Method for energy optimization of a system

Abstract

The present invention relates to a method for energy optimization of a system (1), a computer program, a device (9) and a system (1).

Description

[0001] TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0002] Methods for optimizing the energy consumption of a plant

[0003] Technical field

[0004] The present invention relates to a method for optimizing the energy of a plant, a computer program, a device and a plant.

[0005] State of the art

[0006] Energy management systems and methods for energy optimization are known from the state of the art.

[0007] Description of the invention

[0008] One aspect concerns a method for optimizing the energy consumption of a plant.

[0009] The system can be a building, such as a single-family home, an apartment building, or an industrial facility. The system can include local energy consumers and local energy producers. The local energy producers can, for example, generate electricity that can be used by the local energy consumers. The method can provide optimized control of system components, whereby the optimized control can be adapted with respect to various optimization aspects. Energy optimization can describe the distribution and use of energy, determined according to a specific optimization aspect. The method can be computer-implemented. The steps of the method can be executed by a computer using appropriate computer software, where the computer can be a control unit or part of a control unit.

[0010] The procedure comprises a step of reading in a current electricity consumption parameter set. The procedure also comprises a further step of inputting this current electricity consumption parameter set as input signals into a trained first artificial intelligence model. This first artificial intelligence model was trained based on training data comprising combinations of a historical electricity consumption parameter set and a historical electricity consumption. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0011] and provides as input data an estimate of expected electricity consumption, including a temporal profile.

[0012] The reading process can encompass both acquisition and reception. Acquisition can involve data recording, for example, using a data acquisition device. The data acquisition device can be configured for the specific data to be acquired. Reception can involve data transmission, for example, from a storage unit, particularly a database. The term "reading" as a process step retains this meaning in the subsequent process steps as well.

[0013] The current electricity consumption parameter set can include data representative of current electricity consumption. The current electricity consumption parameter set can include relevant data for estimating expected electricity consumption. The historical electricity consumption parameter set can include data representative of historical electricity consumption. The data in the historical electricity consumption parameter set and the data in the current electricity consumption parameter set can correspond to each other but have been collected at different times. The historical electricity consumption may have been available at the time the data for the historical electricity consumption parameter set was collected. The current electricity consumption parameter set, the historical electricity consumption parameter set, and the historical electricity consumption can each include a time series.The expected electricity consumption can describe a future electricity consumption that is expected based on the current electricity consumption parameter set.

[0014] The procedure includes a further step of determining a prediction of the expected electricity consumption, encompassing a temporal profile, based on the initial data of the trained first artificial intelligence model.

[0015] Determining the prediction can involve the input data of the trained artificial intelligence model. Determining the prediction can also include a further step of processing the input data of the trained artificial intelligence model. Post-processing can, for example, include making the determined prediction of expected electricity consumption usable for subsequent process steps. This can apply to all predictions. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0016] These values ​​are determined based on the initial data of a trained artificial intelligence model.

[0017] The procedure includes a further step of reading in a current set of electricity generation parameters. This step involves inputting the current set of electricity generation parameters as input signals into a trained second artificial intelligence model. This second artificial intelligence model was trained on training data comprising combinations of historical electricity generation parameters and historical electricity generation data. The output data provided is an estimate of expected electricity generation, including a time-dependent profile.

[0018] The current electricity generation parameter set can include data representative of current electricity generation. The current electricity generation parameter set can include relevant data for estimating expected electricity generation. The historical electricity generation parameter set can include data representative of historical electricity generation. The data of the historical electricity generation parameter set and the data of the current electricity generation parameter set can correspond to each other but have been collected at different times. The historical electricity generation can have existed at the time the data for the historical electricity generation parameter set was collected. The current electricity generation parameter set, the historical electricity generation parameter set, and the historical electricity generation can each include a time series.The expected electricity generation can describe a future electricity generation that is expected based on the current electricity generation parameter set.

[0019] The procedure includes a further step of determining a forecast of the expected electricity generation, encompassing a temporal profile, based on the input data of the trained second artificial intelligence model. The procedure includes a further step of inputting the forecast of expected electricity generation, the forecast of expected electricity consumption, and a set of energy system parameters as input data into a determination model. The procedure includes a further step of determining, for an energy utilization scenario of the plant to be optimized, a temporal profile of control variables. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0020] the system for controlling the use of available energy as an output of the determination model.

[0021] The energy system parameter set can be representative of the plant's current energy state. This set can incorporate geographic data, which can include two-dimensional or three-dimensional location data. The geographic data can describe the plant's location or the location of its components, such as electricity consumers or generators. This allows the location of the plant or its components, such as electricity consumers or generators, to be considered during data acquisition, processing, and forecasting. Location can be a relevant influencing factor for the plant and / or its individual components. Geographic data can be particularly relevant when using a decentralized computing system.

[0022] The determination model can be configured to determine the temporal profile of the system's control variables based on the input data. The system's energy utilization scenario to be optimized can define an optimization objective. This scenario can describe which usage patterns, at what times, and to what extent are optimal for achieving the optimized objective. The energy utilization scenario can be specified by a user at the system and / or centrally. The system can have an input device for this purpose. For example, the energy utilization scenario can be selected from a predefined list via the input device. The process can include further steps such as reading the energy utilization scenario to be optimized and entering it into the determination model.The data import and input steps can be performed in a single process step. The energy utilization scenario of the system to be optimized can be adjusted over time. The energy utilization scenario of the system to be optimized can be the selected optimization aspect. The determination model can be configured to achieve the optimization objective as closely as possible. The energy utilization scenario of the system to be optimized can encompass an optimization period. The optimization period can be limited in duration. For example, the optimization period could be 24 hours or 48 hours. The time course of the estimates can be adapted to the optimization period. The time course of the estimates can correspond to the optimization period. DerTQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026.

[0023] The optimization period can be defined by the user at the system and / or centrally. The time course can comprise a sequence of discrete values, particularly time-discrete values, over a period of time, especially the optimization period. The sequence of discrete values ​​can form a continuous curve over the period, especially the optimization period. The time course enables optimized adjustment of the system's control variables over the time course. The values ​​that form the time course can vary over the time course, especially the optimization period.

[0024] The system's control variables can be adjustable parameters that regulate the use of available energy. These control variables can describe the distribution and use of energy over time. The control variables can be adjusted according to the energy utilization scenario of the system to be optimized. The process can include a further step of controlling the system based on the defined control variables. This control can encompass the distribution of available energy usage.

[0025] This enables a process that, based on an adjustable optimization aspect, provides optimal settings for the control variables. This can reduce and / or optimize the energy consumption of the system. The system can thus be operated more energy-efficiently.

[0026] In one embodiment, the system can comprise a means for drawing energy from the power grid, a means for feeding energy into the power grid, a switchable load, and a power generator. The system can preferably include a unidirectional or bidirectional charging station for charging electric vehicles and / or an energy storage device.

[0027] The means of obtaining energy from the electricity grid can be configured to draw electricity from the public electricity grid. The means of obtaining energy from the electricity grid can have a grid connection. The means of feeding energy into the electricity grid can be configured to feed electricity into the public electricity grid. The means of feeding energy into the electricity grid can have an inverter and a grid connection. The switchable load can be an electronic device that consumes electrical energy. The switchable load can be one or more devices connected to a heat pump, a TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0028] The switching device can include a washing machine, a clothes dryer, and a dishwasher. Several such devices can be connected. The power generator can include a photovoltaic (PV) system and / or a wind turbine. The PV system can consist of multiple solar cells. The PV system is designed to generate electricity from solar energy. The wind turbine can be designed to generate electricity from the kinetic energy of the wind. The wind turbine can be, for example, a windmill. The power generator can also include several such systems. The charging station can include multiple charging points. The charging stations can have multiple connections for electric vehicles. The charging station can be an EV charger.A unidirectional charging station can be configured to supply electricity to an electric vehicle and charge its battery. A bidirectional charging station can be configured to supply electricity to an electric vehicle and draw electricity from its battery for other uses. In a unidirectional charging station, charging the electric vehicle can be considered switching a switchable load on or off. An energy storage system can be configured to store electricity and release it at a later time. The energy storage system can include a battery. It can be connected to switchable loads, the charging station, the means of drawing power from the grid, and power generators.The system can also include a heat connection to the public heating network, for example, the public district heating network. This allows the system to also be supplied with heat energy from the public heating network.

[0029] The method can include a further step of determining a prediction of expected charging station usage. The prediction of expected charging station usage can include the charging station's power demand and / or the available power from the electric vehicle connected to the charging station. The method can include a further step of inputting the prediction of expected charging station usage into the determination model. The prediction of expected charging station usage can be determined based on output data from a tenth artificial intelligence model. The method can include a step of reading in a current charging station parameter set. The method can include a further step of inputting the current charging station parameter set as input signals into a trained tenth artificial intelligence model, where the tenth artificial intelligence model is TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0030] An intelligence model was trained on training data comprising combinations of historical charging station parameter sets and historical charging station usage, and provided as output data an estimate of expected charging station usage, including a temporal trend. The method can include a further step of inputting the predicted charging station usage as input data into a determination model.

[0031] The current charging station parameter set can be representative of expected charging station usage. The current charging station parameter set can take calendar data and / or vehicle information into account.

[0032] The historical charging station parameter set can be representative of historical charging station usage. This set can incorporate calendar data and / or vehicle information. Vehicle information can include one or more parameters, such as the electric vehicle's total battery capacity, current battery capacity, or vehicle availability. Availability can be specified by a vehicle owner, for example. Vehicle information can also include vehicle status information. This information, particularly vehicle status, can be accessed and processed via a vehicle interface, including the internet. The electric vehicle's total battery capacity can represent the maximum possible energy that the electric vehicle, specifically its battery, can absorb.The current battery capacity can be representative of the energy currently stored in the electric vehicle's battery. The vehicle's availability can be representative of the period during which the electric vehicle is connected to the charging station.

[0033] In one embodiment, the current power consumption parameter set can include a value for the current power consumption and current power consumption data from the power consumers. The current power consumption data can include time data, calendar data, weather data, and / or weather forecast data.

[0034] The current electricity consumption value can quantify current electricity consumption. The current electricity consumption value can include a time series of current electricity consumption. The current electricity consumption data can include data from electricity consumers and / or time series of current electricity consumption. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0035] The data includes energy consumption data from electricity consumers. For example, this data can allow for the unambiguous assignment of current electricity consumption to specific consumers. The time series of current electricity consumption data can include a time series for each consumer. The current electricity consumption parameter set can incorporate geographical data. For example, the data from electricity consumers can include geographical information. Time data can include a time of day. Calendar data can include a day of the week, a weekend, a holiday, or a value representative of a season. Weather data can include one or more values, such as temperature, solar irradiance, or wind speed.Weather forecast data can be weather data that includes a forecast. Weather data and / or weather forecast data can be particularly relevant when using a heat pump.

[0036] The historical electricity consumption parameter set can include one or more historical electricity consumption values ​​and historical electricity consumption data from electricity consumers. The historical electricity consumption data can include time data, calendar data, weather data, and / or weather forecast data.

[0037] One or more historical electricity consumption values ​​can represent a trend over time. One or more historical electricity consumption values ​​can quantify past electricity consumption. Multiple historical electricity consumption values ​​can quantify electricity consumption over a past period. One or more historical electricity consumption values ​​can include time series of historical electricity consumption data from electricity consumers. Historical electricity consumption data can include data from electricity consumers. The historical electricity consumption parameter set can, for example, allow for a unique assignment of historical electricity consumption to individual electricity consumers.The time series of historical electricity consumption data for each electricity consumer can include a time series with historical electricity consumption for each consumer. The historical electricity consumption parameter set can take geographical data into account. For example, the data for electricity consumers can include geographical data. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026.

[0038] This embodiment allows for a more precise subdivision of the current and historical electricity consumption parameter sets, thereby improving the accuracy of estimating expected electricity consumption. Furthermore, it enables more precise determination of the control variables to better match the energy utilization scenario of the plant being optimized.

[0039] In one embodiment, the current power generation parameter set can include a current power generation value and current power generation data from PV energy and / or current power generation data from wind energy. The current power generation data from PV energy and / or the current power generation data from wind energy can take into account one or more of the time of day, a value representative of a season, a value representative of a solar position trend, weather data, or weather forecast data.

