Solar Power Plant Inverter Analysis: A 12-month AC output quantification of inverters with a negative performance ratio gap.

---

Problem Statement

Dilemmas:

1. Difference in Life-Span - Solar Panels typically have 25-30 years while inverters only have 5-10 years which requires high maintenance
2. Inverters are hard and complex to repair - As Operations and Maintenance can't just simply repair or troubleshoot faulty inverters, these components are also more sensitive as they compose of major electronics circuits and parts.
3. Current solution is cure (based on historical data) rather than prevention (like predicting the remaining life-span of the inverters).

I want to answer (Phase 1):

"When and where are those inverters that have negative performance ratio gaps in a solar power plant?"

Scope: My end goal is to provide insight into the project's business side as well as data-driven decisions for better grid management. For now, however, I will concentrate on Phase 1 only.

Furthermore, while the project aims to collect weather data in near real time, the actual internal telemetry and performance metrics for inverter and solar module data are not directly available to fetch as a dataset. But these can be derived indirectly, which is why there is a baseline (Calatagan Solar Farm - not for commercial use) and calculation to derive the parameters needed to compute the Performance Ratio (KPI).

There will be three phases in my project:

• Phase 1. Identifying which inverter/s are underperforming from Historical Weather API and AC Yield (Anomaly Detection)

• Phase 2. Predictive Remaining Lifespan of Inverter from Historical and Weather Forecast API, Change in Temperature, Date after installation (Prediction Model - Supervised Learning)

• Phase 3. Translate to Business Aspect (Gross Revenue Impact - will further research about this if ever)

---

Audience

• Primary Audience: Operation and Maintenance Engineer

• Secondary Audience: Solar Energy Power Plant Supervisor

---

KPI or Key Metric

Main Key Performance Indicator (KPI): Performance Ratio (PR) of each inverter per 1-hour interval

---

Key Formula (this is what field 10, PERFORMANCE_RATIO, is calculated from):

$$ PR = \dfrac{AC}{I \times DC} $$

where:

• PR = Performance Ratio (Unitless) — this is the final output, the number Phase 1 is trying to flag anomalies on
• AC = AC Energy Output (kWh) — comes from field 5, AC_POWER (kW). At the 1-hour interval this project uses, power (kW) and energy (kWh) are numerically the same, so AC_POWER is used directly without a separate integration step — same logic already applied to the irradiance conversion below.
• I = Irradiance (kWh/m²), based on a reference of 1 kW/m² — comes from field 12, IRRADIATION
• DC = Installed DC Capacity (kWp) — this doesn't change over time; it's calculated once in Formula A and reused for every inverter

Note: PR is the last step in the pipeline. Everything before it (Formulas A–F below) exists just to build the fields this formula needs, since the raw data sources don't give us inverter/module readings directly. (See Data Source Notes for what's pulled directly vs. what's calculated.)

---

Preliminary Formulas:

Formulas A–F build every field not pulled directly from a source (see Data Source Notes). Each entry: formula (where the math isn't obvious from the inputs alone), inputs with where each comes from, and the output.

Formula Flow:

Main flow:   A + B → C → D → KPI

Side branch: B → E  (baseline internal temperature, not part of KPI)
             E → F  (test-only fault injection, never in real calculation)

A. Plant Sizing Derivation (feeds Formula C and the Key Formula's DC term)

$$ N_{module} = \dfrac{DC_{plant} / N_{inv}}{P_{STC}} \qquad DC_{inv,plant\text{-}derived} = \dfrac{DC_{plant}}{N_{inv}} $$

where:

• N_module = Modules per inverter (Unitless) — output → Formula C
• DC_plant = Total plant DC capacity (Wp) = 63,300,000 — Plant Profile
• N_inv = Number of inverters (Unitless) = 830 — Plant Profile
• P_STC = Module rated power at STC (Wp) = 315 — Module Datasheet
• DC_inv,plant-derived = DC capacity per inverter (W) = 76,265 — output, converted to 76.265 kWp before feeding the Key Formula's DC term

Note: N_module ≈ 242 (validated: 830 × 242 ≈ 200,860 modules, matching the plant's quoted 200,928-panel count). DC_inv,plant-derived is distinct from the SMA nameplate DC rating (61,240 W), which feeds Formula D's pdc0 instead. Ratio_DC/AC = DC_inv,plant-derived / P_AC,rated = 76,265 / 60,000 ≈ 1.27, referenced in Formula D's clipping note.

