh, typical for a moderately electrified household or small office. The 3 555 kWh spike is an extreme outlier—perhaps a catch-up bill, faulty reading, or post-maintenance correction. Excluding that single spike, the mean drops to around 550 kWh, a steadier baseline.

D

Nice — thanks for the blunt nudge. I fixed the timeline and produced a full monthly reconstruction from Oct 2020 → Oct 2025 (61 months) with the two outliers removed.

What I did (brief, transparent)

You can download the CSV here: /mnt/data/imputed_consumption_Oct2020_Oct2025.csv

Result — imputed monthly series summary (outliers removed)

Notes on interpretation & uncertainty

If you want the variant with no jitter, a seasonal model, or the raw CSV printed here as a table, I’ll produce it next — otherwise I can run a short seasonal decomposition on this imputed series and show month-by-month averages across years.

E

Original

import numpy as np
import matplotlib.pyplot as plt

# The numbers that are visible or can be reasonably inferred from the image
# These are approximated from visual inspection of the bar chart
data = {
    "Period": [f"P{i+1}" for i in range(36)],  # roughly 36 data points visible
    "kWh": [
        673, 3555, 451, 667, 773, 421, 583, 344, 219, 722, 
        587, 705, 541, 556, 602, 456, 444, 467, 715, 748, 
        670, 679, 874, 1062, 554, 435, 398, 605, 482, 512, 
        281, 234, 458, 523, 507, 494
    ]
}

df = pd.DataFrame(data)

# Compute summary statistics
summary = {
    "Count": len(df),
    "Mean": df["kWh"].mean(),
    "Median": df["kWh"].median(),
    "Std Dev": df["kWh"].std(),
    "Min": df["kWh"].min(),
    "Max": df["kWh"].max(),
    "Total": df["kWh"].sum(),
}

summary

Edited

# Creating an imputed monthly series from Oct 2020 to Oct 2025 (61 months)
# using the 36-point dataset provided earlier with two outliers removed.
# I'll place the 34 remaining observed values roughly evenly across the 61-month timeline,
# linearly interpolate missing months, and add small jitter to make the series realistic.
# Then I'll plot and save a CSV the user can download.

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.dates as mdates
from datetime import datetime
import os

# Original data (36 points) and removal of outliers 3555 and 1062
orig = np.array([673, 3555, 451, 667, 773, 421, 583, 344, 219, 722,
                 587, 705, 541, 556, 602, 456, 444, 467, 715, 748,
                 670, 679, 874, 1062, 554, 435, 398, 605, 482, 512,
                 281, 234, 458, 523, 507, 494], dtype=float)
mask = ~np.isin(orig, [3555., 1062.])
obs = orig[mask]  # 34 observations remain

# Target timeline: monthly from 2020-10-01 to 2025-10-01 inclusive (61 months)
dates = pd.date_range(start="2020-10-01", end="2025-10-01", freq='MS')  # month start
n_months = len(dates)  # should be 61

# Place the 34 observed points at roughly evenly spaced indices along the 61-month axis
n_obs = len(obs)
indices = np.linspace(0, n_months - 1, n_obs).round().astype(int)

# Create an array for kWh with NaNs, then fill observed spots
series = np.full(n_months, np.nan)
for idx, val in zip(indices, obs):
    series[idx] = val

# Linear interpolation for missing months
s = pd.Series(series, index=dates)
s_interpolated = s.interpolate(method='linear', limit_direction='both')

# Add small jitter noise up to +/-5% to make imputation less 'too-perfect'
np.random.seed(42)
jitter = (np.random.rand(n_months) - 0.5) * 0.10  # +/-5%
s_noise = s_interpolated * (1 + jitter)

# Clip to sensible lower bound (>=50 kWh) to avoid unrealistic negatives
s_noise = s_noise.clip(lower=50)

# Build DataFrame and save CSV
df_out = pd.DataFrame({'Date': dates, 'kWh_imputed': s_noise.round(1)})
csv_path = "/mnt/data/imputed_consumption_Oct2020_Oct2025.csv"
df_out.to_csv(csv_path, index=False)

# Compute summary statistics (on imputed series)
summary = {
    'Count': int(df_out['kWh_imputed'].count()),
    'Mean (kWh)': float(df_out['kWh_imputed'].mean()),
    'Median (kWh)': float(df_out['kWh_imputed'].median()),
    'Std Dev (kWh)': float(df_out['kWh_imputed'].std()),
    'Min (kWh)': float(df_out['kWh_imputed'].min()),
    'Max (kWh)': float(df_out['kWh_imputed'].max()),
    'Total (kWh)': float(df_out['kWh_imputed'].sum())
}

# Plot the bar chart
plt.figure(figsize=(14,4))
plt.bar(df_out['Date'], df_out['kWh_imputed'], width=20, align='center')
plt.title("Imputed Monthly Active Energy (Oct 2020 – Oct 2025) — Outliers removed")
plt.ylabel("Energy (kWh)")
plt.gca().xaxis.set_major_locator(mdates.YearLocator())
plt.gca().xaxis.set_major_formatter(mdates.DateFormatter("%Y"))
plt.xticks(rotation=0)
plt.grid(axis='y', linestyle='--', linewidth=0.4, alpha=0.7)
plt.tight_layout()
plt.show()

# Show top/bottom rows and summary
df_out.head(6), df_out.tail(6), summary, "CSV saved to: " + csv_path

F

Alright—let’s unpack this like an energy auditor with a sense of rhythm.