[0040] PV energy can describe photovoltaic energy. PV energy can be generated by converting sunlight into electricity. Wind energy can be generated by converting the kinetic energy of the wind into electricity.

[0041] The value of current electricity generation can quantify current electricity generation. The value of current electricity generation can also encompass a time-based progression of current electricity generation.

[0042] The current electricity generation data for PV energy and / or the current electricity generation data for wind energy can include data from electricity producers and / or time series of current electricity generation from the electricity producers. The electricity producer data can, for example, allow for a unique attribution of current electricity generation to the electricity producers. The current electricity generation parameter set can take geographical data into account. For example, the electricity producer data can include geographical data. The time of day can include a specific time. The weather data can include one or more values, such as temperature, solar irradiance, or wind speed. Weather forecast data can be weather data that includes a forecast. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0043] Weather data and / or weather forecast data can be taken into account, especially when using a heat pump.

[0044] The historical electricity generation parameter set can include one or more historical electricity generation values ​​and historical electricity generation data for PV energy and / or historical electricity generation data for wind energy. The historical electricity generation data for PV energy and / or the historical electricity generation data for wind energy can include one or more values ​​from a specific time of day, a value representative of a season, a value representative of a solar position trend, weather data, or weather forecast data.

[0045] One or more values ​​of historical electricity generation can represent a time series. One or more values ​​of historical electricity generation can quantify past electricity generation. Multiple values ​​of historical electricity generation can quantify electricity generation over a past period. One or more values ​​of historical electricity generation can include time series of historical electricity generation data from power producers. Historical electricity generation data for PV energy and / or historical electricity generation data for wind energy can include data from power producers. The historical electricity generation parameter set can, for example, allow for a unique assignment of historical electricity generation to power producers. The historical electricity generation parameter set can take geographical data into account.For example, data from electricity producers can include geographical data. The time of day can include a specific time. Weather data can include one or more values ​​such as temperature, solar radiation, or wind speed. Weather forecast data can be weather data that includes a forecast. Weather data and / or weather forecast data can be particularly important when using a heat pump.

[0046] In one embodiment, the first artificial intelligence model can be a neural network. Preferably, the neural network can be a Long Short-Term Memory (LSTM), a Recurrent Neural Network (RNN), a Gated Recurrent Unit (GRU), and / or a Convolutional Neural Network (CNN). The second artificial intelligence model can be a neural network. Preferably, the neural network is a Long Short-Term Memory (LSTM), a Recurrent Neural Network (RNN), a Gated TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0047] A Recurrent Unit (GRU) and / or a Convolutional Neural Network (CNN). The determining model can be a third artificial intelligence model, a statistical model using statistical and / or stochastic methods, or a combination of a third artificial intelligence model and a statistical model using statistical and / or stochastic methods.

[0048] The third artificial intelligence model can be configured to determine, based on input data, a time-dependent profile of the plant's control variables for an energy utilization scenario to be optimized. This third artificial intelligence model can be a neural network. The neural network can incorporate reinforcement learning, a long short-term memory (LSTM), a recurrent neural network (RNN), a gated recurrent unit (GRU), and / or a convolutional neural network (CNN). The statistical model, employing statistical and stochastic methods, can include a combined application of the neural network, simulation-based optimization, and stochastic methods, such as Monte Carlo simulation.When used in combination, the step of determining the temporal profile of control variables can include the steps of identifying possible temporal profiles of the plant's control variables and determining, for an energy utilization scenario to be optimized, a temporal profile of the plant's control variables based on the identified possible temporal profiles as output of the determination model. This allows various possible energy utilization scenarios to be simulated and evaluated, with the one that most closely matches the objective being selected. Using a combination of the third artificial intelligence model and the statistical model can increase the prediction accuracy.

[0049] The process can include a further step of reading the time course of the manipulated variables as training data for the third artificial intelligence model. This enables continuous training of the third artificial intelligence model.

[0050] Long Short-Term Memory (LSTM) can provide efficient processing of time series. Recurrent Neural Networks (RNNs) can incorporate the contents of previous inputs. Gated Recurrent Units (GRUs) may require less computational effort but may have memory requirements. Convolutional Neural Networks (CNNs) can be efficiently adapted to new tasks and datasets. When using a Convolutional Neural Network, time series can be read like a one-dimensional image. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0051] In one embodiment, the energy system parameter set may include one or more values ​​of the current electricity consumption, the current electricity generation, a consumer value, or a state-of-charge value.

[0052] The consumer value can be representative of the presence of switchable consumers. For example, the consumer value can include one or more values ​​representing the presence of a heat pump, a washing machine, a tumble dryer, or a dishwasher.

[0053] The state of charge (SBC) value can represent the state of charge of the energy storage system. The SBC can vary over time. It can increase or decrease over time. The SBC value can represent the state of charge of the electric vehicle's battery. The SBC value can be a value between 0 and 100%, where 100% can represent the maximum charge of the energy storage system.

[0054] This allows additional parameters to be included as input data in the model and thus considered when determining the control variables. This can improve the quality of the model and determine control variables that are closer to an optimal solution for the energy utilization scenario being optimized.

[0055] In one embodiment, the control variables can include one or more from an energy draw from an electricity storage device, an energy charging of the electricity storage device, a use of switchable loads, an energy charging of an electric vehicle, an energy draw from the electric vehicle, an energy draw from the electricity grid or an energy feed-in to the electricity grid.

[0056] Energy consumption from the energy storage system can describe when and how much energy should be drawn from the storage system. Energy charging of the energy storage system can describe when and how much energy should be stored in the storage system. The use of switchable loads can describe when and which switchable load should be used. The use of switchable loads can encompass a variety of usage values ​​for the switchable loads, for example, for each switchable load, a TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0057] A usage value as a control variable, or a functional or spatial grouping of switchable loads with a usage value as a control variable per group. Charging the electric vehicle can describe when and how much energy should be supplied to the electric vehicle. Drawing energy from the electric vehicle can describe when and how much energy should be drawn from the electric vehicle. Charging and drawing energy from the electric vehicle can be linked to the number of charging stations, the number of connections for electric vehicles, or the number of electric vehicles. Drawing energy from the grid can describe when and how much energy should be drawn from the public grid. Feeding energy into the grid can describe when and how much energy should be fed into the public grid. The control variable can include drawing heat energy from the public heating network.This provides various possibilities for controlling the distribution and use of energy.

[0058] In one embodiment, the method can include a further step of reading in the current electricity consumption parameter set as training data for the first artificial intelligence model. Alternatively or additionally, the method can include a further step of reading in the current electricity generation parameter set as training data for the second artificial intelligence model.

[0059] These steps ensure the continuous training of both the first and second artificial intelligence models with current data. This leads to faster adaptation to changing plant conditions. Furthermore, this enables both the first and second artificial intelligence models to provide more accurate estimates.

[0060] In one embodiment, at least some of the control variables can be managed and mapped by agents that are part of a multi-agent control system. Each agent can have the ability to achieve defined goals through interaction with an environment and other agents.

[0061] An agent can control exactly one variable to achieve the defined goals. An agent can control multiple variables to achieve the defined goals. Multiple agents can control the same variable to achieve the defined goals. For this purpose, a higher-level agent (TQ-Systems GmbH WBH-Ref: 343.0029WO, March 26, 2026) can be used.

[0062] Decide which agent has priority for controlling the manipulated variable. The environment can be a real-world plant environment. At least one of the agents can be a mechanical agent. Multi-agent control can include one or more agents such as an electricity storage agent, an electricity consumer agent, a grid monitoring agent, a renewable energy agent, a coordinator agent, or an energy feed-in agent. The agents can be communicatively connected to each other. The agents can communicate and interact with each other to achieve their objectives. The defined objectives can be specified wholly or partially by a user. For example, the defined objectives can be specified by the coordinator agent. The defined objectives can be adjusted over time. The coordinator agent can be the superior agent.The coordinator agent can assume a higher-level task and pursue an overarching plant objective. The electricity storage agent can have the objective of optimizing the use of electricity storage. The electricity consumer agent can have the objective of optimizing electricity consumption. The electricity consumer agent can include one agent for each switchable load or one agent for a group of switchable loads, with the grouping of switchable loads preferably occurring after prioritization. The grid monitoring agent can have the objective of optimizing the use of electricity drawn from the public grid. The renewable energy agent can have the objective of optimizing the use of electricity generation. The renewable energy agent can include a PV agent for the PV systems and a wind power agent for the wind turbines.The renewable energy agent can include one agent for each PV and wind power plant. The feed-in agent can aim for optimized use of electricity fed into the public grid.

[0063] A multi-agent controller can provide parallel processing of tasks and thus operate efficiently. This can improve the efficiency of the plant. Furthermore, control variables and estimations can be determined precisely, and the controller can be effectively adapted to new conditions.

[0064] In one embodiment, the energy utilization scenario to be optimized for the plant can be a maximization of self-consumption of electricity, an optimization of net energy consumption, an optimization of CO2 emissions, an optimization of net resource utilization, a TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0065] a weighted combination of optimizing net energy consumption, CO2 emissions, and net resource use.

[0066] Maximizing self-consumption of electricity can involve prioritizing the use of generated electricity. Maximizing self-consumption can also involve minimizing energy drawn from the public grid. Net energy consumption can describe self-generated energy, specifically self-generated electricity and / or heat. Self-generated electricity can include the electricity generated by the photovoltaic system and / or wind power plant. It can also include energy generated by a dispatchable generator. Self-generated heat can include energy generated by a dispatchable heat generator. Optimizing net energy consumption can involve optimizing self-generated energy. Optimizing net energy consumption can also consider the use of the dispatchable generator and / or the dispatchable heat generator.Optimizing net energy consumption can take into account the consumption and / or generation of heat energy. Optimizing net energy consumption can provide for the optimized use of available energy. This can result in a lower reliance on electricity and / or heat from the public grid.

[0067] Net resource utilization optimization can assign a resource value to each instance of energy consumption from the public grid and each instance of energy feed-in, particularly electricity. This optimization can consider the current electricity consumption value and / or the current feed-in balancing. Net resource utilization can also assign a net resource value to the controllable power generator and / or the controllable heat generator for providing electricity and / or heat, as described below. Optimizing net resource utilization can also minimize net resource consumption, specifically the net resource value. Net resource utilization can be negative if the resource value of the feed-in is greater than the resource value of the consumption. This can optimize technical resource utilization and increase system capacity.

[0068] Optimizing CO2 emissions can involve minimizing total CO2 resource input or minimizing CO2 emissions. Total CO2 resource input can be representative of the accumulated resource input for a total amount of CO2 emitted, for example, in tonnes of CO2. Total CO2 - TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0069] Resource input can be determined based on CO2 emissions, for example in tons, and a CO2 resource input, for example, the resource input per ton of CO2 emitted. This allows CO2 emissions to be optimized, and in particular minimized. The CO2 resource input can have a fixed value. The CO2 resource input can be provided by an external source. The CO2 resource input can have a fixed value for a specific period. The CO2 resource input can have a variable value.

[0070] The weighted combination can weight all optimizations equally. Alternatively, the weighted combination can weight certain optimizations more or less heavily than others. Alternatively, the energy utilization scenario of the plant to be optimized can also include a weighted combination of other optimizations.

[0071] The procedure can additionally include the following steps: Reading in a current CO2 emission parameter set. Inputting the current CO2 emission parameter set as input signals into a trained fourth artificial intelligence model, where the fourth artificial intelligence model was trained based on training data comprising combinations of a historical CO2 emission parameter set and a historical CO2 emission value, and provides an estimate of expected CO2 emissions as output data. Determining a prediction of expected CO2 emissions, including a time course, based on the output data of the trained fourth artificial intelligence model. Inputting the prediction of expected CO2 emissions as input data into the determination model. This allows for a prediction of expected CO2 emissions, enabling precise optimization of CO2 emissions.

[0072] The current CO2 emission parameter set can include a current CO2 emission value and current CO2 emission data. The current CO2 emission data can include current energy consumption and current energy production of the dispatchable power generator and current energy production of the dispatchable heat generator. The electricity generation of the PV system and the wind turbine can be considered CO2-neutral. The current CO2 emission data can include one or more weather data points, weather forecast data, a value representative of a season, or a value representative of a solar path. Preferably, the weather data and the TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0073] Weather forecast data includes a temperature value, a solar radiation value and / or a wind speed value.