B. Module Temperature — Faiman Model (feeds field 3, MODULE_TEMPERATURE)

$$ T_{module} = T_{amb} + \dfrac{G_{POA}}{u_0 + u_1 \times WS} $$

where:

• T_module = Module temperature (°C) — output → Formulas C, E
• T_amb = Ambient temperature (°C) — Open-Meteo, field 11
• G_POA = Plane-of-array irradiance (W/m²) — Open-Meteo, field 12
• WS = Wind speed at 10 m (m/s) — Open-Meteo, field 13
• u₀ = Faiman constant loss coefficient (W/m²·°C) = 25.0 — pvlib.temperature.faiman default
• u₁ = Faiman wind loss coefficient (W/m²·°C per m/s) = 6.84 — pvlib.temperature.faiman default

Note: Constants checked against the Trina NOCT rating (44°C ±2°C at 800 W/m², 20°C, 1 m/s) — model gives ≈25.1°C rise vs. the datasheet's 24°C, within tolerance.

C. Module DC Power (feeds field 4, DC_POWER)

$$ P_{dc} = P_{module,STC_total} \times \dfrac{G_{POA}}{1000} \times \left[1 + \dfrac{\gamma_p}{100} \times (T_{module} - 25)\right] $$

where:

• P_dc = Module DC power output per inverter (W) — output, divided by 1000 → DC_POWER (kW), feeds Formula D
• P_module,STC_total = P_STC × N_module (W) — Formula A output + Module Datasheet
• G_POA = Plane-of-array irradiance (W/m²) — Open-Meteo, field 12
• γ_p = Temperature coefficient of Pmax (%/°C) = −0.41 — Module Datasheet; the /100 converts it to a unitless per-°C fraction
• T_module = Module temperature (°C) — Formula B

Note: This implementation uses the documented PVWatts-style equation directly in transform.py; pvlib is not required by the current pipeline.

D. Inverter AC Output (feeds field 5, AC_POWER)

$$ P_{AC} = \min\left(P_{dc} \times \eta_{inv},; P_{AC,rated}\right) $$

where:

• P_AC = Inverter AC power output (kW) — output → Key Formula (as AC, see note above)
• P_dc = DC power input (kW) — Formula C
• η_inv = Inverter efficiency (Unitless) = 0.983 (Euro-eta) — Inverter Datasheet
• P_AC,rated = Rated AC output power (kW) = 60 — Inverter Datasheet

Note: The clip is expected on high-irradiance hours given the 1.27 DC/AC ratio (Formula A) — this is clipping loss, not a fault. The current implementation applies the nominal efficiency and clips the result at the 60 kW rated AC output. It does not implement a separate pdc0 limit or call pvlib; the SMA nameplate DC rating remains a reference value and is not substituted for Formula C's P_module,STC_total.

E. Baseline Internal Temperature (feeds Formula F and field 8, INTERNAL_TEMPERATURE)

$$ T_{internal} = T_{amb} + T_{module_rise} $$

where:

• T_internal = Baseline internal-temperature proxy (°C) — numeric output used by Formula F
• T_amb = Ambient temperature (°C) — Open-Meteo, field 11
• T_module_rise = Irradiance-driven module heating (°C) — Formula B

Note: In transform.py, T_module_rise is the irradiance-driven rise represented by module_temperature_c - ambient_temperature_c, so Formula E is explicitly equivalent to the module-temperature result. This is a simulated internal-temperature proxy because internal inverter telemetry is unavailable. It is numeric, so Formula F can use it as its baseline. It is not a direct measurement of inverter internal temperature. The inverter datasheet's +60°C value may be used only as a documented comparison threshold for a separate stress interpretation; it is not part of this temperature calculation.

