An AI-supported energy management system (EMS) coordinates the PV system, battery storage and loads in real time, based on weather forecasts, a learned load profile and price signals. Industrial businesses thereby reach self-consumption ratios of 70-90% instead of 30-40% without storage, and save typically CHF 15'000-40'000 per year through peak shaving at a 300+ kW peak load. Since the Mantelerlass revision, certified EMS interfaces are mandatory for systems from 100 kWp.
Key Takeaways
10 min read
Key Takeaways
10 min read- An AI-supported EMS optimises storage, PV and loads in real time. Manual energy management cannot keep up with volatile feed-in tariffs and dynamic grid fees.
- With battery storage and an intelligent EMS, industrial businesses reach self-consumption ratios of 70-90%, compared with 30-40% without storage.
- Peak shaving through AI-controlled pre-discharge saves businesses with a 300+ kW peak load typically CHF 15'000-40'000 per year.
- Since the Mantelerlass revision, EMS systems must be certified and meet the grid connection requirements (NIV). Selection and configuration require technical know-how.
Modern industrial and commercial businesses have three energy sources to coordinate: the PV system, battery storage and the public grid. Manual energy management cannot optimise these three systems at the same time.
PV yield fluctuates with cloud cover and season, grid electricity prices vary by the hour, and the business’s own load peak decides the monthly demand charge. An intelligent energy management system (EMS) is the missing link between expensive hardware and maximum economic benefit. AI algorithms lift that benefit to a level rule-based controls do not reach.
What is an intelligent energy management system?
An EMS is the control intelligence that coordinates all energy sources and consumers of a business. Without an EMS, PV, storage, loads and grid connection work independently of one another.
With an EMS, the photovoltaic system, battery storage, building loads, EV charging stations, heat pumps and the grid connection become one integrated system. The shared goal: minimum energy costs at maximum self-supply.
Classic, rule-based EMS systems work with fixed thresholds: “store when PV surplus > X kW”, “discharge when grid load > Y kW”. That logic works for stable, predictable load profiles. In complex businesses with changing shifts, seasonal production cycles and external price signals, it fails.
AI- and ML-supported EMS systems go one decisive step further: they learn from historical data, produce short-term forecasts and adapt their control strategy dynamically. The difference is not gradual but qualitative. A rule-based system reacts to events. An AI-supported system anticipates them.
How AI algorithms optimise storage and PV
Four mechanisms form the heart of an AI-supported EMS.
Weather-based PV yield forecasting: The EMS processes weather data for the next 24-72 hours and calculates the expected PV yield. This forecast determines how much storage capacity must be reserved for self-consumption and how much can remain available for peak-shaving duties.
Load profile learning: Every business has a characteristic load pattern with production starts in the morning, midday peaks, night shifts or weekend operation. The EMS recognises these patterns, distinguishes between regular operating days and exceptions, and produces daily load forecasts. A manufacturing business with predictable shifts benefits particularly strongly from this mechanism.
Dynamic battery dispatch strategy: Based on the PV forecast, load forecast and current storage state, the EMS coordinates the storage’s charging and discharging strategy. The aim is not a simple “charge until full, discharge on demand” logic but a multi-stage optimisation: when to charge (PV surplus vs low tariff), when to discharge (peak shaving vs self-consumption), how much capacity to reserve for which duty?
Tariff arbitrage: Where dynamic electricity tariffs are available, the EMS can integrate grid price signals into the dispatch strategy. It charges the storage at low-price times and discharges it at high-price times. Without an EMS this function is simply not feasible.
The result of these mechanisms: the storage coordinates in real time and reaches a self-consumption ratio that manual or rule-based systems cannot.
Use case: peak shaving
Peak shaving is for many Swiss industrial businesses the economically most attractive EMS function, because it acts directly on the demand charge in the grid fee.
Most Swiss grid operators bill, alongside the energy price (Rp./kWh), a demand charge on the monthly peak load (CHF/kW). A single short power peak from a compressor start, a production start or charging infrastructure starting up at the same time can raise that monthly charge for good.
