Renewable Energy Grid Integration Week 2023

In battery pack manufacturing the main time-consuming aspects is the long test time for discharge and charge cycles. Making an AI that can assist by predicting the result of these tests while they are running is the end goal of this series of papers, where this is step one. To achieve this, the classification of the different battery pack types and the different tests must be performed. In collaboration with the company Banke Aps., who is a battery pack manufacturer, a database of tests has been build forming the foundation of this paper. Using the Gramian Angular Field (GAF) and Markov Transition Field (MTF) methods to transform the time series data into image form gives the possibility to utilize the standard convolutional neural network (CNN) structures to classify the battery pack and test type. Furthermore, building an algorithm that can distinguish between pass or fail tests with as high accuracy as possible, is important. Some question, that needs to be investigated, is which mathematical technique is best, GAF or MTF? Can these mathematical techniques identify important test features in the generated images and be used to estimate the test outcome?

The performance of the AI is measured in percent accuracy and the aim is to achieve as high accuracy as possible (above 90%) giving the confidence to apply the AI in the battery factory test system. Here the AI will evaluate the battery pack in real time and give the operator an indication if the battery pack will fail in the future or not.

To get above 90% accuracy, 10 pretrained neural networks will be tested and the best selected for this specific task. The networks range from “simple” (25 layers) to more complex (1243 layers) giving a good overview of the whole range of networks.

The German Government plans to have at least 15 million battery electric vehicles (EV) on German roads by 2030. Without intelligent, flexible charging this would require tremendous investments in energy generation and grid infrastructure. This paper outlines a solution for controlled charging of EVs in the form of event driven tariffs that offers market-driven and grid-serving flexibility. We suggest a regulatory framework for the yellow “grid traffic light phase”, which is coherent with the currently proposed regulatory framework of the German Ministry of Economic Affairs and Climate Protection (BMWK) and the Federal Network Agency (BNetzA). Within this proposed regulatory framework, a process structure for the suggested tariffs based on EV charging is described. To enable this tariff structure, the Smart Meter Gateway (SMGW) infrastructure needs to be adopted and combined with a home energy management system (HEMS). The paper outlines necessary functional extensions of the SMGW and HEMS infrastructure for the implementation of event driven tariffs. Furthermore, a proposal is made for the information exchange between the technical components and stakeholders. An outlook to the lab test of the enhanced SMGW and HEMS, as well as to the field test for the prioritised charging tariff is provided.

The process of electrification of the public transportation sector is resulting in a growing number of electric buses on the streets. Modeling and simulating the electric bus fleets can not only identify possible issues in time but can also provide valuable inputs for the optimal integration of these buses into existing operational plans and management systems. One of the important requirements for accurate modeling is knowledge of the energy consumption of the buses. This paper uses a data-driven approach to analyze the factors impacting energy consumption. The considered factors are: average daily temperature, trip length, total trip time, state of charge at the beginning of the trip, and average vehicle speed during the trip. Additionally, the impact of different buses and routes is analyzed by considering their ID numbers. The data from 96 different electric buses were collected in the city of Hamburg for 13 months. The analysis of individual parameters provides an insight into the actual operation of electric bus fleets. Additionally, using correlation analysis, it is possible to understand the relationship among all mentioned parameters. The analysis of the energy consumption of electric buses provided in this paper offers valuable inputs for future studies and the successful electrification of further bus fleets

The integration of renewable energy sources (RES) and new load-flexible consumers, coupled with the zonal electricity market design in Europe, presents a prospective challenge of increasing congestion situations in the power grid. In response, this paper proposes an approach for the consideration of intelligently controlled battery electric vehicle (BEV) charging as an additional congestion management measure within conventional redispatch simulations on transmission grid level. An aggregation method for individual BEV charging processes through the utilization of superimposed clustering algorithms is introduced. This allows the integration of a large number of temporally and spatially distributed charging processes into a cost-minimization based congestion management simulation under consideration of time coupling constraints. The presented methodology is implemented and tested within the Model of International Energy Systems (MILES), developed at TU Dortmund. It is demonstrated, how the introduced approach enables the analysis of the impact and optimal utilization of BEV charging control in congestion management. It is shown, that the approach allows a runtime efficient consideration of BEVs in national scale scenarios while ensuring a high spatial and temporal resolution of the adjustments.