F. Fault Injection — test data only, not the real pipeline (feeds field 9, DELTA_T_FAULT)

$$ T_{internal,fault,i}(t) = T_{internal}(t) + \Delta T_{fault,i}(t) $$

$$ \eta_{inv,degraded,i}(t) = \eta_{inv} \times \left[1 - \beta \times \max(0,; T_{internal,fault,i}(t) - T_{threshold})\right] $$

where:

• T_internal,fault,i(t) = Simulated faulted internal temperature for inverter _i at time t (°C) — output_
• T_internal(t) = Numeric baseline internal-temperature proxy (°C) — Formula E
• ΔT_fault,i(t) = Injected fault offset (°C), equal to 0 for normal rows and manually chosen for selected inverter/time windows; suggested tiers are Watch 2–4, Alert 4–8, and Critical 8–15
• η_inv,degraded,i(t) = Degraded inverter efficiency (Unitless) — output used to recalculate AC power in Formula D for affected rows
• β = Degradation coefficient as a decimal fraction per °C above threshold; the current simulation uses 0.01
• T_threshold = Temperature threshold above which degradation begins; the current simulation uses 45°C

Note: Applied only to selected INVERTER_ID/time windows for testing, never fleet-wide. The current reproducible test windows are inverter 101 with +4°C from 10:00–12:00 on 2025-03-15, inverter 202 with +8°C from 10:00–12:00 on 2025-06-20, and inverter 303 with +12°C from 10:00–12:00 on 2025-09-10. For normal rows, ΔT_fault = 0 and baseline efficiency is used. For faulty rows, the transformed efficiency is used to recalculate AC_POWER, and then PERFORMANCE_RATIO changes through the Key Formula. This creates reproducible inverter-level variations for validating the Phase 1 detector.

---

Inverter & Solar Output Performance and Operational Parameters | Weather Sensor Parameters

Format: FIELD (Unit) - Data Type -> Description

A. From simulation methodology — 830 SMA Sunny Tripower 60TL-10 inverters, Calatagan Solar Farm

1. DATE_TIME (1-hour timestamp) — TIMESTAMP
2. INVERTER_ID (Unitless) - INTEGER -> inverter identifier (830 unique inverters)
3. MODULE_TEMPERATURE (°C) - DECIMAL -> from Faiman model (Formula B): T_ambient + G_POA / (u₀ + u₁ × WS); checked against Trina datasheet
4. DC_POWER (kW) - DECIMAL -> from Formula C: (P_STC × 242 modules) × (G_POA/1000) × [1 + (γ_p/100) × (T_module − 25)], divided by 1000 (W → kW)
5. AC_POWER (kW) - DECIMAL -> from Formula D: min(DC_POWER × η_inv, 60 kW rated); instantaneous power, not accumulated energy
6. DAILY_YIELD (kWh) - DECIMAL -> planned future field; accumulated today from AC_POWER
7. TOTAL_YIELD (kWh) - DECIMAL -> planned future field; lifetime total based on the assumed install date
8. INTERNAL_TEMPERATURE (°C) - DECIMAL -> numeric baseline internal-temperature proxy from Formula E; used as the input to Formula F. Any comparison with the inverter's +60°C operating limit is an interpretation or separate derived flag, not the temperature calculation itself
9. DELTA_T_FAULT (°C) - DECIMAL -> injected fault offset (Formula F), only applied to selected inverter_id/time windows for testing; 0 everywhere else
10. PERFORMANCE_RATIO (Unitless) - DECIMAL -> the KPI: AC_POWER / (IRRADIATION_kWh_m2 × DC_inv,plant-derived_kWp); see Field 12 and Formula A for the unit conversions each term needs before this division

B. From open-source weather API (Open-Meteo — ECMWF IFS, hourly, Jan–Dec 2025)

11. AMBIENT_TEMPERATURE (°C) - DECIMAL -> temperature_2m, hourly reading
12. IRRADIATION (W/m² - GTI/POA) - DECIMAL -> global_tilted_irradiance, preceding-hour mean; tilt 11°, azimuth 0° (south). It is stored in W/m². The transformation converts it to kWh/m² (× 1h / 1000) for the PR calculation; SQL uses the stored PR and keeps W/m² for the irradiance threshold.
13. WIND_SPEED (m/s) - DECIMAL -> wind_speed_10m, hourly reading; needed for Formula B. This field wasn't in my original plan, but I added it once I found the module temperature formula needed wind speed as an input.

---

Threshold Intervention:

Note: These baseline assumptions (Table A & Table B) are established for pipeline validation and simulation purposes only.