[0074] The historical CO2 emission parameter set can include historical CO2 emission data. This data can include historical energy consumption and output from the dispatchable power generator and historical energy output from the dispatchable heat generator. The historical CO2 emission data can incorporate one or more weather data points, weather forecast data, a value representative of a season, or a value representative of a solar position trend. Preferably, the weather data and weather forecast data include a temperature value, a solar irradiance value, and / or a wind speed value.

[0075] Alternatively or additionally, the procedure can include a step of determining a forecast of expected CO2 emissions based on the expected energy consumption of local energy consumers and / or the expected energy consumption of local energy producers to supply the energy, and an emission factor. The emission factor can depend on the energy source used to supply the energy, for example, in kg CO2 per kWh. The energy source could be, for example, gas, liquid fuel, or coal. The emission factor can also be provided externally, particularly if energy for supplying local energy consumers is drawn from the public grid. The generated energy, which is stored, for example, in the electricity storage system or the electric vehicle and then drawn from it again, is only considered once in the forecast of expected CO2 emissions.This can occur during the generation and storage of energy or the consumption of energy through switchable loads or feed-in.

[0076] The procedure can include a further step of entering the forecast of the expected CO2 emissions as input data into the determination model.

[0077] CO2 emissions can thus be directly determined from the use of the energy sources relevant to CO2 emissions. When determining the energy usage scenario to be optimized, the expected CO2 emissions of local energy consumers and / or the expected CO2 emissions of local energy producers can therefore be taken into account. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0078] Using a prediction of expected CO2 emissions via the fourth artificial intelligence model and a forecast of expected CO2 emissions based on expected energy consumption can, for example, be used to compare one of the two values ​​and thus improve the accuracy of the expected CO2 emissions. This allows for precise optimization of CO2 emissions.

[0079] The multi-agent control system can additionally include a CO2 emissions agent. The CO2 emissions agent can have the goal of optimizing CO2 emissions.

[0080] In one embodiment, the method can include the step of reading the current CO2 emission parameter set as training data for the fourth artificial intelligence model. This ensures continuous training of the fourth artificial intelligence model based on current data.

[0081] The procedure can include the following steps: Reading in a current CO2 resource input parameter set. Inputting the current CO2 resource input parameter set as input signals into a fifth trained artificial intelligence model, where the fifth artificial intelligence model was trained based on training data comprising combinations of a historical CO2 resource input parameter set and a historical CO2 resource input, and providing an estimate of expected CO2 resource input as output data. Determining a prediction of expected CO2 resource input, including a time course, based on the output data of the trained fifth artificial intelligence model. Inputting the prediction of expected CO2 resource input as input data into the determination model.

[0082] The current CO2 resource input parameter set can be representative of the expected CO2 resource input. The current CO2 resource input parameter set can include current CO2 resource input and current CO2 resource input data. The current CO2 resource input data can include one or more weather data points, weather forecast data, a value representative of a season, or a value representative of a solar position trend. The weather data and weather forecast data can preferably include temperature, solar irradiance, and / or wind speed. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0083] The historical CO2 resource input parameter set can be representative of historical CO2 resource input. This parameter set can include historical CO2 resource input data. The historical CO2 resource input data can include one or more weather data points, weather forecast data, a value representative of a season, or a value representative of a solar position trend. The weather data and weather forecast data can preferably include temperature, solar irradiance, and / or wind speed.

[0084] The process can include the step of reading the current CO2 resource input parameter set as training data for the fifth artificial intelligence model. This enables continuous training of the fifth artificial intelligence model based on current data. This ensures continuous training of the fifth artificial intelligence model.

[0085] Multi-agent control can include a CO2 resource utilization agent. The CO2 resource utilization agent can aim for optimized use of CO2 resources.

[0086] By taking into account the forecast of expected CO2 resource consumption, the process can be optimized with regard to overall CO2 resource use. This allows, in particular, the future tradable CO2 emission allowances to be considered in the plant's optimization.

[0087] In one embodiment, the method can include a further step of reading in a current consumption value and / or reading in a feed-in compensation value. The current consumption value can include a current consumption value and preferably a prediction of an expected consumption value. The prediction of the expected consumption value can preferably be based on output data from a trained sixth artificial intelligence model. The feed-in compensation can include a current feed-in compensation value and preferably a prediction of an expected feed-in compensation value. The prediction of the expected feed-in compensation can preferably be based on output data from a trained seventh artificial intelligence model. The method can include a further step of inputting TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0088] the electricity reference value and / or the feed-in compensation as input data into the determination model.

[0089] The electricity reference value can be a value representing the resource expenditure for drawing electricity from the public grid. The electricity reference value can define a specific value for the consumption of one unit of energy, for example, 1 kWh. The feed-in compensation can be a value representing the resource expenditure for feeding electricity into the public grid. The feed-in compensation can define a specific value for feeding in one unit of energy, for example, 1 kWh. The electricity reference value and / or the feed-in compensation can each have a fixed value. The fixed value can be a day-ahead value, defined the day before for the following day. The day-ahead value can be a fixed value, for example, on an hourly basis. The fixed value can be provided by an external source.The current electricity consumption value and / or the current feed-in compensation can each have a fixed value for a fixed period. This fixed period could, for example, be one hour. The fixed value could then be updated hourly. The fixed period can be shorter, longer, or equal to the optimization period. Alternatively, the fixed period can be so short that the current electricity consumption value and / or the current feed-in compensation can each have a variable value.

[0090] If the optimization period is longer than the fixing period, the forecast of the expected electricity consumption and the forecast of the expected feed-in compensation can be considered to determine optimal control parameters for the optimization period. Alternatively, the forecast of the expected electricity consumption and / or the forecast of the expected feed-in compensation can be considered independently of the fixing period.

[0091] The procedure may include a step of checking whether the optimization period is longer than the fixing period. If the optimization period is longer than the fixing period, the procedure may include the following steps. Alternatively, the procedure may include the following steps regardless of the result of the checking step: Reading in a current reference value parameter set. Inputting the current reference value parameter set as input signals into the trained sixth artificial intelligence model, wherein the sixth artificial intelligence model was trained based on training data which combinations of each historical TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0092] The process includes a set of power consumption parameter values ​​and a historical power consumption value, and provides an estimate of an expected power consumption value as input data. It determines a prediction of the expected power consumption value, including a time course, based on the output data of the trained sixth artificial intelligence model. The prediction of the expected power consumption value is then fed into the determination model as input data. This allows for a prediction of the expected power consumption value, which can enable the precise determination of the control variables.

[0093] The current electricity reference value parameter set can be representative of the expected electricity reference value. The current electricity reference value parameter set can include a current electricity reference value. The current electricity reference value parameter set can include current electricity reference value data. The current electricity reference value data can include one or more of the following: time data and calendar data, weather data in a relevant location area of ​​the electricity grid (preferably temperature, solar irradiance, and wind speed), a value representative of the season, or a value representative of the sun's position.

[0094] The historical electricity reference value parameter set can be representative of the historical electricity reference value. The historical electricity reference value parameter set can include historical electricity reference value data, wherein the historical electricity reference value data includes one or more of the following: time data and calendar data, weather data in a relevant location area of ​​the electricity supply network, preferably temperature, solar irradiance and wind speed, a value representative of the season, or a value representative of the course of the sun's position.

[0095] The procedure can include a step of checking whether the optimization period is longer than the fixing period. If the optimization period is longer than the fixing period, the procedure can include the subsequent steps. Alternatively, the procedure can include the following steps regardless of the result of the checking step: Reading in a current feed-in balancing parameter set. Inputting the current feed-in balancing parameter set as input signals into the trained seventh artificial intelligence model, wherein the seventh artificial intelligence model was trained based on training data comprising combinations of historical feed-in balancing parameter sets and a historical feed-in balancing. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0096] The system provides an estimate of expected feed-in compensation as input data. It then determines a prediction of this expected feed-in compensation, including a time-dependent progression, based on the input data of the trained seventh artificial intelligence model. This prediction is then fed into the determination model as input data. This allows for the precise determination of the control variables.

[0097] The current feed-in balancing parameter set can be representative of the expected feed-in balancing. The current feed-in balancing parameter set can include a value of current feed-in balancing. The current feed-in balancing parameter set can include current feed-in balancing data. The current feed-in balancing data can include one or more of the following: time data and calendar data, weather data in a relevant location area of ​​the power grid (preferably temperature, solar irradiance, and wind speed), a value representative of the season, or a value representative of the sun's position.

[0098] The historical feed-in balancing parameter set can be representative of historical feed-in balancing. The historical feed-in balancing parameter set can include historical feed-in balancing data, wherein the historical feed-in balancing data takes into account one or more of the following: time data and calendar data, weather data in a relevant location area of ​​the power grid, preferably temperature, solar irradiance and wind speed, a value representative of the season, or a value representative of the course of the sun's position.

[0099] Using a central computing system allows for the prediction of expected electricity consumption and feed-in compensation for all plant components connected to the central system, and even for different plants. This can save computing power and increase plant efficiency.

[0100] The procedure can include a step of reading the current electricity reference value parameter set as training data for the sixth artificial intelligence model. The procedure can also include a step of reading the current feed-in balancing parameter set as training data for the seventh artificial intelligence model. This enables continuous training of the TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0101] sixth and seventh artificial intelligence models will be provided based on current data.

[0102] The multi-agent control system can additionally include a feed-in balancing agent and / or an electricity purchase value agent. The feed-in balancing agent can aim for optimized use of feed-in compensation. This could, for example, be maximizing the compensation received for feeding electricity into the public grid. The electricity purchase value agent can aim for optimized use of electricity purchased from the grid. This can take into account when the electricity purchase value is expected to be low or high and adjust usage accordingly.

[0103] Considering the electricity consumption value and feed-in balancing can also account for the possibility of charging an electricity storage system with grid-generated electricity when the current electricity consumption value is low, and conversely, feeding energy from the storage system into the grid, particularly during peak load times. Furthermore, this allows for the consideration of additional input data. This can refine the determination of the control variables, leading to improved adaptability of the method to different scenarios. Additionally, using the sixth artificial intelligence model, an expected electricity consumption value, specifically the specific value for energy consumption per unit of energy, can be determined. Similarly, using the seventh artificial intelligence model, an expected feed-in balancing, specifically the specific value for feed-in per unit of energy, can be determined.This can provide an optimization opportunity, especially if the optimization horizon is larger than the fixing period, for example, of the day-ahead value.

[0104] In one embodiment, the system can include the controllable generator. The method can include a further step of reading in an electricity consumption value of the controllable generator. The method can also include a further step of inputting the electricity consumption value of the controllable generator as input data into the determination model. The control variables can include a utilization value of the controllable generator.

[0105] The controllable generator can be the plant's local generator. The controllable generator can comprise one or more controllable generators. The controllable generator can supply the plant with additional power by drawing from the energy source. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0106] Provide. The controllable generator can be, for example, a generator set based on gas, liquid fuel, or other energy sources. The local generator can be part of a combined heat and power (CHP) system. The electricity consumption value can include an electricity operating resource for using the controllable generator. The electricity operating resource can describe resource input for operating the controllable generator. The electricity operating resource can consider the use of the energy source of the controllable generator. The utilization value of the controllable generator can, for example, define a utilization rate of 0 to 100%. The utilization value of the controllable generator can, for example, be an electricity generation value in kWh.

[0107] The multi-agent control system can additionally include a controllable generator agent. The controllable generator agent can aim for optimized utilization of the controllable generator.

[0108] In one embodiment, the method can include a further step of allocating a first resource input for energy consumption. The method can also include a further step of allocating a second resource input for CO2 emissions. The energy utilization scenario to be optimized can involve minimizing a combination of the first and second resource inputs.

[0109] The first resource input can describe resource input for energy consumption. The first resource input can include net resource input. The second resource input can describe resource input for CO2 emissions. The second resource input for CO2 emissions can include CO2 resource input. Assigning the first resource input can involve entering the first resource input as input data into the determination model. Assigning the second resource input can involve entering the second resource input as input data into the determination model.

[0110] This allows the energy usage scenario to be optimized to include a combined optimization and minimize overall resource consumption. Alternatively, the TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0111] The energy utilization scenario includes minimizing either the first or second resource input.

