Dynamic pricing scheme for energy balancing in microgrid using an intelligent system

Abstract

The random nature of both renewable energy supply and demand patterns poses a significant challenge to energy balancing in microgrids. This unpredictability impacts key power system parameters, reflecting real-time imbalances between supply and demand. Such fluctuations can affect the performance of dynamic pricing, which plays a crucial role in optimizing real-time energy usage within a microgrid environment. Traditional pricing models fail to adapt to these variations, resulting in inefficiencies such as overconsumption during peak hours and underutilization during low-demand periods. These limitations hinder efforts to maintain grid stability, cost-effectiveness, and resource sustainability. To address these challenges, an intelligent pricing approach is essential for effectively responding to fluctuating system conditions. Therefore, this research proposes a dynamic pricing scheme based on real-time variations in supply, demand, and the state of charge (SOC) of a battery. An exponential pricing model is developed using key factors such as the supply–demand ratio and SOC. To enable price forecasting, training data is generated using this model under a 5 kW load during peak and off-peak periods. The data is then used to train and test machine learning algorithms, including Support Vector Regression (SVR), Random Forest (RF), and Artificial Neural Networks (ANN). Among these, ANN demonstrated the highest accuracy and effectiveness in predicting energy prices. To implement and validate the proposed AI-based pricing approach; a MATLAB-based hybrid microgrid system has been developed. This system consists of a 2 kW solar source, a 240 V, 100 Ah lithium-ion battery, and a 3 kW grid connection. The results show that this approach enables faster, more accurate and more efficient pricing, thereby facilitating improved energy balance and benefiting both suppliers and consumers.

Introduction

India, which has traditionally relied on fossil fuels, is now facing growing environmental challenges such as carbon emissions and resource depletion. Because of this, finding new ways to generate clean energy is crucial for promoting sustainable growth. An integral part of India’s energy revolution is the move into Renewable Energy Sources, which include solar and wind power etc. The inherent problems of renewable power systems, such as intermittency and variability, make grid management and demand–supply balance more complex than with traditional power systems. Because of these problems, sophisticated control systems that can adjust to changes in energy generation are required¹. At the same time, achieving a balance between the three competing goals of affordable, sustainable, and secure energy is crucial to India’s energy future. To achieve this equilibrium, we must minimize our reliance on fossil fuels and maximize our use of renewable energy sources, which in turn minimize our carbon emissions. An essential part of India’s energy transformation will be micro grids that have adaptive technology, dynamic pricing models, and smart control systems². Strong energy management strategies utilizing demand response, machine learning, and artificial intelligence (AI) are necessary to make this vision a reality. With these advancements, energy flows can be precisely controlled, resources can be optimally used, and grid stability can be guaranteed, even when renewable power sources are intermittent. One potential response to the challenges posed by climate change and limited resources is the rise of localized energy systems, such as micro grids. Micro grids especially when coupled with renewable energy sources, allow for more localized energy generation, less transmission losses, and improved grid resilience. Stabilizing power infrastructure, reducing climatic impacts, and fostering energy independence are all made possible by these systems³. As an example, grid-connected micro grid shown in Fig. 1 offer significant benefits because they can easily connect to main grids. They allow for real-time energy control and ensure backup during grid outages, all while improving efficiency and dependability. Even in the face of unpredictable and ever-changing grid conditions, an efficient energy management system (EMS) developed for grid-connected micro grid can maximize the benefits of renewable generation while keeping the grid stable⁴. In addition, micro grids that have deftly drawn boundaries can accommodate both standalone and networked operations, meeting the ever-changing demands of both urban and rural areas⁵. Particularly promising for energy independence are urban micro grids, which can make use of local resources while decreasing dependence on centralized networks and environmental concerns. By producing clean energy, these technologies not only make cities more resilient, but they also encourage sustainability⁶. Despite the challenges posed by renewable energy sources in micro grids, dynamic pricing is essential for real-time energy use. A new effective technique for energy management, dynamic pricing allows users to adjust their energy consumption based on the current state of the grid. Dynamic pricing models promote energy efficiency and aid grid stability by modifying customer costs in response to changes in supply and demand.

Fig. 1.

Structure of a grid-connected Microgrid.