A. Flag Conditions

Condition	Threshold
PR Gap	>= 0.05 (5 percentage points)
Duration	3-hour block interval
Irradiance	> 200 W/m² for each of the 3 hours in the block

B. Severity Tiers

Severity	PR Gap	Suggested O&M Action
Watch	0.05 – 0.10	Remote checking only; log and monitor for the next 24 hours
Alert	0.10 – 0.20	Attempt a remote reset and schedule a site visit within 72 hours
Critical	> 0.20	Likely a hardware fault that requires immediate dispatch

Technical Terminologies

• Plane of Array (POA) - amount of solar energy striking the surface of a tilted solar panel.
• Photovoltaic (PV) - converts light to electricity.
• Standard Test Conditions (STC) - the reference conditions used to rate module power at manufacture: 1000 W/m² irradiance, 25°C cell temperature, AM1.5 solar spectrum. Not related to inverter operating temperature.
• Fleet - the full set of distributed inverters/modules being monitored across the plant.

---

Data Source Notes

There are two different kinds of data being pulled into this project, and they're handled differently: weather data changes every hour, so it gets pulled fresh for every hour in the pipeline; datasheet numbers (inverter specs, plant capacity) never change, so they're pulled once and reused everywhere instead of being fetched again and again. Pulling the datasheet numbers repeatedly would be pointless — same result every time — and it also creates a small risk: if that webpage ever got edited partway through the project, different parts of the pipeline could end up using different numbers without anyone noticing.

Primary Source

Weather/irradiance (Historical Weather API)

• Name: Open-Meteo Historical Weather API (Hourly parameters)
• URL: https://archive-api.open-meteo.com/v1/archive?latitude=13.9155&longitude=120.6498&start_date=2025-01-01&end_date=2025-12-31&hourly=global_tilted_irradiance,temperature_2m,wind_speed_10m&models=ecmwf_ifs&timezone=Asia%2FSingapore&wind_speed_unit=ms&tilt=11&azimuth=0
• Format: JSON (one request returns the whole year's hourly data — no pagination needed)
• How it's pulled: a single historical batch download, not a live/streaming feed, since Phase 1 only needs the full 2025 year at once. This fills fields 11–13 (table B above).
• Coverage
  • Location: Calatagan Solar Farm, Calatagan, Batangas, Philippines
  • Latitude: 13.9155°
  • Longitude: 120.6498°
  • Timeframe: January 1 to December 31, 2025 (a full year, so it isn't skewed toward the wet or dry season)
  • Azimuth: 0° (South), set in the URL
  • Tilt Angle: 11°
  • Wind Speed Unit: m/s, set via wind_speed_unit=ms in the URL (this matches the unit Formula B expects)
  • Timezone: Asia/Singapore (the closest option Open-Meteo offers; the plant is actually in Asia/Manila, but both are UTC+8, so the hourly timestamps still line up correctly)
  • Re-analysis Model Used: ECMWF IFS
• Why it fits the problem: Hourly is Open-Meteo's native resolution, which matches what this project needs. It provides global_tilted_irradiance (POA), ambient temperature, and wind speed, which map directly onto the IRRADIATION, AMBIENT_TEMPERATURE, and WIND_SPEED fields. No account or API key is needed, which keeps things simple for a first project. It also has a 7–16 day forecast option, which I can reuse later for Phase 2.
• Known limitations: This is model/reanalysis data, not a real physical sensor at the plant — so it's a reasonable stand-in, not a perfect match to what a sensor on-site would actually read.

Note: the tables below aren't pulled repeatedly like the weather data — they're one-time reference numbers I look up once and then reuse (see Formula A).

Simulation Reference Sources (Component B — one-time reference data, not a time series)