This is a presentation of the research project SekQuaSens³ of the German Aerospace Center that started in 2023 and that will run until the end of 2025. In this project, a concept to aggregate data via networked sensors to reap the benefits of coupling transportation, heating and energy supply in a district will be developed. The decentralized, networked sensor data inputs will be used to facilitate decisions with neural networks based on simulation models, present vehicles and movement profiles of the residents to optimize energy flows in a replicated (urban) district. The sensors we want to employ will continuously send state-of-charge and capacity of each vehicle for example. In addition, especially the presence and absence of the vehicles are considered for the control of the bidirectional energy flows. The models we will combine in the project are TAPAS (Travel and Activity Patterns Simulation), SUMO (Simulation of Urban MObility), a model for district thermal energy requirements based on Modelica and oemof (open energy modelling framework) and MTRESS (Model Template for Residential Energy Supply Systems). Autonomous communication between vehicles, other stakeholders and facilities in a district will be built and validated, while evaluating the potential benefit in terms of reducing the use of fossil-based energy in different scenarios. The combined model will be used to make statements about peak load reduction and flexibility options for energy and load shifting in the district.

Net load forecasting has become increasingly complex due to the high penetration of renewables and evolving data subject to so-called concept drift, i.e., sudden and large changes in the energy flow pattern. Accurate net load forecasting is essential to prevent unexpected imbalances across all voltage levels of the electricity grid, as well as to promote the stability and reliability of the power system. Therefore, the use of accurate forecasting techniques is essential to manage and optimize the use of available resources at the TSO-DSO interface. This paper evaluates several forecasting methods, including an adaptive random forest method based on incremental learning incorporating a drift detector, a method based on a recurrent neural network using long short-term memory (LSTM), and a method based on an ensemble of models including decision trees (DT), support vector machines (SVM), extreme gradient boosting algorithms (XGBoost), and Lasso regressions. The experiment was conducted using the net load data collected at the TSO-DSO interface in Portugal, where concept drift can be observed, possibly due to increasing integration of distributed energy resources behind the meter. The study examined two scenarios. In the first scenario, the models were trained using a large training set that included significant drifts, while in the second scenario, the models were trained prior to the occurrence of the drifts. The results showed that the approach using the adaptive model is more robust to the concept drift and performs better compared to the other traditional methods, especially in scenarios where there are significant changes in the net load patterns over time.

Purpose: The potential of Artificial Intelligence (AI) has attracted more and more attention from the industry and the seemingly endless possibilities of AI are rapidly being discovered. The energy industry is undergoing a transformation from plannable conventional energy sources to volatile renewable energy sources. As volatility increases, so does the need to adjust energy consumption. Energy needs to be purchased when it is available. The potential of artificial intelligence and dynamic pricing models, which can be used to adjust cost accounting, will make a significant contribution to this necessary change. General scope of this within the framework of the EU-financed research project "KI-Werkstatt Mittelstand" is to develop an innovative AI-based platform to predict regional supply and demand of renewable energy and offer dynamic pricing based on weather forecasts, production data and consumption patterns. The project is supported by two major regional companies in the mobility sector - both offering electric cars and charging stations with their own PV systems.

Methodology/Approach/Current work: The most important aspect of training an AI is a sufficiently large set of training data. The first step for the realization of the project was to create a general AI model capable of predicting the production of a PV system and the amount of electricity fed into the power grid. For this purpose, a large dataset of training data from the past years (15 minutes cycle) was provided by the project partners, including basic data such as the size and production of the PV system and the owner's internal consumption. Combined with relevant past weather data from the owner's location, a well-fitting set of training data was created. A number of different AI models were used to train the AI, ranging from a simple linear regression model to more complicated deep learning models. Ultimately, a random forest algorithm was used, which provided accurate predictions on the training data while having a comparatively low runtime. The random forest model uses ensemble learning methods to create multiple randomly drawn decision trees and averages these results to create a strong prediction.

Major conclusions drawn: Once a reliable AI-based forecast is available for both PV power generation and PV power consumption, a dynamic pricing model is stored to ensure that in times of excess PV power generation in the region, it is offered to consumers at low cost and when there is low generation, less electricity is demanded due to higher prices.

Growing electricity consumption by electric vehicles (EV) could create a challenging situation for low-voltage (LV) grids when
high peak demand occurs. Therefore, the development of appropriate charging infrastructure and LV grids is necessary. Due to
spatial constraints and socio-economic factors, EV electricity demand and charging infrastructure options vary between different
city quarters. In this conceptual paper we design a traffic survey concept to generate data on mobility behaviour and EV electricity
consumption within archetypical city quarters, based on theoretical considerations. We review different types of traffic survey
methods to determine their characteristics and suitability for these objectives. In order to achieve these goals, the key figures
parking behaviour and distance travelled by the population are selected. These metrics can be collected using survey methods
that collect data on individual trips. Therefore, questionnaires, GPS tracking or floating car data can be used as a basis. As all of
these methods can have the disadvantage of potentially small sample sizes, validation should be carried out using survey methods
that collect aggregated data. These include licence plate recognition and machine learning based surveys.