These systems promote demand response programs, which mitigate intermittency by adjusting user consumption habits to coincide with the availability of renewable power⁷. A conceptual foundation for balancing grid operations and limiting resource wastage is provided by the synchronization of everyday energy supply and demand through dynamic pricing⁸. Micro grids have achieved amazing success by demand response in an energy management system. These systems optimize the utilization of renewable energy sources, enhance grid stability, and instantly the supply and demand gets balanced. Operational efficiency, consumer needs are both met by demand response, which aligns user consumption with pricing schemes. By utilizing Energy Management System technologies, grid-connected micro grids can integrate renewable energy sources seamlessly, guaranteeing reliable operations regardless of environmental fluctuations⁹. Dynamic pricing helps maintain a stable grid and save energy by taking into account factors like technology limits and usage trends. In order to provide grid stability and promote energy efficiency under changeable conditions, such methods are crucial for micro grids that incorporate renewable energy sources. The energy management techniques of today are being radically altered by the recent innovations in technology, particularly the ML. Enhancing efficiency and decreasing operational expenses; these solutions make possible real-time data collecting, predictive analytics, and automated energy control. Micro grids are able to adapt to changing conditions and make better use of their resources with the help of machine learning algorithms that optimize energy flows by integrating data from consumption, storage, and generation¹⁰. The integration of machine learning (ML) with environmentally sustainable technologies significantly accelerates the global energy transition. ML techniques enhance system efficiency, optimize resource allocation, and enable intelligent control of distributed energy systems. This not only leads to cost reduction but also contributes to substantial reductions in carbon emissions, aligning with global sustainability goals. In particular, urban microgrids benefit immensely from recent technological advancements that support the utilization of local renewable resources, such as rooftop solar and community-based storage. The integration of machine learning (ML) with environmentally sustainable technologies is playing a transformative role in accelerating the global energy transition. ML techniques enhance system efficiency, optimize resource allocation, and enable intelligent control of distributed energy resources. These advancements not only reduce operational costs but also contribute to significant reductions in carbon emissions, aligning with global sustainability goals. In particular, urban microgrids benefit substantially from recent innovations that enable the effective utilization of local renewable resources such as rooftop solar and community-based storage. Recent studies have shown that particle swarm optimization (PSO)-tuned XGBoost significantly improves forecast accuracy compared to classical approaches such as ARIMA and long short-term memory (LSTM). Other research has reported that simpler models, including LEAR and XGBoost, often perform more reliably than deep learning methods, which tend to struggle with abrupt price fluctuations¹¹. The effectiveness of hybrid approaches has also been highlighted, where combining XGBoost with ARIMA was found to enhance forecasting performance compared to ARIMA alone¹². These findings reinforce the growing preference for ensemble techniques such as XGBoost, which have emerged as powerful tools for electricity price prediction due to their ability to handle non-linear data interactions, manage missing values systematically, and mitigate overfitting while maintaining high scalability¹³.Further advancements have been proposed by integrating XGBoost with error-correction features, Bayesian hyperparameter optimization, and explainable AI methods such as SHAP and LIME, which achieved substantial reductions in forecasting error on ISO-NE market data while improving interpretability for stakeholders. In addition, reinforcement learning-based energy management systems have been developed, where Soft Actor–Critic with normalizing flows was employed to dynamically tune ADMM penalty parameters, resulting in accelerated EMS convergence and enhanced demand response capability under conditions of variable renewable generation¹⁴. Together, these studies illustrate how the integration of machine learning, explainability, and reinforcement learning is advancing the accuracy, transparency, and adaptability of modern energy management and pricing systems. These innovations reduce dependency on centralized utility grids, thereby lowering transmission losses and promoting energy decentralization.

Moreover, ML-driven microgrid technologies contribute to solving various urban environmental challenges, including pollution control, grid instability, and peak load issues. By enabling real-time adaptive management and predictive analytics, these systems can proactively respond to demand fluctuations and external conditions. Beyond improving operational efficiency, such technologies also foster renewable energy self-sufficiency and strengthen the resilience of urban infrastructures to disruptions—whether from grid failures, natural disasters, or energy market volatility¹⁵.

In today’s energy management systems, artificial intelligence (AI) plays a pivotal role in achieving intelligent, automated, and adaptive control. AI-powered algorithms are increasingly used to automate energy optimization operations, enabling real-time coordination between renewable energy sources, energy storage systems, and demand response mechanisms. These intelligent control systems leverage data-driven techniques to dynamically balance supply and demand, adapt to uncertain generation patterns, and optimize operational efficiency. By incorporating predictive analytics and adaptive decision-making, AI technologies contribute significantly to enhancing microgrid performance, reducing operational risks, and fostering sustainable energy usage^16,17.The present research proposes a novel dynamic pricing strategy developed using the MATLAB platform, aiming to improve microgrid efficiency through real-time pricing adaptation. The strategy integrates machine learning models, particularly artificial neural networks (ANN), to determine electricity prices dynamically based on system parameters such as battery state of charge (SOC) and the supply-to-demand ratio etc. as illustrated in Table 1.

Table 1.

Abbreviations, symbols and terms.