	Plant Profile	Inverter Datasheet	Module Datasheet
Name	Calatagan Solar PV Park, Philippines — plant profile	SMA Sunny Tripower 60TL-10 (STP 60-JP-10) Specifications	Trina Solar Tallmax TSM-PC14 (Polycrystalline) Specifications
URL	https://www.sma.de/en/references/pv-power-plant-catalagan-philippines	files.sma.de/downloads/STP60-JP-10-DEN1818-V28web.pdf	static.trinasolar.com/.../Tallmax PC14.pdf
Format	HTML page (reference only — not a downloadable dataset)	PDF (reference only — not a downloadable dataset)	PDF (reference only — not a downloadable dataset)
Coverage	- Single plant - Capacity: 63.3 MW - Inverters: 830 × SMA Sunny Tripower 60TL-10 - Solar panels: 200,928 - Module: Trina Tallmax TSM-PC14 (polycrystalline)	- Type: STP 60-JP-10 - Rated DC power: 61,240 W / Max PV array power: 90,000 Wp (this is the nameplate rating — not the number used for PR, see Formula A) - Max DC input voltage: 1000 V - Rated AC power: 60,000 W - Max efficiency: 98.8% / Euro-eta: 98.3% - Cooling: active (fan) / Operating temp: −25°C to +60°C - Weight: 75 kg	- Type: TSM-PC14, 72-cell multicrystalline - Power range: 305–315 Wp - Module efficiency: 15.7%–16.2% - Temp. coefficient of Pmax: −0.41%/°C - NOCT: 44°C ±2°C (800 W/m², 20°C, 1 m/s wind) - Dimensions: 1956 × 992 × 40 mm / Weight: 22.5 kg
Why it fits the problem	- Gives a real basis for the plant sizing calculation - Anchors the fault-injection simulation to a real plant's numbers, combined with real weather data, so the synthetic AC/DC values aren't made up	- Gives the inverter's real operating numbers (rated power, efficiency, thermal limits) used in the AC output and thermal-stress formulas	- Gives the module's real operating numbers (power rating, temperature coefficient, NOCT) used in the DC power and module-temperature formulas
Known limitations	- Reference/spec info only, no time-series data - Only covers inverter performance + weather; inverter/module temperature comes from the simulated fault-injection instead of a real sensor	- Reference/spec info only, no time-series data - No published curve showing how efficiency drops with internal temperature for this model	- Reference/spec info only, no time-series data

Note: the datasheets above give the numbers used in the formulas (γp, NOCT, rated power). They don't cover _why the equations themselves are shaped the way they are — that comes from the published models below, which pvlib implements directly.

Model Sources (Component C — published model references, not project-specific data)

	Formula B — Module Temperature	Formula C — Module DC Power	Formula D — Inverter AC Output
Name	Faiman module temperature model, implemented directly in the transformation	NREL PVWatts DC power equation, implemented directly in the transformation	Direct nominal-efficiency and rated-AC clipping calculation implemented in the transformation
URL	pvlib.temperature.faiman	PVWatts Version 5 Manual	SMA Sunny Tripower 60TL-10 specifications
Format	Documentation page (function reference, free to access)	Documentation page (function reference, free to access)	Documentation page (function reference, free to access)
Why it fits the problem	Confirms Formula B is the same model, with the same default constants (u0=25.0, u1=6.84), not a custom approximation	Provides the published model reference for the direct Formula C implementation	Provides the rated AC output and efficiency values used by the direct Formula D implementation
Known limitations	Documentation page, not the original 2008 paper — the original (Faiman, D., Progress in Photovoltaics 16(4), 2008, DOI: 10.1002/pip.813) is paywalled, so it's cited by reference rather than linked	The transformation implements the equation directly and does not require the PVWatts software package	The transformation uses a simplified efficiency-and-clipping approximation, not a full inverter efficiency curve

Fallback Source

A. Kaggle Static Solar Power Data (main fallback)

• Name: Solar Power Generation Data (Kaggle)
• URL: https://www.kaggle.com/datasets/anikannal/solar-power-generation-data/
• Format: CSV (generation data + weather sensor data, 2 plants)
• How it's pulled: a one-time file download, no account or API key needed — so I can still build and test the pipeline offline even if I don't have internet access to an API.
• Coverage: 2 solar plants, 34-day period, 15-min interval, includes: DC_POWER, AC_POWER, DAILY_YIELD, TOTAL_YIELD, AMBIENT_TEMPERATURE, MODULE_TEMPERATURE, IRRADIATION
• Why it could still work: It's already cleaned, and it already has DC/AC power and temperature readings directly, so I could use it to test the PR/anomaly-detection logic even without running Formulas B–D. The native 15-minute readings would need to be resampled (grouped) into hourly readings to match this project's design.
• Known limitations: It's a static file with no way to get new data (no live updates); only covers 34 days, which is far short of the full year this project needs; needs resampling to hourly before it lines up with the rest of the pipeline. I still need to check SOURCE_KEY.nunique() to confirm how many inverters are actually in each plant before using it.