The AI EMS anticipates these peaks before they occur. Based on the learned load profile, it calculates in the hours before a forecast peak event whether, and how far, the storage should be pre-discharged. When the peak load occurs, the storage discharges exactly the necessary power into the site network. The measured peak load stays below the defined threshold, and the demand charge falls for good.
For a deeper analysis of the demand tariff system and concrete savings potential, see our guide Peak shaving for commerce and industry. The calculation basis and the comparison with alternative measures are fully documented there.
Use case: self-consumption optimisation
Self-produced solar electricity costs 8-12 Rp./kWh, grid electricity 20-30 Rp./kWh. Every kilowatt-hour consumed on site saves the difference.
Without storage and without an EMS, the self-consumption ratio of typical industrial and commercial businesses sits between 30-40%. The PV yield exceeds the midday load, and surpluses are fed in at 5-12 Rp./kWh.
The AI EMS raises that ratio through two mechanisms: first through the storage, which absorbs PV surpluses and shifts them in time; second through active load shifting. Flexible consumers such as air conditioning, process cooling and hot water preparation are switched preferentially into periods of high PV yield.
The result: self-consumption ratios of 70-90% are achievable with combined PV and storage systems and an intelligent EMS. An extreme case is BACHMANN GROUP with 96% self-consumption, though as a 24/7 manufacturing business with a constantly high base load it is a special case. For typical businesses with a daytime-dependent profile, 70-90% is a realistic target with an optimised system design.
How ZEV structures extend self-consumption optimisation to several properties or tenants is explained in our article Optimising self-consumption: ZEV, EMS and load shifting.
Use case: e-mobility and charging management
Electric vehicles bring high peak power when charging, but are more flexible in time than production processes.
A 40-tonne truck charges at a 400 kW DC charger in around 90 minutes. The charging station creates a power peak that, without storage, would overload the grid connection and drive up the demand charge.
The AI EMS coordinates charging management with the other system components: it prioritises charging from PV surpluses, controls the charging speed dynamically on the basis of the current storage state and the forecast PV yield, and caps charging peaks that would exceed the peak-shaving threshold. The storage acts as a buffer and enables powerful charging infrastructure without a proportional expansion of the grid connection.
For the detailed technical requirements of commercial charging infrastructure, from charging point sizing to grid-side integration, see our guide Charging infrastructure for commerce and industry.
Market overview: which EMS systems suit industry?
The EMS market for industrial applications is fragmented. For businesses with PV, storage, EV charging infrastructure and heat pumps, the choice is not trivial.
Battery manufacturers supply proprietary EMS with their products, software providers offer manufacturer-independent solutions, and some grid operators develop their own platforms.
Decisive criteria for industrial EMS systems:
Scalability: The system must reliably manage systems from 100 kWp and storage from 100 kWh. Simple home storage EMS solutions are not suited to this scale.
Open interfaces (API): An EMS that only communicates with hardware from the same manufacturer restricts manufacturer independence. Industrial EMS should support standardised protocols (Modbus, SunSpec, OCPP for charging management).
Grid code certification: As described in the regulatory note, EMS systems must meet the grid connection requirements. A current DSO approval is the precondition for smooth commissioning.
Multi-asset control: For businesses with PV, storage, EV charging infrastructure and thermal loads, the EMS must coordinate all assets at the same time, not optimise them in isolation.
Ampere Dynamic is a system integrator and selects the EMS per project. No single system is optimal in every constellation. What matters is the configuration and integration into the overall system. The comparison Battery storage vs grid expansion illustrates how the system design affects the economics.
Why AI-controlled energy management pays off
An AI-supported EMS is the precondition for PV and battery storage to realise their full economic potential.
The technology is mature, the regulatory requirements are defined, and the economic arguments are proven: 70-90% self-consumption, CHF 15’000-40’000 in peak-shaving savings per year, charging infrastructure without grid expansion.
The decisive question is not whether, but which EMS, and how it is integrated into the existing or planned energy system. That decision depends on the load profile, the system size, the grid operator and the business’s medium- and long-term energy plans.
As an entry point into the economics of commercial storage, we recommend the commercial storage in Switzerland hub. It summarises all revenue streams, payback scenarios and sizing recommendations. We offer complete EMS advice, including load profile analysis and system selection, as part of an initial consultation.