Abbreviation	Full form
R	Ratio
SOC	State of charge
SVR	Support vector regression
RF	Random forest
ANN	Artificial neural network
AI	Artificial intelligence
EMS	Energy management system
ML	Machine learning
PLL	Phase locked loop
MPPT	Maximum point power tracking
NiMH	nickel-metal hydride
PV	Photo-voltaic
A	Difference of solar and grid power
	Tuning parameter for ratio
	Tuning parameter for SOC
D	Demand
S	Supply
S/D	Supply to demand ratio
	Maximum capacity of power
T	Time in hours
Po, i	Average cost for energy generation by source
	Time constant
	Total net energy
	Net production of energy through energy source
P	Price
	Low price
	High price
KW	Kilo-watt
KWh	Kilo-watt-hour

Unlike traditional pricing mechanisms, which are often static or slow to respond to real-time fluctuations, the proposed approach ensures pricing that is responsive, stable, and economically efficient. Simulation results demonstrate that the ANN-based pricing model outperforms conventional methods and other algorithms, delivering more accurate predictions and smoother price transitions. This makes the proposed model a robust and practical solution for modern micro grid energy management, where intelligence, responsiveness, and economic viability are essential.

The structure of this work is as follows: The Introduction is in Section I, while the Related Work is in Section II. Section III describes the Problem Statement, while Section IV gives about existing and suggested methods. The dynamic pricing strategy is explained in Section V. Section VI discusses the Dataset and Simulation; Section VII gives the Simulation Results and Performance evaluation. Then, Machine learning integration is covered in Section VIII, and the task is eventually concluded in Section IX.

Related work

Modern developments in pricing and energy management have brought new approaches to improving customer involvement and streamlining grid operations. In order to optimize demand responsiveness in the face of renewable energy variations, differentiating pricing methods were investigated. However, there are still obstacles to adopting real-time adaptive pricing that has improved prediction skills. Not enough attention given to multi-period optimization techniques, which are crucial for the stability of the system in the long run. It is even more important to have a model that can adapt to changing prices by using renewable energy data in real-time¹⁸. Smart micro grid energy balancing adaptive control technologies have solidified dynamic pricing as a crucial tool. However, concerns about scalability in larger grid configurations and modifications made by customers remain unresolved. It is essential to have scalable adaptive pricing models in order for these strategies to work with different types of energy systems. Energy modifications for end users and greater grid-level requirements should both be accounted for in these models¹⁹. These issues bring to light the necessity for an all-encompassing plan that links micro grid operations with the broader dynamics of the energy market. Micro grid powered by renewable energy sources may be managed via dynamic price-based demand response. Although this method successfully optimized micro grid operations, it lacks the scalability and adaptability that would be achieved with further integration of predictive and adaptive methods²⁰. The studies, however, failed to delve deeply enough into cutting-edge tech that has the potential to radically improve pricing strategies’ precision and flexibility, such as ML algorithms and artificial neural networks (ANN).The use of these technologies could pave the way for micro grids to dynamically alter prices in order to accommodate significant penetration of renewable energy sources²¹. Future development must focus on integrating ML-based prediction with adaptive algorithms. Although studies on residential demand-side dynamic pricing have provided valuable insights, they have not yet fully accounted for renewable energy integration and storage systems. To tackle this, energy management solutions need to be cohesive and effective, which means using integrated pricing models that optimize storage while simultaneously combining renewable energy forecasts with demand response²². Such integration would also lend credence to a wider adoption in a variety of residential settings. At the same time, virtual energy storage systems that use real-time pricing have shown promise in terms of storage capacity optimization. But there is still no solid framework that takes user-level demand patterns and the unpredictability of renewable generation into consideration. So, use of user-centric demand response techniques linked into real-time pricing models is necessary to better match energy generation and consumption in renewable energy systems²³. Grid stability and energy efficiency could be greatly enhanced by this alignment. Incorporating AI into the energy management sector will open up new and exciting possibilities. While their approach did improve energy balance and demand responsiveness, it might be much more effective with the use of hybrid ML models and predictive analytics²⁴. Not much has been done to encourage user adaptation to demand response schemes or to integrate pricing systems that can react to changes in renewable energy sources. Integrating these models with adaptive pricing systems and renewable energy forecasts could lead to more demand flexibility and improved energy balance in micro grids. The accuracy of pricing projections, especially in systems that incorporate renewable energy sources, has not improved significantly despite the adoption of sophisticated machine learning algorithms. For more precise and versatile predictions, ML-based methods like hybrid models show a lot of potential. Building upon this work, future energy sector advances may develop sophisticated hybrid systems that combine the benefits of many algorithms to improve dynamic pricing projections for energy storage systems even further²⁵. Hybrid ML models used for energy price prediction have not yet adequately addressed real-time dynamic demand response. By incorporating real-time demand response into hybrid ML-based price estimates, energy management approaches could become more responsive and users could be more engaged²⁶.Although hybrid regression models that combine statistical and machine learning techniques have proven effective for day-ahead electricity price forecasting, several limitations remain. A key challenge lies in their lack of real-time flexibility and limited ability to integrate demand response dynamics. Operating largely on static data inputs and fixed time horizons, such models are often unsuitable for applications requiring instantaneous adjustments to fluctuating grid conditions. Recent advancements in short-term price forecasting have attempted to address these issues. For example, error-correction features were integrated into an XGBoost model, with parameters optimized through Bayesian search and interpretability achieved using SHAP/LIME. While this approach improved forecasting accuracy and transparency in wholesale market contexts, it did not incorporate the forecasting outputs into a real-time control or pricing feedback loop¹¹. Similarly, microgrid operation optimization has been advanced by accelerating the convergence of ADMM-based scheduling through the learning of optimal penalty parameters with Soft Actor–Critic and normalizing flows, thereby enhancing demand response and reducing reliance on diesel generators¹⁴.To move beyond these forecasting- and optimization-centric frameworks, real-time adaptive pricing models have emerged as a promising alternative. By enabling immediate responsiveness to changes in generation, consumption, and storage status, they enhance both the reliability and stability of the power grid²⁷. Unlike static forecasting approaches, adaptive models continuously update price signals in response to instantaneous grid fluctuations, thereby improving demand–supply balancing and mitigating risks associated with volatility and uncertainty. Moreover, the integration of demand response with intelligent storage management allows these models to utilize renewable resources more effectively, reduce dependency on costly peaking units, and optimize operating costs. When combined with machine learning, real-time adaptive pricing can further capture nonlinear system behaviours and learn from evolving operational conditions, making them scalable from localized micro grids to large-scale smart grid infrastructures. This evolution toward real-time adaptive frameworks represents a crucial step toward the development of intelligent, resilient, and consumer-centric energy markets. Building on this foundation, the proposed model advances real-time adaptive pricing by integrating demand response mechanisms with intelligent energy storage management, ensuring efficient energy utilization and accurate alignment of pricing signals with grid requirements. Specifically, machine learning in particular; Artificial Neural Networks (ANN) is employed to predict electricity prices with high accuracy and adaptability. Unlike traditional methods, the model dynamically incorporates real-time system parameters such as the supply–demand ratio and Battery State of Charge (SOC) into the forecasting process. This enables a continuous and responsive pricing strategy, enhancing coordination between generation sources, storage systems, and loads. Consequently, the proposed approach not only overcomes the limitations of conventional price prediction techniques but also establishes a scalable and intelligent pricing framework that aligns seamlessly with the evolving needs of smart grid infrastructure.