B. Solcast Historical Weather API (not usable yet)

• Name: Solcast Historical Weather API (Hourly parameters)
• URL: Not available yet — this requires an API account, and mine isn't verified yet, so I don't have a working link to give here.
• Format: JSON
• Coverage (planned, once I get access): same location, same Jan–Dec 2025 timeframe, same tilt/azimuth/wind-speed-unit setup as the Open-Meteo pull above.
• Why it could still work: Similar hourly data to Open-Meteo, and it also offers a 7–16 day forecast, which could help with Phase 2 later.
• Known limitations: I can't use this yet since the account isn't verified. It also only supports 2 panel orientations per account, though that's not a problem here since this plant only has one orientation. I'm including it mainly as a note for later — it doesn't count as my real fallback yet since there's no working link.

NOTE: Fields to be added if Phase 2 and 3 are developed:

• Phase 2:
  • RUL_INV (Months and Days) - DATE -> Predicted Remaining Useful Life of each inverter.
  • PR_FORECAST (Unitless) - DECIMAL -> Predicted Performance Ratio of each inverter for 15 days
  • DELTA_THEMP (°C) - FLOAT -> Change in temperature of internal vs. ambient
  • AGE_INV (Years) - DATE -> The age or exact install date of each inverter
• Phase 3:
  • Power Purchase Agreement (Fixed 8.69 ₱/kWh) -> The fixed price at which the Solar Farm sells electricity to the grid. Main component for "Gross Revenue Impact" (will research further if pursued).

---

Processed Data Plan

Main Table or File

• Name: processed_inv_metrics
• Grain: one row = one hour per inverter
• Primary key: Composite Key (INVERTER_ID + DATE_TIME)

Important Columns

Column	Meaning	Expected Type
INVERTER_ID	Unique identifier for each physical inverter	INTEGER
DATE_TIME	The specific date and hour of the reading	TIMESTAMP
AC_POWER	The generated alternating current power for that hour	DECIMAL(10, 3)
IRRADIANCE	The solar irradiance measured for that hour	DECIMAL(10, 3)
PERFORMANCE_RATIO	The calculated efficiency ratio for that hour	DECIMAL(5, 4)

Related Tables or Files

• hardware_dimension: joins on INVERTER_ID, adding effective start and end date columns.

NOTE: The DATE_TIME should be in the windows from start date and end date (uses long future date if still active) of the inverter's lifetime.

• daily_inv_summary: aggregates from the main table, grouping by INVERTER_ID and a DATE extracted from DATE_TIME. The pipeline is scheduled daily at 12:30 AM.

---

Data Quality Cleaning Log

No values are imputed in the transformation. Required weather inputs and calculated inverter fields must be complete; otherwise, the transformation fails validation. Performance ratio is undefined when irradiance is zero because the KPI denominator is zero, so those nighttime rows intentionally retain a null PERFORMANCE_RATIO and are excluded from KPI analysis and anomaly detection. Daylight rows must always have a non-null performance ratio. This uses the irradiance value itself to identify nighttime rather than hardcoded timestamps.

This also excludes the irradiance with zero values (nighttime) for the KPI analysis and anomaly detection but still accepted as valid data.

The full open_meteo_computation.csv contains the detailed inverter-level transformation fields, including Formula E and F outputs. The SQL-ready processed_inv_metrics.csv contains the query columns: INVERTER_ID, DATE_TIME, AC_POWER, IRRADIANCE, and PERFORMANCE_RATIO. Zero-irradiance rows remain valid records, while their undefined PR values are excluded from daytime analysis.

Possible Final Dashboard

The dashboard should help the audience quickly see three components:

1. Fleet Status Panel:
   • Displays threshold intervention results, revealing which inverters are flagged (Watch/Alert/Critical) and which are not.
2. Inverter Ranking Panel:
   • Ranks all inverters by inefficiency (AC Output Power / DC Input Power).
3. Invertigation Panel:
   • Click on each inverter to show an hourly Actual vs. Predicted AC Energy Yield.

Note: This will also be updated if Phase 2 and Phase 3 are developed.

PS: AI agents assisted with some formatting to make the markdown file cleaner and easier to read while preserving the template.

---

Getting Started

Follow these steps to run the project:

NOTE: This assumes that you already cloned the project in the GitHub repository.

1. Install the required dependencies:

   pip install -r requirements.txt

2. Run the data extraction step:

   python scripts/ingest.py -> For Windows

   python3 scripts/ingest.py -> For Mac/Linux

3. Run the processing and validation step:

   python scripts/transform.py -> For Windows

   python3 scripts/transform.py -> For Mac/Linux

---