Vietnamese manufacturing entities will soon have a new electricity tariff that explicitly charges them for the peak demand (in kW), as opposed to just energy consumption (in kWh). These new two-component tariffs will be in effect from July 2026 and a trial tariff has been issued now to support the transition.
The case study analyzes the load/demand of an hypothetical manufacturing facility, and evaluates the impact of the trial two-component tariffs relative to the current baseline tariffs
The analysis highlights the importance of quantifying the intuitive ideas (e.g., two component tariffs will drive reduction in system peak demand), and gain further insights into the economic value of private investments.
The study is divided into several sections
The sections below also include some of the prompts and Agentic AI outputs, as well as the insights derived from the analysis (a joint AI + human endeavor)
Analysis of Factory Electricity Consumption
First, let's examine the load profile of a hypothetical manufacturing facility, connected at 22 kV voltage level to EVN.
Below is a time-series 30 min interval data for the whole year for the facility, as well as the Load Duration Curve.
(developed using XanhTerra's AI analytics tool)
XanhTerra AI Prompt and Execution Plan:
Results from the Agentic AI analysis:
Insights:
(Combination of AI analysis + Human Interpretation)
- Energy consumption distribution by TOU categories shows majority of the demand in Standard (57.2%), Off-Peak (25.0%), Peak (17.8%)
- Consumption is quite low on Sundays and holidays, indicating that the factory is mostly closed on those days
- Peak demand range: min = 2,507.974 kW; max = 30,066.219 kW; mean = 14,441.215 kW; std = 6,322.575 kW.
- Monthly maximum peak values are tightly clustered: mean of monthly maxima = 28,855.67 kW, std ≈ 735 kW, indicating consistent high-load events each month
- Highest Annual peak: 30,066.219 kW at 14 November 12:30 — ~2.02x the mean (30,066 / 14,441).
- Top 25% of intervals have peak demand >= 20,051.816 kW (75th percentile).
- Median peak demand is 12,383.057 kW, indicating a right-skewed distribution (mean > median).
- Standard deviation (6,322.575 kW) shows substantial variability across 30-minute intervals—important for demand management and peak shaving strategies.
Let's further examine the Top 0.5% and 10% of the peaks by TOU category and the peaks for each month
The graph below shows the Top 90 (0.5%) demand intervals, color coded by Standard and Peak.
Insights
- Highest-demand interval (rank 1, 30,066.22 kW) is a Standard rather than Peak. High-demand intervals (top ranks) are often in Standard TOU windows, indicating high operational peaks can occur outside declared Peak windows.
- Standard intervals dominate the dataset (≈57% of all intervals) and also dominate the top 10% subset (1,139 of 1,752 ≈ 65%) and 48% of the top 90 demand intervals.
Like what you see here? Talk to us at contact@xanhterra.com
Analysis of Monthly Demand Maxima
What are the monthly peak demand maxima and their characteristics?
Insights
- The daily peak demand happens mostly in the afternoon (12:30 to 14:00) or a half-hour to an hour before the morning Peak (9:30 to 11:30).
- TOU Category for monthly maxima: Peak in 9 of 12 months, Standard in remaining months
- 11:00 is often the maximum peak period (see heatmaps below), and when the daily occurs in the middle of the night, it is only on Sundays (indicating the low usage of electricity on Sundays)
Let's further examine the timing of the peak demand by month and by day of the week.
Below are the AI Prompts, Execution Plan, and Progress Summary:
XanhTerra AI Prompt and Execution Plan for Heatmaps:
Visualization Results:
Like what you see here? Talk to us at contact@xanhterra.com
Baseline vs. Trial Tariff Cost Comparisons
Now, let's calculate the monthly and annual costs under the Baseline and Trial tariffs.
XanhTerra's AI tool made the calculations, and the comparisons are show in the graph below. The full analysis summary from the Agentic AI tool is also provided below to highlight its capabilties.
(The left axis is for monthly costs in billions of VND and the right axis is for annual costs (the last columns) in billions of VND)
XanhTerra AI Analysis Summary:
Insights
(Combination of AI analysis + Human Interpretation)
- Baseline average monthly total is 21.62 B VND (min 18.08, max 24.11), with an annual total cost of 259.41 B VND
- Baseline energy cost is more sensitive to TOU energy volumes in Peak periods; months with higher peak kWh show larger Baseline totals (e.g., March).
- EVN Trial average monthly total is 21.49 B VND (min 19.13, max 23.37), with an annual total cost of 257.83 B VND, a difference of only -0.6% annual.
- Monthly demand charges are stable in magnitude (mean 6.793 B VND/month) because monthly Pmax values vary within a limited range; demand contributes ~30–35% of total cost.
- Average monthly difference: -0.243 B VND (Trial tariffs slightly lower on average); however, the Trial tariff is higher in February, which had lower consumption than other months due to the Tet holidays.
Like what you see here? Talk to us at contact@xanhterra.com
Comparison across Voltage Connection Levels
What if the factory was connected to 110 kV or 6 kV voltage level, instead of 22 kV?
Below is XanhTerra's AI Thought Process (including consulting a Financial Expert) and a draft Plan
Results:
Insights
- Under Trial tariff, the factory would pay more only at the lowest-voltage case (Voltage below 6kV) by +2.40 B VND annually; all other voltage levels show annual savings under Trial (largest: -8.18 B VND at 6–22 kV).
- The higher cost for the Voltage below 6kV is driven by high demand charge, despite the energy-rate reductions.
- Demand charges constitute a material share of total Trial costs (approx 29%–35% depending on voltage).
- Recommendation: Target peak reduction measures (demand response, load-shifting) to reduce monthly Pmax: each kW reduced lowers monthly demand charge directly and materially reduces Trial total cost (demand share ~30%+).
Case Study:
Impact of Two-Component Tariffs for Manufacturing Facilities in Vietnam