Problem statement

The electrical grid’s growing incorporation of renewable energy sources, especially solar power presents key challenges such as balancing energy supply and demand, and ensuring economic feasibility. In order to keep the grid stable and efficient, innovative energy management solutions are required due to the intermittent nature of renewable energy and varying customer demand¹⁹. When there are sudden changes in supply and demand, traditional control systems like PID controllers have a hard time keeping up, which causes operational expenses to rise and efficiency to suffer²⁸. Moreover, existing pricing models fail to dynamically adapt to real-time changes in demand, supply, and SOC of battery, resulting in suboptimal energy utilization. When demand exceeds supply, the price goes up, and vice versa when supply goes down²⁹. Calculating energy pricing is a complex and time-consuming process that involves manually recording meter readings before the price can be computed. This often requires physically collecting data from households, causing delays and might be less accuracy. Hence, the proposed system aims to bridge this gap by dynamic pricing concept combined with coordinated energy source management. This is implemented through simulation, which generates results more quickly and accurately. Furthermore, the implementation of an AI-based dynamic pricing system enables real-time cost calculation based on actual energy consumption patterns and system conditions. Unlike traditional time-block or flat-rate pricing mechanisms, the proposed model allows prices to be computed and updated within seconds, providing consumers with instant visibility into their current energy usage and associated costs. Through smart devices or energy management dashboards, users can continuously monitor this information, empowering them to make informed decisions about their consumption behaviours in real time. This real-time introduces a new level of efficiency, transparency, and user convenience. Consumers can proactively respond to pricing signals reducing or shifting their load when prices are high and optimizing their usage during low-cost periods. For example, during peak demand hours, the model automatically increases prices to reflect grid stress, discouraging unnecessary usage. Conversely, during off-peak periods, if system usage is unexpectedly high, prices may still increase, ensuring that the grid remains balanced and economically optimized. This behaviour is governed by the proposed exponential pricing function, which adapts to both the supply–demand ratio and the state of charge (SOC) of the energy storage system. From a technical standpoint, this level of adaptability is achieved through machine learning algorithms, particularly Artificial Neural Networks (ANN), which process historical and real-time data to forecast prices accurately and adaptively. The result is a highly responsive, usage-based pricing framework that reflects actual grid conditions and consumer behaviour, rather than relying on static assumptions or predefined schedules. This system benefits both consumers and suppliers: consumers gain cost-saving opportunities and greater control over their energy usage, while suppliers benefit from improved demand predictability, grid stability, and economic efficiency. Moreover, the model’s simplicity and automation make it suitable for deployment in residential, commercial, and industrial microgrids. Ultimately, this AI-powered dynamic pricing model represents a transformative shift toward intelligent energy systems, paving the way for more sustainable, resilient, and customer-centric power distribution networks.