[0112] This embodiment can also include the steps of reading and entering the power consumption value and / or the feed-in compensation. Furthermore, according to this embodiment, the system can also include the controllable power generator and the associated steps.

[0113] In one embodiment, the method can include a further step of obtaining a user input. The user input can be configured to specify a boundary condition for the control variables.

[0114] The user specification can be entered as input data into the determination model. The user specification can be entered via a human-machine interface. The human-machine interface can include an input device for entering the user specification. The input device for entering the user specification can also be the input device for entering the energy utilization scenario to be optimized. The input device can, for example, be a touchscreen. The input device can be located at various points within the system. The input device can be movable. The human-machine interface can include a receiver for receiving the user specification. The human-machine interface can include a device for processing the user specification. The human-machine interface can be partially implemented by an application.The application can be run on a smartphone, tablet, or other device, particularly a mobile device. The human-machine interface can also be implemented via a computing unit connected to, for example, the switchable loads, and derives a user input from the interaction with and / or use of the switchable loads. The boundary condition can be temporal and / or functional. Temporally, the boundary condition can refer to a time of day, a calendar day, or a time period. Functionally, the boundary condition can refer to a requirement, suitability, or property. The user input can include direct user input. Direct user input can be input from a user, for example, via the input device, where the input defines a direct relationship to a boundary condition. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026.

[0115] Direct user input can include, for example, a time or period for the use of a consumer. User input can include entering the boundary condition and an associated time range. The boundary condition can define the availability of a system component, preferably the switchable consumer or the controllable heat generator. The availability of a system component can define a period during which the component can be used. The boundary condition can specify heating and / or cooling of a room within the system. The energy system parameter set can include a value representative of the current opening position of a window. The control variables can include a value representative of a window opening position to be adjusted. This allows a window to be controlled, for example, to comply with ventilation requirements and also for temperature control.User input can be used to modify an existing boundary condition. Modification can include adjusting, replacing, or removing the existing condition. This allows for the individualization of existing boundary conditions. By entering a user input, the control variables can also take user preferences into account. This improves user satisfaction, can increase safety, and allows for the optimal determination of the system's energy utilization scenario.

[0116] An eighth artificial intelligence model can be configured to learn the boundary condition from the user input. This eighth artificial intelligence model can, for example, be communicatively connected to the human-machine interface. It can be configured to read the user input and to input the learned boundary condition as input data into the determining model.

[0117] The multi-agent control system can additionally include a maintenance agent and / or a ventilation / cooling agent. The maintenance agent can aim for optimized system maintenance. It can provide predictive maintenance based on collected data, enabling the determination of the most suitable time for repairing and / or replacing components. The ventilation / cooling agent can aim for optimizing the ventilation and cooling of buildings. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0118] In one embodiment, the aforementioned process steps can be carried out wholly or partially by a central computing system, by a decentralized computing system, or by a combination of a central computing system and a decentralized computing system.

[0119] A central computing system can handle calculations and data processing at a central location. It can receive data from various locations or plant components via server services or cloud solutions. A central computing system can increase overall security, simplify system management (for example, through standardization), and reduce costs.

[0120] A decentralized computing system can comprise an energy management system with a local computing system, for example, with machine learning functionality and neural networks. The decentralized computing system can provide good fault tolerance and rapid adaptation to local requirements.

[0121] A combination of a central computing system and a decentralized computing system can enable distributed use, allowing the respective advantages of the systems to be utilized.

[0122] A second aspect concerns a computer program. The computer program comprises instructions that, when executed by a computer, cause it to carry out the procedure according to the first aspect.

[0123] The respective advantages and further features can be found in the description of the first aspect, whereby elaborations of the first aspect also form elaborations of the second aspect and vice versa.

[0124] A third aspect concerns a device that is set up to carry out steps of the method according to the first or second aspect. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0125] The respective advantages and further features can be found in the description of the first and second aspects, whereby elaborations of the first and second aspects also form elaborations of the third aspect and vice versa.

[0126] A fourth aspect concerns an installation with a device for carrying out a method according to the first aspect or the second aspect. The installation comprises a means for drawing energy from the power grid, a means for feeding energy into the power grid, a switchable load, preferably one or more from a heat pump, a washing machine, a clothes dryer, or a dishwasher, and a photovoltaic system for generating electricity and / or a wind turbine for generating electricity. The installation preferably includes a unidirectional or bidirectional charging station for charging electric vehicles and / or an electricity storage system for energy storage.

[0127] At least parts of the system can communicate via an EEBUS. The EEBUS can be a standards-based communication interface designed for optimized energy management. The system can also use other IoT protocols for communication. User settings can be communicated via the EEBUS and / or these other IoT protocols. These other IoT protocols can include one or more of MQTT (Message Queuing Telemetry Transport), AMQP (Advanced Message Queuing Protocol), XMPP (Extensible Messaging and Presence Protocol), OPC UA (OPC Unified Architecture), Z-Wave, Zigbee, Thread, CoAP (Constrained Application Protocol), LoRa and LoRaWAN, Bluetooth and BLE (Bluetooth Low Energy), or LWM2M (Lightweight M2M).

[0128] The respective advantages and further features can be found in the description of the first, second and third aspects, whereby elaborations of the first, second and third aspects also form elaborations of the fourth aspect and vice versa.

[0129] In one embodiment, the method may comprise the following steps: Reading in a current heat demand parameter set. Inputting the current heat demand parameter set as input signals into a trained ninth artificial intelligence model, wherein the ninth artificial intelligence model was trained based on training data which combinations of a historical heat demand parameter set and a historical TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0130] The process includes the heat demand and provides an estimate of the expected heat demand as input data. It determines a forecast of the expected heat demand, including a temporal profile, based on the input data of the trained ninth artificial intelligence model. The forecast of the expected heat demand is then fed as input data into the determination model.

[0131] The current heat demand parameter set can be representative of the expected heat demand. This set can include a current heat energy demand value and current heat energy demand data from heat energy consumers. The current heat energy demand data can incorporate one or more weather forecast data points, preferably temperature, solar radiation, and / or wind speed, a value representative of a season, or a value representative of a solar position profile.

[0132] The historical heat demand parameter set can be representative of historical heat demand. This set can include historical heat energy demand data from heat energy consumers. The historical heat energy demand data can include one or more weather data points, preferably temperature, solar radiation, and / or wind speed, a value representative of a season, or a value representative of a solar position profile.

[0133] Heat energy consumers can, for example, use heat energy for one or more systems such as hot water preparation, space heating, or process heat. A heat energy consumer can be one or more radiators, underfloor heating systems, household appliances such as washing machines, dryers, cooking appliances or ovens, instantaneous water heaters, or steam generators. The consumer value can include the presence of the heat energy consumers. The control variables can include the utilization of the heat energy consumers. Utilization can include a usage value for each heat energy consumer.

[0134] The process can include a further step of reading the current heat demand parameter set as training data for the ninth artificial intelligence model. This enables continuous training of the ninth artificial intelligence model. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0135] In one embodiment, the system can include the controllable heat generator. The method can include the steps of reading in a heat energy consumption value of the controllable heat generator and inputting this heat energy consumption value as input data into the determination model. The manipulated variables can include a utilization value of the controllable heat generator.

[0136] The controllable heat generator can provide the system with additional heat energy by utilizing an energy source. The controllable heat generator can comprise one or more controllable heat generators. The heat energy consumption value can include a heat operating resource. The heat operating resource can describe resource input for operating the controllable heat generator. The heat operating resource can take into account the use of the energy source of the controllable heat generator. The controllable heat generator can be decoupled from the controllable power generator. The controllable heat generator can be coupled to the controllable power generator.

[0137] In one embodiment, the controllable heat generator and the controllable power generator can form a combined heat and power (CHP) system, for example, a cogeneration unit. To form a CHP system, the controllable heat generator and the controllable power generator can be combined, for example, in a CHP plant. By implementing combined heat and power, electricity and heat can be generated simultaneously. The CHP plant can be a cogeneration unit. The controllable heat generator or the controllable power generator can specify a controllable target value. This target value can be a specification according to which the CHP plant is operated. The target value can be a predetermined heat energy generation or a predetermined electricity generation. For example, the controllable target value can be a heat energy demand specification for the plant. The target value can be controlled centrally or decentrally.The generation of heat energy to meet the heat energy demand can depend on the necessary resource input. For this purpose, the necessary resource input for providing heat energy using combined heat and power (CHP) and using only the controllable heat generator can be compared. The method with the lower required resource input can be prioritized. A comparative value can be determined based on this comparison between heat generation using CHP and using only the controllable heat generator. This comparative value can be used as input data in the determination model (TQ-Systems GmbH WBH-Ref: 343.0029WO, March 26, 2026).

[0138] The comparison value can be entered. It can comprise a pair of values, where one value describes heat generation using combined heat and power (CHP) and the other describes heat generation using only the controllable heat generator. Alternatively, the comparison value can be a single value that considers both pairs. The comparison value can be determined based on the plant's energy utilization scenario to be optimized. For example, the comparison value can be determined based on optimizing net energy consumption and / or CO2 emissions, specifically minimizing overall CO2 resource use and / or CO2 emissions.For example, based on minimizing CO2 emissions as the energy utilization scenario to be optimized for the plant, a CO2 emission value for heat generation using combined heat and power (CHP) and a CO2 emission value for heat generation using only the controllable heat generator can be determined to establish a benchmark value. These values ​​can then be combined to form the benchmark value. The resulting benchmark value can be entered as input data into the determination model to be considered when determining the temporal profile of the plant's control variables for the energy utilization scenario to be optimized. This allows the control variables to be optimized over time for the energy utilization scenario to be optimized.The necessary resource input using combined heat and power (CHP) can take into account the feed-in tariff or the electricity reference value, as well as the electricity generated. The electricity generated by the CHP plant can be entered into the calculation model. Alternatively, the control variable for electricity generation can be overridden subsequently. This ensures that the current heat demand is met.

[0139] Combined heat and power (CHP) plants can achieve high fuel efficiency and enable the combined production of heat and electricity. This allows the controllable heat generator to meet a heat energy demand and additionally utilize any electricity generated during heat production. By coupling the energy sources, a minimum energy output of both heat and electricity can be achieved.

[0140] In one embodiment, the method can include the step of reading in a heat energy demand specification. The method can further include a step of entering the heat energy demand specification as input data into the determination model. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0141] The process can include a further step of inputting a minimum electricity generation as input data into the determination model. The controllable target value can be the heat energy demand specification. The minimum electricity generation can be based on the input heat energy demand specification. The minimum electricity generation can be the amount of electricity generated by fulfilling the heat energy demand specification in combined heat and power (CHP). This can reduce the amount of electricity drawn from the grid or the energy storage system. Additionally or alternatively, the amount of energy fed into the grid can be increased. This can reduce resource input, such as the initial resource input.

[0142] Alternatively, the process can include a step for reading in the electricity demand specification. The process can further include a step for entering the electricity demand specification as input data into the determination model. The process can further include a step for entering a minimum heat generation value as input data into the determination model. The controllable target value can be the electricity demand specification. The minimum heat generation value can be based on the read-in electricity demand specification.

[0143] This allows the heat energy consumption value of the controllable heat generator to be taken into account when providing the heat energy demand specification, taking into account a minimum electricity generation.

[0144] In one embodiment, the method may include a step of determining the usable waste heat generated during the operation of the controllable power generator and / or the usable waste heat generated during the operation of the controllable heat generator. The method may further include a step of inputting the usable waste heat generated during the operation of the controllable power generator and / or the usable waste heat generated during the operation of the controllable heat generator as input data into the determination model. The resulting usable waste heat can be used for heat generation or for electricity generation.

[0145] This can reduce resource input, such as the initial resource input, and increase the plant's efficiency, particularly fuel utilization. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0146] The multi-agent control system can additionally include one or more heat consumption agents, controllable heat generator agents, CHP agents, or waste heat agents. The heat consumption agent can aim for optimized utilization of heat consumption. The controllable heat generator agent can aim for optimized utilization of the controllable heat generator. The CHP agent can aim for optimized utilization of combined heat and power. The waste heat agent can aim for optimized utilization of waste heat.