Industries are expected to be one of the early adopters of electric vehicles and vehicle-to-grid technology. With most industries typically being characterized by low power factors, investment in various power factor compensation solutions becomes a necessity for these industries. However, DC electric vehicle (EV) chargers have the potential to inject/consume reactive power that can be exploited to the benefit of the industry. The objective of this study is to analyse the potential of a captive EVCS in providing reactive power support to an industry. A control algorithm and a controller are developed to control reactive power exchange from the EVCS to maintain the power factor (PF) at the point of common coupling between the industry and the electric utility at a predefined target. In this study, a food processing industry based in Delhi, India has been taken into consideration with a captive EV charging station (EVCS). For quantification of the benefits of reactive power support from EVCS, the electricity tariff in Delhi, India for 2021-2022 has been considered. The findings suggest that a dedicated EVCS can enhance the industry's power factor, although the extent of this support depends on the interplay between the industry's operational hours and the period and duration of EV charging, in addition to the type and rating of the chargers in the EVCS.

Flexibility of distributed energy resources (DERs) for fast demand response like secondary reserve has been highlighted to support large scale integration of renewable energy sources into the power system. To control such resources in real time, IoT (Internet of Things)-based demand response (DR) is necessary. One important issue with IoT-based DR is with regard to their full activation time (FAT) and its ICT latency. Although many studies have reported the inherent delay of various DERs, our research review has not found many integrated measurement results of both ICT latency and the inherent delay of the DER. The relationship between the ICT response time (IRT) of the DER and the FAT has not been clearly discussed. To the well understanding of both the ICT latency and FAT of DERs, an experimental system is constructed to measure both the ICT latency and the electrical delay of DERs, synchronously with the local timeserver. This paper reports the experimental system, the characteristics of measured IRT and the FAT of one electric vehicle charger as an example.

For the market ramp-up of battery electric trucks, the public infrastructure construction is one of the most important elements and the installation must take place timely. In the future, break and rest periods at rest areas will be used to charge the vehicle battery of long-haul trucks. In order to identify the site-specific charging and energy demands at rest areas, a load modelling approach was developed and transferred into a simulation model that is based on the parking behaviour of long-haul truck traffic. The focus of the approach is the transferability to other sites. The forecast for site-specific parking occupancy is based on a combination of dynamic parking data and counting point data on road utilisation and is supplemented by the derivation of parking behaviour from synthetic trip chains. Depending on the duration of the break or rest period, a charging strategy is assigned to the arriving battery electric truck, megawatt charging for intermediate charging during short breaks and combined charging system for overnight charging during longer rest. As a result, the load modelling approach provides forecasts of site-specific parking occupancy, energy demand and charging profiles of battery electric trucks.

As companies start to invest in EV fleets, the need for smart charging systems to distribute high power demand while ensuring safety of the charging infrastructure gained importance in the last years. To increase efficiency of existing smart charging systems, predictions about events in the near future can help better prioritize and schedule charging sessions. To achieve this, a short-term forecasting framework for a PhD thesis is presented in this paper that creates system states of fixed intervals in the near future, e.g. every 15 minutes in the next two hours. These system states can be considered as images of the future situation within a charging infrastructure, created by applying short-term forecasts on data gathered at a company's parking area. Integrating the system states into smart charging to detect and eliminate potential problems in a timely manner helps to react to future events, e.g., EVs can be charged despite high energy prices to additionally serve a high number of arriving vehicles later on. Therefore, suitable areas of smart charging for short-term forecasting are analyzed and the proposed system state is described, including the general idea of short-term forecasting and its subsequent integration into smart charging.

This paper aims to investigate electric vehicles (EVs) as a means for energy provisioning, with a focus on company fleets: EVs can be used as mini power plants to feed energy into a local power grid when connected. Electric fleets have the potential to access a huge number of EVs, including cars, trucks, buses, shuttles, delivery vans, etc. to participate in the energy market as a Virtual Power Plant (VPP). Aggregating multiple fleets can improve the trading of energy for fleet operators with volume discounts and enhanced consumption flexibility. The federated approach from this research framework proposes open standards to enable simple participation in the VPP. Consequential results help to process the available data from the EVs to assemble ancillary services for supporting the transition to renewable energy generation and improving the stability of the power grid.