Existing and proposed work

Renewable energy sources, due to their intermittent nature and varying consumption patterns, pose considerable hurdles when integrated into smart micro grids in terms of balancing supply and demand. Although conventional control systems such as Proportional-Integral-Derivative (PID) controllers can manage continuously changing energy conditions, they are often inefficient and limited in adapting to dynamic scenarios. In contrast, real-time dynamic pricing mechanisms offer the potential for more economically feasible and efficient resource allocation in energy systems¹⁹. However, despite the superior forecasting capabilities of modern machine learning (ML) techniques such as Artificial Neural Networks (ANN) and Support Vector Regression (SVR) over traditional methods, their practical implementation faces challenges due to data availability constraints and high computational complexity²⁹. At the same time, pricing mechanism is not determined through machine learning approaches. To address this, the proposed system integrates a photovoltaic (PV) array with a single-phase grid through an inverter and a boost converter, ensuring grid synchronization and efficient energy conversion³⁰. Energy harvesting efficiency is increased by PV array’s which has the ability to boost the energy in response to changing weather conditions by using Maximum Power Point Tracking (MPPT) algorithms³¹.. To further improve power generation and system adaptability, this approach enhances the system’s ability to respond dynamically to environmental changes, optimizing energy distribution across solar, grid, and battery³². A lithium-ion battery serves as the energy storage unit, playing a crucial role in maintaining system reliability and extending operational lifespan. The system operates within a 40% to 100% state-of-charge (SOC) range, dynamically managing charging and discharging to stabilize power distribution and balance supply and demand fluctuations^33,34. Compared to traditional batteries, lithium-ion technology offers superior energy density, efficiency, and longevity, ensuring consistent performance while reducing the environmental impact of energy storage³⁵. The battery operates in charge–discharge cycle by analysing system parameters such as Ratio which is due to variations of supply and demand. Now, to obtain the accurate price, the above system is implemented which comprising a 2 kW solar source, a 240 V, 100Ah lithium-ion battery, with a 3 kW grid connection. Through this system, real-time electricity prices are adjusted based on changing energy conditions. This model adapts energy pricing according to SOC levels and the supply-to-demand ratio, ensuring economically optimized energy distribution and system stability without using controller³⁶. Unlike static pricing models, which fail to account for real-time grid conditions, the proposed pricing strategy dynamically adjusts costs based on Ratio-SOC-tuned parameters, aligning with recent techno-economic evaluations of hybrid renewable energy systems³⁷. Grid synchronization is maintained through a robust Phase-Locked Loop (PLL), which ensures stability by continuously adjusting phase and frequency variations^38,39. Grid stability, operational costs, and energy efficiency can all be improved by incorporating AI-driven forecasts, according to research. Additionally, machine learning is employed for the prediction of price because Machine learning algorithms, provide greater accuracy and flexibility by efficiently managing non-linearity, compared to others controllers, which have difficulty functioning in dynamic situations. In order to train ML models that improve system efficiency, the simulation model produces dataset that include Ratio, SOC and Price. To facilitate better decision-making and the advancement of micro grid technologies, the model presents a new pricing structure that incorporates these elements into a dataset. More sustainable and feasible results can be achieved with dynamic pricing mechanisms in Micro grid operations. This will increase their efficiency, accuracy, and adaptability. In contrast to proportional-integral-derivative (PID) controllers, dynamic pricing models adapt prices in response to changes in supply, demand, and the percentage maintenance of charge in the battery. The majority of the current literature uses static pricing models but the battery’s state of charge changes has not been considered in real time. Furthermore, there is a lack of research comparing the efficacy of ML-based techniques with conventional PID controllers for managing the dynamics of energy systems. To dynamically reflect real-time swings in energy prices, this model incorporates two changeable parameters that dynamically showed the variation in price. Then it has been analysed with algorithms like SVR, RF and ANN which in turn guarantees the robust modelling and precise error predictions, making it a significant improvement over more conventional approaches. The suggested method fills voids in the literature by incorporating many new features, while the dynamic pricing model strives to facilitate equitable and long-term adjustments to energy costs. Finally, by integrating AI-based model with pricing mechanism, the proposed system ensures a reliable, cost-effective, and adaptable micro grid framework as compared in Table 2. These improvements set the stage for a sophisticated energy management system that can easily execute dynamic pricing strategies in future that are efficient as well as flexible.