[0147] In one embodiment, the system can comprise a microgrid solution with at least two connected households. The two households can be connected both communicatively and energetically. The energy connection can include electricity and heat. The microgrid solution with at least two connected households can include a local power generator and / or a controllable heat generator, for example, in the form of a combined heat and power plant. The local power generator and / or the controllable heat generator can be connected separately to each of the at least two connected households.

[0148] The multi-agent control can additionally include a coordinator agent for each of the at least two connected households and / or a microgrid coordinator agent. The microgrid coordinator agent can have the goal of optimizing the microgrid solution. The coordinator agent can also have the goal of optimizing the respective assigned household.

[0149] The proposed method can be used for one or more of the following: transmitting the electric vehicle's energy demand and departure time to the energy management system, charging the electric vehicle when PV generation is high, bidirectional charging of the electric vehicle (vehicle-to-home, V2H), overload protection through charging current reduction (OPEV), recording the electric vehicle's charging current consumption (EVCEM), transmitting the electric vehicle's state of charge (EVSOC), predicting expected energy generation, controlling battery storage systems, optimizing self-consumption, monitoring inverters and PV strings, visualizing aggregated PV data, wind energy data and battery data, configuring and monitoring the operating mode of a heating circuit, setting and reading a target temperature, and controlling and monitoring a hot water system.A temperature function and system function of the heating, ventilation and air conditioning (HVAC) systems, a recording of the total energy consumption of the system, a creation of optimal operating schedules for TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0150] Components of the system, a limitation of power at a grid connection point, self-consumption optimization through intelligent control of switchable consumers and producers, control and monitoring of switchable consumers, preferably the washing machine, refrigerators, tumble dryer or dishwasher, transmission of power forecasts and power schedules, aggregation of individual devices into larger power pools for grid services, grid-supporting operation of system components or monitoring and control of Smart Grid Ready conditions (MCSGRC).

[0151] Another aspect concerns a method for storing a data set for use in a process according to the first aspect. The process includes a step of reading the data set. The data set comprises data and at least one tag.

[0152] The at least one tag includes one or more pieces of geographical data, preferably a two-dimensional or three-dimensional position of a component of the plant, plant information, preferably one or more pieces of plant size, a value representative of an energy efficiency class of the plant, a value representative of materials used, a value representative of a performance parameter and / or a size specification of the PV plant, a value representative of a performance parameter, a size specification of the wind turbine, time data, calendar data, a time of day, a value representative of a season, a value representative of a course of the sun's position, weather forecast data or weather data.

[0153] The data includes one or more of the following parameters: current electricity consumption parameter set, historical electricity consumption parameter set, historical electricity consumption, forecast of expected electricity consumption, current electricity generation parameter set, historical electricity generation parameter set, historical electricity generation, forecast of expected electricity generation, energy system parameter set, electricity reference value, feed-in balancing, current electricity reference value parameter set, historical electricity reference value parameter set, historical electricity reference value, forecast of expected electricity reference value, current feed-in balancing parameter set, historical feed-in balancing parameter set, historical feed-in balancing. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0154] Prediction of the expected feed-in balance, the electricity consumption value of the controllable power generator, a current heat demand parameter set, a historical heat demand parameter set, a historical heat demand, the prediction of the expected heat demand, the heat consumption value of the controllable heat generator, a current CO2 emission parameter set, a historical CO2 emission parameter set, a historical CO2 emission, the prediction of the expected CO2 emission, a CO2 resource input, a current CO2 resource input parameter set, a historical CO2 resource input parameter set, a historical CO2 resource input, the prediction of the expected CO2 resource input, a current charging station parameter set, a historical charging station parameter set, a historical charging station usage, a prediction of the expected charging station usage, or the temporal profile of the manipulated variables.

[0155] The process includes the further step of saving the data record. This data record can be used when the initial data for a plant or plant component is unknown.

[0156] The data in a record can contain a timestamp. The record can be stored in a time series database. This allows for faster querying and more efficient data processing. Furthermore, the time series database can be used for forecasting.

[0157] The data records can be saved continuously.

[0158] The process may include a further step of reading data from a data set.

[0159] Another aspect concerns a database encompassing a large number of data records, with each data record including at least one tag and data.

[0160] The at least one tag comprises one or more pieces of geographical data, preferably a two-dimensional or three-dimensional position of a component of the plant, plant information, preferably one or more pieces of plant size, a value. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0161] representative of an energy efficiency class of the system, a value representative of materials used, a value representative of a performance parameter and / or a size specification of the PV system, a value representative of a performance parameter, a size specification of the wind turbine, time data, calendar data, a time of day, a value representative of a season, a value representative of a course of the sun's position, weather forecast data or weather data.

[0162] The data includes one or more of the following: current electricity consumption parameter set, historical electricity consumption parameter set, historical electricity consumption, forecast of expected electricity consumption, current electricity generation parameter set, historical electricity generation parameter set, historical electricity generation, forecast of expected electricity generation, energy system parameter set, electricity reference value, feed-in balancing, current electricity reference value parameter set, historical electricity reference value parameter set, historical electricity reference value, forecast of expected electricity reference value, current feed-in balancing parameter set, historical feed-in balancing parameter set, historical feed-in balancing, forecast of expected feed-in balancing, electricity consumption value of the dispatchable generator, and a current heat demand parameter set.a historical heat demand parameter set, a historical heat demand, the forecast of the expected heat demand, the heat energy consumption value of the controllable heat generator, a current CO2 emission parameter set, a historical CO2 emission parameter set, a historical CO2 emission, the forecast of the expected CO2 emission, a CO2 resource input, a current CO2 resource input parameter set, a historical CO2 resource input parameter set, a historical CO2 resource input, the forecast of the expected CO2 resource input, a current charging station parameter set, a historical charging station parameter set, a historical charging station usage, a forecast of the expected charging station usage, or the temporal profile of the manipulated variables.

[0163] New or additional data records can be continuously stored in the database. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0164] Another aspect concerns training an artificial intelligence model of an energy management system to determine an optimized energy utilization scenario. The training process includes a first step of reading a dataset containing comprehensive information about the energy management system's deployment location. The process then involves a second step of generating training data based on the imported dataset. This training data is derived from a separate dataset in a database, based on a tag of data from the second dataset and data from the imported dataset. Finally, the process includes a third step of training the artificial intelligence model using this training data.

[0165] The deployment location can include one or more geographical data, preferably a two-dimensional or three-dimensional position of a component of the system, system information, preferably one or more of a system size, a value representative of an energy efficiency class of the system, a value representative of materials used, a value representative of a performance parameter and / or a size specification of the PV system, a value representative of a performance parameter, a size specification of the wind turbine, time data, calendar data, a time of day, a value representative of a season, a value representative of a course of the sun's position or weather data, preferably one or more of a temperature, a value representative of solar radiation or a wind speed.

[0166] This allows for the identification of highly identical datasets from the database for training the artificial intelligence model of a new plant or plant component, and for the artificial intelligence model to be trained on these datasets. This enables new plants or plant components to be operated more efficiently and more quickly, saving training and optimization time.

[0167] Another aspect concerns a method for training an initial artificial intelligence model for use in predicting expected electricity consumption, including a temporal profile, in a procedure following the first aspect. The training method includes a step of inputting training data into the initial artificial intelligence model, where the training data comprises combinations of a historical electricity consumption parameter set and a historical electricity consumption. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0168] and provides as input data an estimate of expected electricity consumption, including a temporal profile.

[0169] Another aspect concerns a method for training a second artificial intelligence model for use in predicting expected electricity generation, including a time course, in a procedure according to the first aspect. The training method comprises a step of inputting training data into the second artificial intelligence model, where the training data includes combinations of a historical electricity generation parameter set and a historical electricity generation value, and provides as output data an estimate of expected electricity generation, including a time course.

[0170] Another aspect concerns a method for training a third artificial intelligence model, for use in determining, for an energy utilization scenario of the plant to be optimized, a temporal profile of the plant's control variables in a process following the first aspect. This method includes a further step of inputting training data into the third artificial intelligence model, where the training data comprises combinations of historical input data and historical temporal profiles of the control variables, and provides as output data a temporal profile of the plant's control variables for an energy utilization scenario to be optimized.

[0171] Another aspect concerns a procedure for training a fourth artificial intelligence model for use in predicting expected CO2 emissions in a process following the first aspect. The training procedure includes a step of inputting training data into the fourth artificial intelligence model. The training data comprises combinations of a historical CO2 emission parameter set and a historical CO2 emission value, and provides an estimate of expected CO2 emissions as output data.

[0172] Another aspect concerns a method for training a fifth artificial intelligence model for use in predicting an expected CO2 emission value in a procedure following the first aspect. The training method includes a step of inputting training data into the fifth artificial intelligence model, where the training data consists of combinations of a historical CO2 resource input parameter set and a TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0173] historical CO2 resource use and provides an estimate of expected CO2 resource use as input data.

[0174] Another aspect concerns a procedure for training a sixth artificial intelligence model for use in estimating an expected current reference value in a procedure according to the first aspect. The training procedure includes a step of inputting training data into the sixth artificial intelligence model, where the training data comprises combinations of a historical current reference value parameter set and a historical current reference value, and provides an estimate of an expected current reference value as output data.

[0175] Another aspect concerns a procedure for training a seventh artificial intelligence model for use in predicting an expected feed-in compensation in a procedure following the first aspect. The training procedure includes a step of inputting training data into the seventh artificial intelligence model, where the training data comprises combinations of historical feed-in compensation parameter sets and a historical feed-in compensation, and provides as output an estimate of an expected feed-in compensation.

[0176] Another aspect concerns a procedure for training a ninth artificial intelligence model for use in predicting an expected heat demand in a procedure according to the first aspect. The training procedure includes a step of inputting training data into the ninth artificial intelligence model, where the training data comprises combinations of a historical heat demand parameter set and a historical heat demand, and provides an estimate of an expected heat demand as output data.

[0177] The training data can include measurements of a multitude of relevant states and / or environmental parameters of the system and / or the system component containing an artificial intelligence model. The training data can thus provide a representative picture of the relevant states and environmental parameters. Representing a multitude of relevant states and / or environmental parameters of the system and / or system component can, for example, enable a more accurate estimation of the output data of the artificial intelligence models. The training data can be prepared before input. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0178] The respective advantages and further features can be found in the description of the preceding aspects, whereby elaborations of previous aspects also form elaborations of the next aspect and vice versa.

[0179] Brief description of the characters

[0180] Figure 1 shows a schematic perspective view of a system according to one embodiment.

[0181] Figure 2 schematically illustrates a method for optimizing the energy of a plant, wherein one energy utilization scenario to be optimized for the plant includes maximizing self-consumption of electricity.

[0182] Figure 3 schematically shows a method for optimizing the energy of the plant, wherein the energy utilization scenario to be optimized includes an optimization of net resource use.

[0183] Figure 4 schematically shows a method for optimizing the energy of the plant, wherein the energy utilization scenario to be optimized includes the optimization of CO2 emissions.

[0184] Figure 5 schematically shows a method for optimizing the energy of the plant, wherein the energy utilization scenario to be optimized includes an optimization of the net resource input taking into account thermal energy and electrical energy.

[0185] Figure 6 schematically shows a method for optimizing the energy of the plant, wherein the energy utilization scenario to be optimized comprises minimizing a combination of the first resource input and the second resource input.

[0186] Figure 7 schematically shows a procedure for the continuous training of the artificial intelligence models described in Figures 2 to 6. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0187] Figure 8 schematically shows a method for storing a data set for use in a method according to Figures 2 to 7.

[0188] Figure 9 schematically shows a database.

[0189] Figure 10 schematically shows a procedure for training an artificial intelligence model of an energy management system.

[0190] Figure 11a schematically shows exemplary time courses of the forecast of expected electricity consumption and the forecast of expected electricity generation.

[0191] Figure 11b schematically illustrates exemplary temporal profiles of input variables of the determination model.

[0192] Figure 12 schematically illustrates exemplary time profiles of control variables of the system.