The speed of electron is the one of the most important parameters of this particle. Another effective parameter that makes electron have a very high ability to perform various tasks is its density. Considering the magnitude of its density, it can be said that the ability of an electron to do work is due to its very high density. We have calculated the energy of 1 kg of electrons and according to the results a lot of energy can be obtained. It is efficient and replaceable in all cases where electricity is used. So, by this energy which we could be stored in electron tank, the vehicles and everything that works with electricity can be charged for months and years.

Smart-Charging is becoming an increasingly important feature in charging installations as the size and cost of infrastructure, energy, and utility connections increases. Smart charging is not a single formula, but a range of algorithms and implementations. A range of approaches described in the literature and industry are surveyed and grouped based on their inputs, outputs, and approaches.
In almost all cases the output for smart charging is the charge power for each charger over time, often sent via OCPP protocol, and these charging profiles are periodically updated as the situation evolves. A key consideration in algorithm development is what inputs are used and the uncertainty of predictions – how much does the algorithm need to predict what will happen later to control effectively? Key inputs are vehicle departure times and the required state of charge. Solar production can usually be estimated as an upper bound, but reductions in power due to clouds are unpredictable and need to be addressed in any algorithm. Likewise other loads which may subtract from available utility capacity may have typical daily trends (air conditioning is higher in the afternoon) but may vary unpredictably. Utility prices may be fixed months in advance or evolve rapidly. Vehicle movements may be highly planned in fleets, or unpredictable as passenger car drivers arrive and depart. Which inputs are considered, is one of the differentiators between algorithms.
There are varied objective functions depending on application. Maximizing revenue, minimizing energy cost, demand charges, overall electricity costs, or CO2 production are all potential objective functions. For fleets vehicles dispatching on time is an essential requirement, whereas for private users providing smart charging on a best effort basis can be acceptable, but a dissatisfier if it is inadequate.
This paper will look at a range of algorithms and their inputs, outputs, objective functions, and mathematical approaches to help evaluate suitability for given applications.

This Paper describes results of the research project “SMECON-Box” (Smart-MEter-CONtrol Box). In this project, an algorithm for a smart FNN control box, the so called SMECON-Box, is implemented. The algorithm curtails the charging power of an EV in case of grid overloading. To keep the communication needs as low as possible, the SMECON-Box should decide by itself, based on local measurements, if the grid is overloaded.
The first part of the paper describes the implementation of the algorithm and the updates, that have been applied since the previous publication. In the second part of the paper the challenges of the actual implementation of the project are highlighted. There have been several obstacles mainly related to communication between the different devices which prevented an easy implementation in the field. These obstacles and the remedial actions that have been applied to start the field tests are described here.

The rapid deployment of charging infrastructure for electric vehicles (EV) leads to increasing peak loads in the distribution grids. The Electrify Buildings for EVs (ELBE) project is therefore investigating the controllability of EV supply equipment (EVSE) by the distribution grid operator (DSO). Here, grid-friendly load reduction is communicated via OpenADR from the DSO to the charge point operator (CPO). The communication between the supply equipment communication controller and the electric vehicle communication controller on-board the EV is dependent on the charging mode. Mode 3 charging uses basic signaling via pulse width modulation, while mode 4 charging uses high-level communication via a powerline communication link. Therefore, devices are required to independently validate this controllability.
In prior work, test devices for laboratory environments, and a concept for a mobile modular test device for the combined verification of the signal chain as well as real resulting load reduction for AC and DC EVSE was presented. In this paper, the realization of the described concept as well as validating field measurements on controllable AC and DC EVSE are presented.
For this purpose, the special circumstances of the EVSE roll-out in the semi-public area in Hamburg via the ELBE project are presented, followed by a brief overview of the necessary control equipment for EVSE. Furthermore, the software and hardware concept for the mobile test device is described and the implementation in the real setup is depicted. Finally, the concept validating measurements on controllable AC as well as DC EVSE are presented and the performance and suitability of the device for mobile field tests are discussed.
The device resembles the relevant grid-related functionalities of a charging EV: Both means of communication are implemented, and a setup incorporating different electronic devices emulates the electric behavior of a charging EV in both AC and DC charging. To emulate AC charging, three single phase rectifier bridges and dynamic DC-loads in constant resistance control are used. To emulate DC charging, the same DC-loads in constant voltage control are switched in parallel with a DC-source in constant voltage control. This setup is electrically equivalent to an EV in AC or DC charging and capable of a power dissipation of up to 9kW. This is sufficient to validate the load reduction functionality and to test the basic functions of the EVSE. As the developed test device is designed for mobile use throughout the city of Hamburg, the entire setup is enclosed in modular stackable protective boxes that can be moved by a single operator.
Several validating tests have been performed in a laboratory environment. The setup is shown to be a valuable tool to validate the controllability of EVSE, not only via the signal chain, but also testing an actual reduction of charging power. The device was subsequently used in in-field test campaigns throughout the city of Hamburg. The results will be presented in this work.