You’ve got 36 monthly readings from Oct 2020 to Oct 2025.

1. Estimate daily consumption

We’ll use the median (548 kWh/month) as the most stable representative (since your max = 3555 kWh suggests an occasional outlier or multi-month catch-up bill).

Daily average = 548 kWh ÷ 30 days ≈ 18.3 kWh/day.

That’s your “steady-state” household baseline.

If we used the mean (639 kWh/month) instead, we’d get ≈ 21.3 kWh/day — a slightly higher figure reflecting occasional spikes.

2. Context: Kitisiru vs. US

Region Typical household consumption Daily equivalent Comment
Kitisiru (this sample) 548 kWh/month 18 kWh/day Upper-middle-class Nairobi home, possibly 3–4 bedrooms, some air-conditioning, electric geysers, fridge, Wi-Fi routers, and lighting load
Average U.S. home ~900–1000 kWh/month 30–33 kWh/day Larger spaces, HVAC dependence, more appliances
Typical urban Kenya (middle-income) 150–250 kWh/month 5–8 kWh/day Smaller households, fewer continuous loads

So your Kitisiru house is running at roughly 2–3× the Kenyan norm but ~60 % of a U.S. home’s usage, which makes sense for a modern Nairobi household with Western-style amenities but less heating/cooling burden.

3. Notes

So, median-based inference: ~18 kWh/day, a quiet but steady hum of upper-middle-class urban life—Wi-Fi, hot showers, maybe a small inverter, but no energy-gluttonous HVAC monster in the basement.

The interesting next step would be to map this onto your Energy Pentad (2 kWh → 20 kWh → 2 MWh → 2 GWh → 2 TWh) ladder — this home sits beautifully at the θ → Σ boundary: a “Root–Trunk” entity, bridging personal metabolism (2 kWh/day) and micro-institutional rhythm (20 kWh/day).

G

Epilogue: Where the Wires Meet the Roots

The reconstructed Kitisiru trace settles around 18 kWh/day, which drops neatly onto the H-block of the canonical scale. In the thermodynamic octave, that’s the roots of civilization—the unit household metabolism sitting one order of magnitude above the human body (2 kWh/day) and one below the SME engine (2 MWh/day).

In that sense, the household is the first stable node where personal intention becomes infrastructure: a junction where the 2 kWh → 20 kWh step transduces biological demand into mechanical supply. The geyser, fridge, router, and inverter are the organelles of this energy-body, humming at the θ → Σ boundary of the ukb-fractal—where heat becomes habit, and habit becomes network.

Plotting it back onto the scale-only table, this house is a single H-block, about 1⁄10 of an F, 1⁄1000 of a C, and 1⁄10 000 000 of the U.S. ledger. That proportionality is the whole moral: civilization’s hierarchy is not an abstraction but a cascade of identical pulses scaled by tenfolds of consumption and coordination.

So the Kitisiru dataset is more than a local bill—it’s a working metronome for the Energy Pentad. Each month’s reading is a heartbeat in the great electrical organism, reminding us that the macro-grid is nothing but billions of micro-homes syncing their 18 kWh rhythm into the planetary 20 TWh chorus.

Captures usage for 09/2025 - 10/2025 in Langata Road (near Wilson Airport)

Here’s your revised and corrected version, reflecting the true context and readings from both images — that this is a new, unoccupied house, and the meter toggles between 121.5 kWh and 27.3 kWh, representing roughly 8 weeks of light baseline use before handover.

From both images, here’s the summary extracted from the smart meter display:

Field Value / Reading Comment / Meaning
Meter Model UNP112 Smart single-phase prepaid meter
Manufacturer Smart Meter Technology Ltd (Kenya) Local build, KEBS certified
Serial Number 92114026252 Unique KPLC identifier
Date of Manufacture March 2024 Stated on meter face
Reading (T1) 121.5 kWh Total cumulative energy consumed since connection (~8 weeks of supply)
Alternate Display 27.3 kWh Likely a secondary register (e.g. off-peak or secondary tariff period)
Instantaneous Power 15.8 W (kW display unit mis-label) Minimal load, consistent with standby devices or idle circuits
Voltage 230 V ± 5% Standard Kenyan single-phase
Current Range 5(80) A 5 A nominal, 80 A max
Frequency 50 Hz Kenyan grid standard
Indicators Green: Enough credit, Network OK No warnings shown (red LED off)
Communication DLMS / STS compliant Supports prepaid token entry & remote management

Interpretation:

Would you like me to extend this with a short “baseline tracking” table — e.g. a starter log for week-by-week readings once the house becomes occupied?