[0193] Detailed description of embodiments

[0194] The system 1 shown in Figure 1 comprises a residential building 4 with a device 9 for carrying out a method described in Figures 2 to 8. System 1 further comprises a means 12 for drawing energy from the power grid and a means 13 for feeding energy into the power grid, which are indicated by the arrows. System 1 also includes switchable loads, in this case a washing machine 5 and a dishwasher 6. Additional switchable loads are possible. For electricity generation, System 1 includes a photovoltaic system 2 and a wind turbine 3 as electricity generators. System 1 also includes a bidirectional charging station 8 for charging electric vehicles and an energy storage device 7 for energy storage. The energy storage device 7 can release the stored energy at any time. Energy can be supplied to the energy storage device 7 for storage.The system 1 further comprises a controllable heat generator 14 and a controllable power generator 15. The controllable heat generator 14 is designed as a gas-fired boiler and provides heat energy by using an energy source, in this case, the combustion of gas. The heat energy can be used by heat energy consumers, such as the heater 11. The controllable power generator 15 is designed as a generator set and provides electricity by using an energy source, in this case, the combustion of gasoline. The electricity can then be used by the switchable consumers or the charging station 8. (TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026)

[0195] Electricity storage 7 is stored and / or fed into the public electricity grid via the means 13 for energy feed-in into the electricity grid.

[0196] Annex 1 also includes an input device 10 for entering a user preference. A user can also define an energy usage scenario for Annex 1 to be optimized via input device 10. This is selected from a predefined list of energy usage scenarios to be optimized.

[0197] The components of system 1 are at least communicatively connected to device 9 of system 1. This allows the components to exchange data with device 9. For example, the power generators can communicate data about their current power generation with device 9. The same applies to the heat energy generated by the controllable heat generator 14 and the electricity generated by the controllable power generator 15. The switchable loads and the heating system 11 can also communicate their status or availability to device 9. Communication can take place via EEBUS and other IoT protocols.

[0198] The components are also partially interconnected via power transmission. For example, the switchable loads are connected to the electronics of house 4, enabling them to be supplied with power. Similarly, the controllable power generator 15, the charging station 8, and the energy storage unit 7 are connected to the electronics of the house and, via this connection, to each other. The power generators are also connected to the electronics. The electronics of the house are connected to the device 12 for drawing energy from the grid and to the device 13 for feeding energy into the grid. The controllable heat generator 14 is connected to the heating system ll of house 4 via a heating system. This makes it possible to distribute and utilize the heat energy and electricity within house 4.

[0199] To optimize the energy consumption of plant 1, it has various control variables. These control variables are determined by a method described in Figures 2 to 7 below for an energy utilization scenario of plant 1 that is to be optimized. Each control variable has a temporal profile.

[0200] The control variables include energy consumption from the energy storage system 7, energy charging of the energy storage system 7, use of the switchable loads, and energy charging of the TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0201] electric vehicle, an energy draw from the electric vehicle, an energy draw from the electricity grid, an energy feed-in to the electricity grid, a value representative of a usage value of the controllable electricity generator 15, a value representative of an adjustable opening position of a window not shown, a use of the heat energy consumers and a usage value of the controllable heat generator 14.

[0202] The control variables are managed and represented by agents that are part of a multi-agent control system. Each agent can have the ability to achieve defined goals through interaction with an environment and other agents.

[0203] The multi-agent control system includes an electricity energy storage agent, an electricity consumer agent, a grid monitoring agent, a renewable energy agent, a coordinator agent, an energy feed-in agent, a CO2 emission agent, a CO2 resource utilization agent, a feed-in balancing agent, an electricity purchase value agent, a controllable electricity generator agent, a maintenance agent, a ventilation / cooling agent, a heat consumption agent, and a controllable heat generator agent. The electricity consumer agent includes an agent for washing machine 5 and an agent for dishwasher. The renewable energy agent includes an agent for PV system 2 and an agent for wind power plant 3.

[0204] The controllable heat generator 14 and the controllable power generator 15 can also be configured as a combined heat and power (CHP) plant in an alternative embodiment. This is particularly useful in industrial plants or in a microgrid solution with numerous interconnected households, such as apartment buildings or multi-unit residential complexes. The CHP plant features combined heat and power generation from the controllable heat generator 14 and the controllable power generator 15. Through this combined heat and power configuration, electricity and heat are generated simultaneously. The CHP plant has a controllable target value, which can be either a predetermined heat energy output or a predetermined electricity output. In the case of CHP, this controllable target value can result in a minimum electricity output or a minimum heat output.In this alternative embodiment, the multi-agent control can additionally include a coordinator agent for each of the networked households, a microgrid coordinator agent, and a CHP agent. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026.

[0205] Figure 2 schematically illustrates a method for energy optimization of the plant 1 described in Figure 1.

[0206] The procedure includes step S1 of reading in a current electricity consumption parameter set. This parameter set comprises data representative of current electricity consumption. Specifically, it includes a value for the current electricity consumption and current electricity consumption data from the energy consumers. This data incorporates time data, calendar data, weather data, and weather forecast data.

[0207] The procedure includes a further step S2 of inputting the current electricity consumption parameter set as input signals into a trained first artificial intelligence model, wherein the first artificial intelligence model was trained based on training data which includes combinations of a historical electricity consumption parameter set and a historical electricity consumption and provides as output data an estimate of an expected electricity consumption, including a temporal profile.

[0208] The historical electricity consumption parameter set comprises data representative of historical electricity consumption. This set includes multiple values ​​of historical electricity consumption and historical electricity consumption data from electricity consumers. The historical electricity consumption data incorporates time data, calendar data, weather data, and / or weather forecast data. The multiple values ​​of historical electricity consumption comprise time series of historical electricity consumption from the electricity consumers. For each of the electricity consumers shown in Figure 1, a time series of electricity consumption data is included. Thus, the multiple values ​​of historical electricity consumption include a time series of the historical electricity consumption of washing machine 5 and a time series of the historical electricity consumption of dishwasher 6.If the system includes additional electricity consumers, these can be taken into account accordingly. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026.

[0209] The procedure includes a further step S3 of determining a prediction of the expected electricity consumption, comprising a temporal profile, based on the initial data of the trained first artificial intelligence model.

[0210] The procedure includes a further step S4 of reading in a current electricity generation parameter set. This current electricity generation parameter set comprises data representative of current electricity generation. Specifically, it includes a value representing the current electricity generation, as well as current electricity generation data for PV energy and current electricity generation data for wind energy. The current electricity generation data for PV energy and the current electricity generation data for wind energy take into account the time of day, a value representative of the season, a value representative of the sun's position, weather data, and weather forecast data.

[0211] The procedure includes a further step S5 of inputting the current electricity-energy generation parameter set as input signals into a trained second artificial intelligence model, wherein the second artificial intelligence model was trained based on training data which includes combinations of a historical electricity-energy generation parameter set and a historical electricity-energy generation and provides as output data an estimate of an expected electricity-energy generation, including a time course.

[0212] The procedure includes a further step S6 of determining a prediction of the expected electricity energy generation, comprising a temporal profile, based on the input data of the trained second artificial intelligence model.

[0213] The procedure includes a further step S100 of entering the prediction of the expected electricity generation, the prediction of the expected electricity consumption and an energy system parameter set as input data into a determination model.

[0214] The energy system parameter set is representative of the current energy state of plant 1. The energy system parameter set includes the current electricity consumption value, the current electricity generation value, a consumer value, or a state-of-charge value. The consumer value is representative of the presence of switchable loads. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0215] Referring to Figure 1, the consumer value includes a value representative of the presence of the washing machine 5 and a value representative of the presence of the dishwasher 6. The charge level value is representative of the charge level of the energy storage device 7. The charge level is variable over time. The charge level value is a value between 0 and 100%.

[0216] The procedure includes a further step, S101, of determining, for an energy utilization scenario of plant 1 to be optimized, a time-dependent profile of the plant 1's control variables for managing the use of available energy, as the output of the determination model. The determination model is configured to determine the time-dependent profile of the plant 1's control variables based on the input data. The energy utilization scenario of plant 1 to be optimized defines an optimization objective. This objective involves maximizing self-consumption of electricity. A user can select the energy utilization scenario to be optimized via input device 10. Alternatively, the energy utilization scenario to be optimized can be centrally predefined. The energy utilization scenario of plant 1 to be optimized encompasses an optimization period. In this case, the optimization period is 24 hours.Accordingly, the control variables are determined in an optimized manner for a 24-hour period with respect to the energy usage scenario to be optimized.

[0217] The procedure includes a further step, S102, of controlling plant 1 based on the specified control variables. This control involves the allocation of available electrical energy and available thermal energy.

[0218] Figure 3 schematically shows a procedure for optimizing the energy consumption of plant 1. The energy utilization scenario to be optimized differs from that in Figure 2. The energy utilization scenario to be optimized comprises the optimization of net resource input. Steps S1 to S7 are identical to the steps described in Figure 2.

[0219] The following steps can be performed before step S101 of determining the time course of the manipulated variables. The steps can be executed sequentially. The order of inputting the data into the determination model is arbitrary. Steps S100-1 to S100-5 each represent the input step into the determination model. Step S101 can follow such a step, as indicated by the dash-dot line. Alternatively, after inputting the data, TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0220] Further steps to determine input data will follow. Alternatively, all input data can first be determined and entered into the determination model in one step.

[0221] The procedure includes a further step S8 of reading in a current reference value. The current reference value comprises a current current value and can include a prediction of an expected current reference value. In step S100-2, the current reference value is entered into the determination model. The prediction of the expected current reference value can be provided by an external source. Alternatively or additionally, the prediction of the expected current reference value can be based on input data from a trained sixth artificial intelligence model.

[0222] The current power reference value has a fixed value for a fixed period. The procedure can include a further step S9 of checking whether the optimization period is longer than the fixed period. If the optimization period is longer than the fixed period, the procedure can include the following steps. Alternatively, the procedure can include the following steps regardless of the result of step S9. This is indicated by the arrow with the dotted line. The procedure can include a further step S10 of reading in a current power reference value parameter set.Input S11 of the current current reference value parameter set as input signals into the trained sixth artificial intelligence model, wherein the sixth artificial intelligence model was trained based on training data comprising combinations of a historical current reference value parameter set and a historical current reference value, and provides as output data an estimate of an expected current reference value. Determine S12 of a prediction of the expected current reference value, comprising a time course, based on the output data of the trained sixth artificial intelligence model. Input S100-3 of the prediction of the expected current reference value as input data into the determination model.

[0223] The procedure includes a further step, S13, of reading in a feed-in compensation. The feed-in compensation comprises a current feed-in compensation and can include a prediction of an expected feed-in compensation. In step S100-4, the feed-in compensation is entered into the determination model. The prediction of the expected feed-in compensation can be provided by an external source. Alternatively or additionally, the prediction of the expected feed-in compensation can be based on input data from a trained seventh artificial intelligence model. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0224] The current feed-in compensation has a fixed value for a specified period. The procedure can include step S14 of checking whether the optimization period is longer than the specified period. If the optimization period is longer than the specified period, the procedure can include the subsequent steps. Alternatively, the procedure can include the following steps regardless of the result of step S14 of the check. This is indicated by the arrow with the dotted line. The procedure can include a further step S15 of reading in a current feed-in compensation parameter set.Input S16 of the current feed-in balancing parameter set as input signals into the trained seventh artificial intelligence model, where the seventh artificial intelligence model was trained based on training data comprising combinations of historical feed-in balancing parameter sets and a historical feed-in balancing event, and provides as output data an estimate of an expected feed-in balancing event. Determine S17 of a prediction of the expected feed-in balancing event, comprising a time course, based on the output data of the trained seventh artificial intelligence model. Input S100-5 of the prediction of the expected feed-in balancing event as input data into the determination model.

[0225] The procedure can include a further step S18 of reading in an electricity consumption value. This allows the resource usage of the controllable power generator to be taken into account when optimizing net energy consumption. The procedure can also include a further step S100-6 of entering the electricity consumption value of the controllable power generator as input data into the determination model.

[0226] Figure 4 schematically shows a method for optimizing the energy consumption of plant 1, where the energy utilization scenario to be optimized includes the optimization of CO2 emissions. Steps S1 to S7 are identical to the steps described in Figure 2. The method can be combined with the methods in Figure 3. Accordingly, the method steps can also be combined in the alternative embodiment.