The envisaged diffusion of electric vehicles and of relevant charging points is posing significant challenges on electric power system operation, in order to ensure the provision of charge service without jeopardizing the network functionality. Most of the plugging standards for private charging directly involve distribution networks, where smart charging and vehicle-to-grid, when enabled, are more and more necessary to ensure proper integration. Therefore, the distribution system operator, dealing with increasing electric vehicle charging facility connections, is in charge of individuating the most appropriate ways for electric vehicle integration. In this research paper, a technical-oriented optimization procedure is carried out, considering distribution operator viewpoint in the presence of electric vehicle diffusion, exploring the ability of smart charging and vehicle-to-grid facilities to provide loss reduction while satisfying line loading and bus voltage targets. The procedure accounts for stochastic samples of realistic electric vehicle charging requirements, in terms of time and energy targets, and includes a sensitivity-based network model in the framework of linear programming. The proposed approach is applied to a set of operating conditions of a real-sized MV radial network, inspecting the impact of different electromobility penetration levels and vehicle usages according to the siting of electric vehicle charging facilities. The analysis allows to point out the interactions of electric vehicle smart charging and vehicle-to-grid processes on the network and among electric vehicles as well, individuating the possible formation of energy community principles while dealing with distribution network operation.

This study presents a novel approach for planning and operation of a smart charging facility for electric vehicles based on user
preferences. The facility, located in Vienna, Austria, features eight chargers from different vendors, including two DC fast-
chargers and six AC chargers, as well as a dedicated photovoltaic system and a local battery storage. To customise the
charging experience and ensure an optimised operation of the charging facility, users utilize a mobile application to specify
their preferences for the charging (e.g., desired end time, desired state of charge at the end of the session, etc.). The facility is
operated based on simulated cost functions for energy, aiming at minimising overall costs while meeting user preferences and
grid constraints. Human-entered user preferences are first checked against correctness (i.e., data consistency and
completeness) and plausibility (i.e., the inserted data meets the specification of the infrastructure). Plausible user-preference
are finally analysed to check if the request was indeed fulfilled. Two months of data show that 87% of total charging events
were plausible, and 59% of plausible events eventually fulfilled the users' preferences.

The interest in the electrification of the African transport and mobility sector with the integration of renewable energy sources is increasing in order to reduce harmful emissions. However, off-road machinery and tractors, which are key to increasing productivity in the agricultural sector have barely been considered in the past. This paper shows an overview of the state of technology of the first industry concepts for battery-electric tractors, which are already available. In addition, it shows how they can be classified in comparison to common diesel-driven tractors in the African market. The comparison reveals that none of the tractors exploits the potential of purpose design for the electric drive train in combination with the power supply infrastructure. However, for long-term viability of battery-electric tractors, a holistic concept is needed. Therefore, the authors propose a morphological box with the design variables of a purpose-built battery-electric tractor and show the technical solutions necessary to derive a holistic concept.

The stringent requirements set by the EU to reduce the environmental impact of the transportation sector require a shift in the mobility paradigm. Electric Vehicles (EVs) will be at the forefront of the energy transition, hence in this article we look at the current and future predicted status of electro-mobility in the EU. Leveraging both market and policy-based scenarios, the current and future number of EVs in the EU, and in each country, are shown. A country-specific analysis on Italy, Spain, and Denmark, the three countries where the FLOW’s project demonstrations, are also conducted. The EV market in the EU is rapidly developing, but it is still far away from both the net-zero emissions and policy targets. The geographical distribution is heterogeneous, with northern European countries ahead of the others in terms of EV adoption. In the three analyzed demo countries one in 2-8 drivers should own an EV by 2030 and the recommended publicly available charging power levels are already met, even though the current values are still far from the 2030 targets. Increases in electricity demand are foreseen by 2030 for e-mobility, especially if we consider the electrification of the goods transportation sector.