Table 2.

Comparative table: dynamic pricing and AI in microgrid systems.

Aspect	Reference							Proposed
	36	37	38	45	46	47	48
Objective	Evaluate evolutionary algorithms under RGDP-DR	Optimize MG using 4 DR models with GRSO	Analyse TDP with variable horizons	Review AI in MG integration	RE forecasting using ML/DL	Compare ML/DL for short-term price prediction	MILP-based MG energy scheduling	Real-time dynamic pricing using SOC & supply–demand ratio
Pricing strategy	RGDP-DR (elasticity-based)	CPP DR models	Time-dependent pricing model	Not included	Not included	Price forecast only	TOU-based dispatch	Exponential pricing function
AI/ML usage	Dandelion optimization	Greedy rat swarm optimizer	No AI	ANN, DNN, GA, MAS (reviewed)	CNN-LSTM	ANN, LSTM, CNN, ARIMA	No AI (MILP only)	ANN, SVR, RF
Battery SOC role	Not involved in pricing	Not linked to pricing	Not considered	Conceptual only	Not used	Not used	Used in dispatch, not pricing	Central to pricing & logic based
Real-time operation	Scenario-based planning	Hourly optimization only	Not real-time behaviour focused	Not implemented	No real-time capability	Not implemented in real-time	Hourly scheduling only	Fully real-time loop using (MATLAB)

Dynamic pricing strategy

According to our model requirement, a Dynamic Pricing Scheme is used to ensure that the price fluctuates in real-time, which is based on Ratio and SOC as drawn as block diagram in Fig. 2. Since Indian Rupee is the default national currency for energy in India, it has been determined that this model’s price would be set in INR. This option aligns with the practical implementation of dynamic pricing systems by India’s Distribution Corporations (DISCOMs), as they typically calculate energy bills in Indian Rupees (INR). For this reason, we have provided the suggested prices for both solar and grid power as mentioned in Table 3. Here, the minimum and maximum price have been set at 4.8 and 10 Indian Rupees, respectively, in accordance with the government of Tamil Nadu, India as on 1^st July 2024. In order to encourage sustainable energy management and efficient energy consumption, these pricing techniques are crucial. This framework is based on regional policy documents of the Central Electricity Regulatory Commission (CERC) of the Indian government. The influence of dynamic pricing on energy management has been particularly beneficial to Micro grid, since power rates may be adjusted swiftly in reaction to shifts in supply and demand. Energy efficiency, grid operations stability, and renewable energy integration can all be enhanced with the ability to dynamically adjust prices. Prior studies have explored various approaches to dynamic pricing. A Renewable Generation-Based Dynamic Pricing Demand Response (RGDP-DR) scheme was proposed using a dandelion optimization algorithm to reduce cost and emissions in grid-connected micro grids; however, the model did not incorporate state-of-charge (SOC)-based pricing or predictive intelligence⁴⁰. An advanced micro grid optimization framework was introduced using four demand response models exponential, hyperbolic, logarithmic, and critical peak pricing (CPP) combined with the Greedy Rat Swarm Optimizer (GRSO) to minimize generation costs and improve the load factor. Nevertheless, SOC was not included as a dynamic pricing variable, nor were AI-based price forecasting techniques employed⁴¹.Time-dependent pricing (TDP) with variable announcement horizons was also investigated to manage procurement risk and enhance consumer flexibility. Yet, the model remained limited to economic risk modelling and did not consider real-time system parameters such as the supply–demand ratio or battery SOC⁴².To enhance microgrid efficiency and alleviate stress on the utility grid, a more responsive and intelligent pricing strategy is essential one that actively encourages demand shifting and resource optimization by adapting in real time to system dynamics. Prosumers to install photovoltaic systems on their properties⁴³. Grid reliability and prosumers alike can reap the benefits of this approach. Another study suggested a pricing mechanism that changes over time. Users can adjust their energy use in response to price signals using linear regression-based technique, which enables effective demand response⁴⁴. In addition, dynamic pricing for power trading facilitates the optimization of regional energy markets through real-time energy exchanges between producers and consumers⁴⁵. The following Table 4 presents a comparison of various pricing strategies (Fig. 3).

Fig. 2.

Block diagram of dynamic pricing based on ratio and SOC.

Table 3.

Parameters used for simulation of 5KW power.

Energy source	Percentage contribution	(KW)	T(hours)	Po, i (INR/KWh)
Solar	40	2	7	2
Grid	60	3	24	4

Table 4.

Comparison of various pricing strategies.