[0227] The procedure includes a further step S19 of reading in a current CO2 emission parameter set. Furthermore, the procedure includes a step S20 of inputting the current CO2 emission parameter set as input signals into a trained fourth artificial intelligence model, where the fourth artificial intelligence model is based on training data. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0228] The fourth artificial intelligence model was trained on combinations of historical CO2 emission parameter sets and historical CO2 emissions, and provided with an estimate of expected CO2 emissions as input data. Furthermore, the procedure includes step S21, which determines a prediction of the expected CO2 emissions, encompassing a time series, based on the input data of the trained fourth artificial intelligence model. Step S100-7 involves inputting the prediction of expected CO2 emissions as input data into the determination model.

[0229] This allows the energy utilization scenario for plant 1 to be optimized by minimizing CO2 emissions. Steps S101 and S102 can then be carried out.

[0230] If the energy utilization scenario to be optimized for plant 1 involves minimizing total CO2 resource input, the procedure additionally includes step S22 of reading in a current CO2 resource input parameter set. Furthermore, the procedure includes step S23 of inputting the current CO2 resource input parameter set as input signals into a trained fifth artificial intelligence model. This fifth artificial intelligence model was trained based on training data comprising combinations of a historical CO2 resource input parameter set and a historical CO2 resource input, and provides an estimate of expected CO2 resource input as output data.The procedure includes a further step, S24, of determining a forecast of the expected CO2 resource input, comprising a temporal profile, based on the input data of the trained fifth artificial intelligence model. Step S100-8 involves inputting the forecast of the expected CO2 resource input as input data into the determination model. Subsequently, steps S101 and S102 follow to determine the temporal profiles of the control variables and to control the plant.

[0231] Figure 5 schematically illustrates a method for optimizing the energy consumption of plant 1, where the energy utilization scenario to be optimized for plant 1 includes optimizing the net resource input, taking into account thermal and electrical energy. The method comprises steps S1 to S18, as described in Figure 3. This allows the net resource input of electricity to be determined and optimized. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0232] The procedure comprises step S25, which involves reading in a current heat demand parameter set. Furthermore, step S26 involves inputting the current heat demand parameter set as input signals into a trained ninth artificial intelligence model. This ninth artificial intelligence model was trained on training data comprising combinations of a historical heat demand parameter set and a historical heat demand, and provides an estimate of the expected heat demand as output data. In step S27, a forecast of the expected heat demand, including a time course, is determined based on the output data of the trained ninth artificial intelligence model. Step S100-9 involves inputting the forecast of the expected heat demand as input data into the determination model.The procedure includes step S28 of reading in a heat energy consumption value from the controllable heat generator 14. Furthermore, the procedure includes step S100-10 of inputting the heat energy consumption value of the controllable heat generator 14 as input data into the determination model. The heat energy consumption value represents a heat operating resource. The heat operating resource describes the resource input for operating the controllable heat generator 14 and takes into account the energy source of the controllable heat generator 14. The optimization of the net resource input can therefore be a weighted combination of optimizing the net resource input of heat energy and optimizing the net resource input of electrical energy.

[0233] In a further embodiment not shown, the method can comprise a step of determining usable waste heat generated during the operation of the controllable power generator and a step of inputting the waste heat into the determination model. Furthermore, the method can comprise a step of determining usable waste heat generated during the operation of the controllable heat generator 14 and a step of inputting the waste heat into the determination model.

[0234] If the controllable heat generator 14 and the controllable power generator 15 are configured for combined heat and power (CHP), the process can include a further step (not shown) of reading in a power demand specification or a heat demand specification. The demand specification depends on a defined target value and whether this target value includes a minimum power generation or a minimum heat generation. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0235] Figure 6 schematically illustrates a procedure for optimizing the energy consumption of plant 1, where the energy utilization scenario to be optimized involves minimizing a combination of the first and second resource inputs. The procedure includes step S29, which assigns a first resource input for energy consumption. It also includes step S30, which assigns a second resource input for CO2 emissions. The first resource input can comprise a thermal energy input, as determined in Figure 5, and an electrical energy input. The procedure further includes step S100-12, which inputs the first and second resource inputs into the determination model. Subsequently, in step S101, the temporal profiles of the control variables are determined based on the energy utilization scenario to be optimized for plant 1.

[0236] In a further embodiment not shown, the method can include a step for obtaining a user input. This step can be performed at any time before step S101. The user input is provided via a human-machine interface. The human-machine interface includes the input device 10 shown in Figure 1 for inputting the energy utilization scenario to be optimized.

[0237] The energy utilization scenario to be optimized for plant 1 can be adapted; in one embodiment (not shown), the method can execute all process steps S1 to S30, input the data into the determination model, and determine the time profiles of the manipulated variables according to the energy utilization scenario to be optimized for plant 1. The process steps shown in Figures 2 to 6 can be combined accordingly.

[0238] Figure 7 schematically shows a procedure for continuously training the artificial intelligence models.

[0239] The procedure includes a further step, TS1, of reading the current electricity consumption parameter set as training data for the first artificial intelligence model. Step TS1 follows step S1 of reading the current electricity consumption parameter set. Alternatively, this step can also be placed at any other TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0240] This will occur after step S1 of reading the current electricity consumption parameter set.

[0241] The procedure includes a further step, TS2, of reading the current electricity generation parameter set as training data for the second artificial intelligence model. Step TS2 occurs after step S4 of reading the current electricity generation parameter set. Alternatively, this step can also be performed at any other time after step S4 of reading the current electricity generation parameter set.

[0242] The procedure includes a further step, TS3, which involves reading in a time course of manipulated variables as training data for the third artificial intelligence model. Step TS3 follows step S101, which determines the time course of the manipulated variables.

[0243] The procedure includes a further step, TS4, of reading the current CO2 emission parameter set as training data for the fourth artificial intelligence model. Step TS4 occurs after step S19 of reading the current CO2 emission parameter set. Alternatively, this step can also be performed at any other time after step S19 of reading the current CO2 emission parameter set.

[0244] The procedure includes a further step, TS5, of reading the current CO2 resource input parameter set as training data for the fifth artificial intelligence model. Step TS5 occurs after step S22 of reading the current CO2 resource input parameter set. Alternatively, this step can also be performed at any other time after step S22 of reading the current CO2 resource input parameter set.

[0245] The procedure includes a further step, TS6, of reading the current electricity reference value parameter set as training data for the sixth artificial intelligence model. Step TS6 is performed after step S10 of reading the current electricity reference value parameter set. Alternatively, this step can also be performed at any other time after step S10 of reading the current electricity reference value parameter set. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0246] The procedure includes a further step, TS7, of reading the current feed-in balancing parameter set as training data for the seventh artificial intelligence model. Step TS7 occurs after step S15 of reading the current feed-in balancing parameter set. Alternatively, this step can also be performed at any other time after step S15 of reading the current feed-in balancing parameter set.

[0247] The procedure includes a further step, TS8, of reading the current heat demand parameter set as training data for the ninth artificial intelligence model. Step TS8 occurs after step S25 of reading the current heat demand parameter set. Alternatively, this step can also be performed at any other time after step S25 of reading the current heat demand parameter set.

[0248] Figure 7 also illustrates that the process steps S1 to S30 described in Figures 2 to 7 can also be carried out in a single embodiment that includes all the process steps described in Figures 2 to 7.

[0249] Figure 8 schematically shows a method for storing a data set for use in a process according to Figures 2 to 7. The process includes a step SD1 of reading the data set. The data set comprises data and at least one tag.

[0250] The at least one tag includes one or more pieces of geographical data, preferably a two-dimensional or three-dimensional position of a component of the plant, plant information, preferably one or more pieces of plant size, a value representative of an energy efficiency class of the plant, a value representative of materials used, a value representative of a performance parameter and / or a size specification of the PV plant, a value representative of a performance parameter, a size specification of the wind turbine, time data, calendar data, a time of day, a value representative of a season, a value representative of a course of the sun's position, weather forecast data or weather data.

[0251] The data includes one or more of the current electricity consumption parameter set, the historical electricity consumption parameter set, TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0252] historical electricity consumption, the forecast of expected electricity consumption, the current electricity generation parameter set, the historical electricity generation parameter set, the historical electricity generation, the forecast of expected electricity generation, the energy system parameter set, the electricity reference value, the feed-in balancing, the current electricity reference value parameter set, the historical electricity reference value parameter set, the historical electricity reference value, the forecast of the expected electricity reference value, the current feed-in balancing parameter set, the historical feed-in balancing parameter set, the historical feed-in balancing, the forecast of the expected feed-in balancing, the electricity consumption value of the controllable power generator, a current heat demand parameter set, a historical heat demand parameter set, a historical heat demand, the forecast of the expected heat demand,the heat energy consumption value of the controllable heat generator 14, a current CO2 emission parameter set, a historical CO2 emission parameter set, a historical CO2 emission, the forecast of the expected CO2 emission, a CO2 resource input, a current CO2 resource input parameter set, your historical CO2 resource input parameter set, a historical CO2 resource input, the forecast of the expected CO2 resource input, or the time course of the control variables.

[0253] The process includes a further step, SD2, of saving the data record. Saving the data records can be done continuously. The process also includes a further step, SD3, of reading data from a data record.

[0254] Figure 9 schematically shows a database D1. Database D1 comprises a multitude of data records D2. Each data record D2 includes at least one tag D3 and data D4.

[0255] The at least one day D3 comprises one or more pieces of geographical data, preferably a two-dimensional or three-dimensional position of a component of the system, system information, preferably one or more pieces of system size, a value representative of an energy efficiency class of the system, a value representative of materials used, a value representative of a performance parameter and / or a size specification of the PV system, a value representative of a performance parameter, a size specification of the wind turbine, time data, calendar data, a time of day, a value representative of a season, a value representative of a solar position progression, weather forecast data, or weather data. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0256] The data D4 includes one or more of the following: current electricity consumption parameter set, historical electricity consumption parameter set, historical electricity consumption, forecast of expected electricity consumption, current electricity generation parameter set, historical electricity generation parameter set, historical electricity generation, forecast of expected electricity generation, energy system parameter set, electricity reference value, feed-in balancing, current electricity reference value parameter set, historical electricity reference value parameter set, historical electricity reference value, forecast of expected electricity reference value, current feed-in balancing parameter set, historical feed-in balancing parameter set, historical feed-in balancing, forecast of expected feed-in balancing, electricity consumption value of the dispatchable generator, and a current heat demand parameter set.a historical heat demand parameter set, a historical heat demand, the forecast of the expected heat demand, the heat energy consumption value of the controllable heat generator 14, a current CO2 emission parameter set, a historical CO2 emission parameter set, a historical CO2 emission, the forecast of the expected CO2 emission, a CO2 resource input, a current CO2 resource input parameter set, your historical CO2 resource input parameter set, a historical CO2 resource input, the forecast of the expected CO2 resource input, or the time course of the control variables.

[0257] Figure 10 schematically shows a procedure for training an artificial intelligence model of an energy management system.

[0258] Training the artificial intelligence model of an energy management system comprises a first step (ST1) of reading a dataset. This dataset contains information about the deployment location of the energy management system. The process then includes a second step (ST2) of determining training data based on the read dataset. This training data is derived from another dataset in a database, based on a tag of data from the second dataset and data from the read dataset. Finally, the process includes a third step (ST3) of training the artificial intelligence model based on this training data. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0259] Figure 11a schematically illustrates exemplary time courses t of the forecasts for expected electricity consumption T1 and expected electricity generation T2. ​​The forecasts T1 and T2 were determined using the first and second artificial intelligence models, respectively. The forecasts T1 and T2 each exhibit a time course t over a period of time. Thus, over the time course t, the forecasts T1 and T2 display a multitude of variable values, each forming a continuous curve.

[0260] Figure 11b schematically shows exemplary time profiles t of input variables T1, T2, T3, and T4, which are fed into the control model. The input variables T1, T2, T3, and T4 comprise the time profile of the predicted expected electricity consumption T1 and the time profile of the predicted expected electricity generation T2, which were determined using the first and second artificial intelligence models, respectively (see Figure 11a). Furthermore, the input variables include the time profile of the value representing the presence of the washing machine T3 and the time profile of a consumer value T4. This illustrates the complexity of the control model when determining the manipulated variables, since the input variables each have a time profile t, which, as shown in Figure 12, allows the manipulated variables to be determined comprehensively over time.