When looking at the electric mobility, many countries have advanced discussions on charging methods and infrastructure, grid integration modelling, the use of artificial intelligence and machine learning, among other technologies. Nevertheless, countries such as Brazil are significantly behind schedule, technologically outdated and have undefined regulatory and legal frameworks. This article highlights the struggle to diverse the technological route for transports in Brazil and presents findings from an national R&D project, which aims to install fast chargers for electric vehicles on highways in Brazil. It mainly focuses on regulatory aspects and business model, pointing out: (i) high costs of infrastructure versus the low financial return, given by the initial simulations, (ii) the complexity of obtaining authorizations, due to piecemeal approach to infrastructure deployment and lack of knowledge and process harmonisation, (iii) the existing electrical grid cannot handle the additional load, and (iv) the lack of consumers demand and interoperability. In order to face these challenges, achieve Brazil’s climate goals, and decarbonize the transport sector, we conclude that a national strategy and collaborative approach is urgent, since EV’s initiatives across the country are decentralised. Finally, there is a need for legal frameworks, standards and regulations to ensure security and encourage private investments.

The most diffused UFC infrastructures can be classified as standalone and satellite charger architectures. Generally, standalone can be considered as independent solutions made of monolithic power converters, directly supplied by the grid. Otherwise, the power conversion units in the satellite ones are modular and distributed, the conversion capability could be shared from the several chargers. The design of satellite architectures offers several degree-of-freedom in terms of the modularity of converters and the topology of the distribution system.
At present, there is not a common reference design criteria to choose the level of granularity of power conversion units and the topologies of distribution units in satellite infrastructures. The main manufacturers design their products based on the production scale economy, but other sizing criteria could improve the overall dependability of the systems.
A design case study has been proposed to compare distributed and concentrated solutions in terms of technical and economic matters, taking into account indexes of efficiency, reliability and performability. This comparative study results in preliminary design guidelines for multi-slot Ultra-fast chargers.

Designing the infrastructure of bus depots involves numerous factors and considerations, but it is often subject to uncertainties that can affect the efficiency, cost, and overall performance of the depots. This study analyzes various sources of such possible uncertainties encountered during the design phase of bus depots and highlights their potential impact. Generally, uncertainties in bus depot infrastructure design can arise from several aspects, including technological advancements, regulatory changes, financial constraints, and evolving operational requirements. The adoption of emerging technologies, such as electric buses, introduces uncertainties regarding the charging infrastructure, energy storage capacity, and compatibility with existing depot layouts. This study considers operational uncertainties, such as changes in the loading of transformers, occurrence of blackouts, which consequently pose challenges to depot design. This is realized by employing many sensitivity case studies to evaluate various operation and design options under different uncertainty scenarios. The performed analysis in this study can be used in calculating the loading of transformers of bus depots in advance. Additionally, it is possible to estimate the required stationary battery in the bus depot for supplying the buses during different blackout times.

Purpose: The individual mobility sector is undergoing a transformation from internal combustion engines to electric vehicles, as the EU has forbidden registering any new car with an internal combustion engine after 2035 to reduce emissions. While the mobility sector is moving more to electric vehicles, the electricity demand and volatility increase. Ideally, for those who can individualise parts of their electricity consumption, that is, charging electric vehicles, electricity should be consumed when there is low consumption and high production of renewable energy to reduce emissions and volatility. To optimally adjust usage, a prediction of future energy consumption would be useful. Within the EFRE-financed research project "KI-Werkstatt Mittelstand" the goal is to create a platform for dynamic electricity pricing based on predicting energy consumption and PV production. The research project is supported by two regional SMEs that offer electric car charging stations and also use them for their own electric vehicle.

Methodology/Approach/Current work: To predict the energy consumption of the industry partners, a suitable amount of training data must be acquired. As consumption patterns differ for each of the locations of the industry partners, each location needs its own trained AI model and training data set to be able to capture individual consumption patterns. For training those models, past data from multiple years was used. The consumption data was provided by the project partners. For each hour, the energy consumed is included in kWh. The prediction was realised with a time series AI, which is able to pick up patterns based on the consumption within an hour, day, or month cycle.

Major conclusions drawn: Once a reliable time-series AI-based energy consumption forecast is available for each of the locations, predictions can be used to adjust consumption behaviour to lower cost and emissions. Furthermore, by combining consumption prediction with predictions for PV plant energy production, a regional renewables market place can be built, including dynamic electricity pricing based on these predictions.