Terms	Flat rate pricing	Time of use pricing	Real time pricing	Exponential pricing
Definition	A constant price per kWh regardless of time or demand	Price varies by pre-defined time blocks (peak/off-peak hours)	Price updates dynamically based only on supply and demand	Price is determined by both Ratio and SOC
Ratio	Not considered	Not considered	Not considered	Considered
SOC	Not considered	Not considered	Sometimes considered in scheduling, rarely in pricing	Directly included as a driver of price
Adaptability	Not adaptive	Semi-adaptive	Adaptive to demand/supply	Highly adaptive, with tunable parameters
Advantages	Simple, easy to implement, predictable bills	Encourages load shifting, reduces peak load	Reflects true system cost, improves grid efficiency	Real-time responsiveness, SOC-aware, flexible, scalable with ML

Fig. 3.

Flowchart of dynamic pricing strategy.

In contrast to these approaches, our research introduces a novel AI-integrated exponential pricing model that dynamically adjusts the price of electricity based on both supply–demand ratio and real-time battery SOC, which are key indicators of grid stress and storage availability as illustrated in Fig. 4. Furthermore, by employing Support Vector Regression (SVR), Random Forest (RF), and Artificial Neural Networks (ANN) on synthetically generated data for a 5 kW system during peak and off-peak conditions, we not only forecast dynamic prices but also determine that ANN yields the most accurate results. This intelligent pricing framework, validated on a MATLAB-based hybrid micro grid consisting of 2 kW solar, 240 V 100 Ah battery, and 3 kW grid, demonstrates improved system responsiveness, efficient pricing adaptation, and better resource utilization offering a more integrated and real-time pricing solution than those proposed in earlier works.

Fig. 4.

Ratio and SOC elasticity curve with price.

The proposed pricing system employs an Exponential Model that dynamically adjusts energy tariffs based on the supply–demand ratio and battery’s State of Charge (SOC). By incorporating two adjustable parameters, alpha (α) and beta (β), this model offers a more responsive and consistent pricing mechanism compared to traditional linear models. The values of α and β can range between 0.1 and 0.5, allowing for flexibility in price adjustments. These parameters play a critical role in managing the sensitivity of price variations in response to changes in supply, demand, and SOC. we have carefully evaluated across multiple scenarios with varying α and β values. Each scenario provides insights into different pricing strategies and their impact on energy management efficiency. By fine-tuning, we have selected alpha (α) and beta (β) as 0.3, ensuring an optimal balance between supplier profitability and customer affordability. These parameters are fundamental in shaping the dynamic pricing structure, directly influencing price adjustments within the defined minimum and maximum limits. While α and β can be fine-tuned beyond 0.5, but their optimization must be strategically aligned with both economic feasibility and system constraints to maintain a sustainable pricing mechanism. Therefore to achieve a flexible price, our proposed model utilized simulation-generated data as a proxy for real-time data, enabling a more rigorous and structured analysis of price fluctuations. Moreover, this approach enhances the pricing accuracy as shown in Fig. 5, while ensuring the scalability and adaptability to various energy market conditions. Also this pricing strategy makes it as a better solution for the prediction by considering the micro grid applications that can be integrated into future energy management system. This intelligent pricing mechanism enhances market efficiency, ensuring a fair and economically viable energy distribution model for evolving global energy demands.

Fig. 5.

Impact of electricity supply and demand on pricing.

Pricing formula:

The following function describes how an exponential function is used to determine the pricing.

^{-----------−1}.

Where,

A = Difference of Solar and Grid power.

Tuning factor for ratio which impacts price.

= Sensitivity of SOC to pricing.

Ratio = impact of the supply-and-demand ratio on dynamic pricing in real time.

The supply–demand ratio’s exponential effect

When the ratio is greater than 1, it indicates an excess supply, causing the price to decrease. Conversely, when the ratio is less than 1, it indicates higher demand, leading to an increase in price as mentioned as flowchart in Fig. 3.

The SOC’s exponential impact

As the State of Charge (SOC) increases, the system relies more on stored energy, which impacts the price. The price tends to rise during charging and decrease during discharging.

Scenario 1: α = 0.1, β = 0.1

When α and β have been set to their minimum feasible values of 0.1, it shows that there is a huge impact on the price calculation, potentially leading to higher electricity prices as displayed in Fig. 6. This increase can directly influence consumer behaviour by discouraging energy usage, thereby limiting daily consumption across society. As a result, prices may fluctuate more quickly in reaction to changes in ratio and SOC, making the pricing system less stable. Such instability could pose challenges for both suppliers who may struggle to forecast demand and consumers who may face unpredictable energy costs, ultimately reducing the overall efficiency and reliability of the energy management system.

Fig. 6.

Scenario 1 of price.