[0261] Figure 12 schematically shows exemplary time profiles t of control variables T5, T6, T7, T8, T9, and T10 of system 1. These control variables represent the time profile of energy consumption from an energy storage device T5, the time profile of energy charging of the energy storage device T6, the time profile of the use of switchable loads T7, the time profile of energy charging of an electric vehicle T8, the time profile of energy consumption from the grid T9, and the time profile of energy feed-in to the grid. The control variables T5, T6, T7, T8, T9, and T10 exhibit variable values ​​over the time profile t, thus ensuring optimal control of the system at every point in time t, based on the energy utilization scenario of system 1 to be optimized.This allows for the implementation of a control mechanism that may appear disadvantageous at a specific point in time t, for example, drawing electricity from the grid even when sufficient electricity generation is available, because this is advantageous for the considered period of time based on the energy utilization scenario of plant 1 being optimized. Plant 1 is controlled advantageously over the period of time t using the control variables T5, T6, T7, T8, T9, and T10, based on the energy utilization scenario being optimized. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026.

[0262] Reference sign

[0263] 1 Annex

[0264] 2 PV systems

[0265] 3 wind turbines

[0266] 4 residential building

[0267] 5 Washing machine

[0268] 6 Dishwasher

[0269] 7 Energy storage

[0270] 8 charging stations

[0271] 9 Device

[0272] 10 Input device

[0273] 11 Heating

[0274] 12 means of obtaining energy from the electricity grid

[0275] 13 means of feeding energy into the electricity grid

[0276] 14 adjustable heat generators

[0277] 15 adjustable power generators

[0278] 51 Reading in a current electricity consumption parameter set

[0279] 52 Input into a trained first artificial intelligence model

[0280] 53 Determining a forecast of expected electricity consumption

[0281] 54 Reading in a current electricity generation parameter set

[0282] 55 Input signals into a trained second artificial intelligence model 57 Determine a prediction of expected electricity generation

[0283] 58 Reading in a power consumption value

[0284] 59 Check if the optimization period is longer than the fixing period 510 Read in a current power reference value parameter set

[0285] 511 Input into a trained sixth artificial intelligence model

[0286] 512 Determining a forecast of the expected electricity reference value

[0287] 513 Reading in a feed-in compensation

[0288] 514 Check if the optimization period is longer than the fixing period 515 Read in a current feed-in balancing parameter set

[0289] 516 Input into a trained seventh artificial intelligence model

[0290] 517 Determining a forecast of the expected feed-in compensation TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0291] 518 Reading in an electricity consumption value

[0292] 519 Reading in a current CO2 emission parameter set

[0293] 520 Input into a trained fourth artificial intelligence model

[0294] 521 Determining a forecast of expected CO2 emissions

[0295] 522 Reading in a current CO2 resource input parameter set

[0296] 523 Input into a fifth trained fifth artificial intelligence model

[0297] 524 Determining a forecast of expected CO2 resource use

[0298] 525 Reading in a current heat demand parameter set

[0299] 526 Input into a trained ninth artificial intelligence model

[0300] 527 Determining a forecast of the expected heat demand

[0301] 528 Reading in a heat energy consumption value

[0302] 529 Allocating an initial resource allocation

[0303] 530 Allocating a second resource

[0304] S100, S100-1 to S100-12 Enter into a determination model

[0305] 5101 Determining the time course of manipulated variables

[0306] 5102 Controlling the plant based on the specified actuators

[0307] TS1 Reading in the current electricity consumption parameter set as training data for the first artificial intelligence model

[0308] TS2 Reading in the current electricity generation parameter set as training data for the second artificial intelligence model

[0309] TS3 Reading in the time course of the control variables as training data for the third artificial intelligence model

[0310] TS4 Reading the current CO2 emission parameter set as training data for the fourth artificial intelligence model

[0311] TS5 Reading in the current CO2 resource input parameter set as training data for the fifth artificial intelligence model

[0312] TS6 Reading the current current reference value parameter set as training data for the sixth artificial intelligence model

[0313] TS7 Reading the current feed-in balancing parameter set as training data for the seventh artificial intelligence model. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026

[0314] TS8 Reading the current heat demand parameter set as training data for the ninth artificial intelligence model

[0315] SD1 Reading the data set

[0316] SD2 Saving the data set (Continuous -> Arrow in a circle)

[0317] SD3 Reading data from a data record

[0318] D1 Database

[0319] D2 data set

[0320] Day 3

[0321] D4 data

[0322] ST1 Reading a data record

[0323] ST2 Determining training data

[0324] ST3 Training the Artificial Intelligence Model

[0325] t time course

[0326] T1 Time course of the forecast of expected electricity consumption

[0327] T2 Time course of the forecast of expected electricity generation

[0328] T3 Time course of the value representative of the presence of the washing machine T4 Time course of the consumer value

[0329] T5 Time course of energy consumption from a power storage system

[0330] T6 Time course of the energy charging of the electricity storage system

[0331] T7 Time course of the use of switchable consumers TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026 T8 Time course of the energy charging of an electric vehicle

[0332] T9 Time course of energy consumption from the electricity grid

[0333] T10 Time course of energy feed-in to the power grid

Claims

TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026 Patent claims 1. Method for optimizing the energy of a plant (1) comprising: - Reading (S1) a current electricity consumption parameter set; - Input (S2) of the current electricity consumption parameter set as input signals into a trained first artificial intelligence model, wherein the first artificial intelligence model was trained based on training data which includes combinations of a historical electricity consumption parameter set and a historical electricity consumption and provides as output data an estimate of an expected electricity consumption, including a time course; - Determining (S3) a prediction of the expected electricity consumption, including a temporal profile, based on the initial data of the trained first artificial intelligence model; - Reading (S4) a current electricity generation parameter set; - Input (S5) of the current electricity generation parameter set as input signals into a trained second artificial intelligence model, wherein the second artificial intelligence model was trained based on training data which includes combinations of a historical electricity generation parameter set and a historical electricity generation and provides as output data an estimate of an expected electricity generation, including a time course; - Determining (S6) a prediction of the expected electricity energy generation, comprising a time course, based on the input data of the trained second artificial intelligence model; - Input (S100) the forecast of expected electricity generation, the forecast of expected electricity consumption and an energy system parameter set as input data into a determination model, - Determine (S101), for an energy utilization scenario of the plant (1) to be optimized, a time course of the plant's control variables (1) for controlling the use of available energy, as output of the determination model. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026 2. Method according to claim 1, wherein the system (1) comprises a means (12) for drawing energy from the power grid, a means (13) for feeding energy into the power grid, a switchable load and a power generator, and wherein the system (1) preferably comprises a unidirectional or bidirectional charging station (8) for charging electric vehicles and / or a power storage device (7) for energy storage.

3. The method of claim 2, wherein the current electricity consumption parameter set comprises a value of the current electricity consumption and current electricity consumption data of the electricity consumers, wherein the current electricity consumption data take into account time data, calendar data, weather data and / or weather forecast data, wherein the historical electricity consumption parameter set includes one or more values ​​of historical electricity consumption and historical electricity consumption data from electricity consumers, wherein the historical electricity consumption data take into account time data, calendar data, weather data and / or weather forecast data.

4. Method according to claim 2 or 3, wherein the current electricity generation parameter set comprises a value of the current electricity generation and current electricity generation data of PV energy and / or current electricity generation data of wind energy, wherein the current electricity generation data of PV energy and / or the current electricity generation data of wind energy take into account one or more values ​​from a time of day, a value representative of a season, a value representative of a solar position profile, weather data or weather forecast data. wherein the historical electricity generation parameter set comprises one or more values ​​of historical electricity generation and historical electricity generation data of PV energy and / or historical electricity generation data of wind energy, wherein the historical electricity generation data of PV energy and / or the historical electricity generation data of wind energy take into account one or more values ​​from a time of day, a value representative of a season, a value representative of a solar position profile, and weather data from weather forecast data. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026 5. A method according to any of the preceding claims, wherein the first artificial intelligence model is a neural network, preferably a Long Short-Term Memory (LSTM), a Recurrent Neural Network (RNN), a Gated Recurrent Unit (GRU) and / or a Convolutional Neural Network (CNN), the second artificial intelligence model is a neural network, preferably a Long Short-Term Memory (LSTM), a Recurrent Neural Network (RNN), a Gated Recurrent Unit (GRU) and / or a Convolutional Neural Network (CNN), and / or the determination model is designed as a third artificial intelligence model, as a statistical model with statistical and / or stochastic methods, or as a combination of a third artificial intelligence model and a statistical model with statistical and / or stochastic methods.

6. A method according to any of the preceding claims, referring back to claims 3 and 4, wherein the energy system parameter set comprises one or more of the current electricity consumption value, the current electricity generation value, a consumer value or a state of charge value.

7. Method according to one of claims 2 to 6, wherein at least one of the electricity generators is designed as a combined heat and power plant and provides useful heat in addition to electrical energy, wherein the determination model takes into account a heat utility value of the provided useful heat, a comparison value of an alternative heat generation and / or a CO2 emission value and / or CO2 cost value associated with an alternative heat generation when determining the control variables.

8. Method according to one of the preceding claims, wherein the control variables comprise one or more of the following: energy consumption from an energy storage device (7), energy charging of the energy storage device (7), use of the switchable loads, energy charging of an electric vehicle, energy consumption from the electric vehicle, energy consumption from the power grid, or energy feed-in to the power grid. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026 9. A method according to any of the preceding claims, comprising the steps of: Reading (TS1) the current electricity consumption parameter set as training data for the first artificial intelligence model; and / or Reading (TS2) the current electricity energy generation parameter set as training data for the second artificial intelligence model.

10. Method according to one of the preceding claims, wherein at least a part of the control variables are controlled and mapped by agents that are part of a multi-agent control system, each agent having the ability to achieve defined goals through interaction with an environment and other agents.

11. Method according to one of the preceding claims, wherein the energy utilization scenario to be optimized of the plant (1) comprises maximizing self-consumption of electricity, optimizing net energy consumption, optimizing CO2 emissions, optimizing net resource use or a weighted combination of optimizing net energy consumption, optimizing CO2 emissions and optimizing net resource use.

12. A method according to any one of the preceding claims, wherein the method comprises the steps: Reading (S8) a current reference value and / or reading (S13) a feed-in compensation, wherein the current reference value comprises a current current reference value and preferably a prediction of an expected current reference value, wherein the prediction of the expected current reference value is preferably based on output data of a trained sixth artificial intelligence model, wherein the feed-in compensation comprises a current feed-in compensation and preferably a prediction of an expected feed-in compensation, wherein the prediction of the expected feed-in compensation is preferably based on input data of a trained seventh artificial intelligence model; and TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026 Enter (S100-2) the electricity consumption value and / or enter (S100-4) the feed-in compensation as input data into the determination model.

13. Method according to any one of the preceding claims, wherein the installation (1) comprises a controllable power generator, and the procedure comprises the steps: Reading (S18) an electricity consumption value of the controllable generator; and entering (S100-6) the electricity consumption value of the controllable generator as input data into the determination model, where the control variables include a utility value of the controllable power generator.

14. Method according to any one of claims 1 to 10, comprising the steps: - Allocating (S29) an initial resource allocation for an energy consumption; - Allocating (S30) a second resource input for a CO2 emission; where the energy utilization scenario to be optimized includes minimizing a combination of the first resource input and the second resource input.

15. Method according to any of the preceding claims, comprising the step of: obtaining a user specification, where the user setting is configured to specify a boundary condition for the control variables.

16. A method according to any of the preceding claims, wherein the aforementioned method steps are performed wholly or partially by a central computing system, by a decentralized computing system, or by a combination of a central computing system and a decentralized computing system. TQ-Systems GmbH WBH-Ref: 343.0029WO March 26, 2026 17. Computer program comprising instructions which, when the program is executed by a computer, cause it to execute the method according to any of the preceding claims.

18. Device (9) configured to perform steps of the method according to any of the preceding claims.

19. System (1) with a device (9) for carrying out a method according to one of claims 1 to 16, wherein the system (1) comprises a means (12) for drawing energy from the power grid, a means (13) for feeding energy into the power grid, a switchable load, preferably one or more from a heat pump, a washing machine (5), a clothes dryer or a dishwasher (6), and a PV system (2) for generating electricity and / or a wind turbine (3) for generating electricity, wherein the system (1) preferably comprises a unidirectional or bidirectional charging station (8) for charging electric vehicles and / or an electricity storage device (7) for storing energy.

Images