Scenario 2: α = 0.3, β = 0.3

Here, the pricing model is more balanced and marginally stable compared to its predecessor, as both α and β are set to 0.3. This configuration results in a moderate and reasonable electricity price for consumers, offering a more sustainable and fair pricing strategy for both customers and suppliers as presented in Fig. 7. It promotes price stability while still allowing for dynamic adjustments based on supply–demand conditions and SOC variations. Customers are more likely to benefit from this approach, as it reduces the risk of sharp price spikes and encourages consistent usage patterns. Moreover, this pricing model fosters a more robust and responsive behaviour, helping to improve overall energy efficiency and grid reliability in the long term.

Fig. 7.

Scenario 2 of price.

Scenario 3: α = 0.5, β = 0.5

In this case, α and β values are 0.5. Higher value (such as 0.5) controls the sensitivity of the pricing model, resulting in only a minimal effect on price changes. This means that energy consumption will remain relatively stable, and price adjustments will be less responsive to sudden fluctuations in the supply–demand ratio and State of Charge (SOC) as indicated in Fig. 8. While this approach benefits customers by providing price stability and predictability, it can also be advantageous for suppliers if the prices remain reasonable, as it ensures a steady and manageable demand pattern without drastic shifts in load or pricing volatility.

Fig. 8.

Scenario 3 of price.

Considering all the above scenarios, as demonstrated in Fig. 17, the selected ideal parameter values of α = 0.3 and β = 0.3 provide a balanced and reasonable pricing structure that benefits both suppliers and customers. This configuration maintains prices at a sustainable level neither too high nor too low while effectively capturing the dynamics of SOC and demand–supply fluctuations. As a result, it promotes efficient energy utilization, ensures supplier profitability, and supports overall system stability.

Fig. 17.

Response of price based on ratio in all three cases.

Simulation and dataset

The proposed micro grid system was developed on the MATLAB simulation platform to model the dynamic interaction between solar, battery, and grid supply. A representative dataset of 50 rows was constructed for a 5 kW system, using three key parameters Ratio, SOC, and Price as mentioned in Table 5, which were identified as the most influential features during the pre-processing stage for price prediction using Machine Learning. The dataset was partitioned into training (70%), validation (15%), and testing (15%) sets to ensure reliable model evaluation. An exponential dynamic pricing model was integrated into the simulation as a tuning mechanism, enabling real-time flexibility under varying operational scenarios. The complete model as revealed in Fig. 9 was then simulated over a 24-h period to generate the required data. Upon completion, it is essential to export the simulation results to an Excel sheet for further analysis. To facilitate this, a structured table is created within the MATLAB workspace to enable seamless data visualization as evidenced in Fig. 10. This dataset is then used for further prediction and analysis using machine learning algorithms such as Support Vector Regression (SVR), Random Forest (RF), and Artificial Neural Networks (ANN), with a focus on evaluating prediction accuracy and error percentage.

Table 5.

Values of the features after the simulation by MATLAB.

Ratio	SOC	Price	Status
S < D (high price)
0.5	40	10	High
0.55	43.4	9.72	High
0.83	65.31	8.33	High
0.61	80.78	8.41	High
0.65	69.15	8.7	High
0.78	77.69	8.12	High
S > D (low price)
1.29	45.91	7.66	Low
1.4	52.93	7.23	Low
1.35	100	6.33	Low
1.73	74.65	6.08	Low
2.11	84.62	5.24	Low
2.2	100	4.8	Low

Fig. 9.

Simulation of price using ratio and SOC.

Fig. 10.

Dataset obtained in workspace from MATLAB simulation.

Simulation results and performance evaluations

MATLAB-based simulations are employed to analyse the outcomes of dynamic pricing. The model was implemented with a 5 kW power supply over a 24-h runtime as shown in Fig. 11. The resulting prices were determined based on varying values of the Ratio (ranging from 0.5 to 2.2) and the State of Charge (SOC), which ranged from 40 to 100% as indicated in Fig. 14 and 16. A sample dataset of 50 values was collected and used for prediction through machine learning algorithms. The objective is to achieve the lowest possible error percentage, enabling more accurate forecasting of price fluctuations based on real-time energy dynamics.

Fig. 11.

Proposed simulation model for dynamic pricing.

Fig. 14.

Variations of ratio and SOC.

Fig. 16.

Response of price based on Ratio and SOC.

The total net energy of the production is given by,

Here, m is the total number of energy sources, and is the output power of each source. Two critical elements are required to guarantee steady and dependable energy production: One benefit of the micro grid is that energy providers are incentivized to supply power to it when the market price is higher than their production costs. Second, there are limitations on the amount of power that can be generated by each energy source due to their maximum installed capacity.

Based on the above simulation results, the system is considered inactive when it does not generate any output power and the total available energy within the system is zero indicating that neither the renewable sources nor the storage components are supplying power to the load. In such a scenario, the system enters a low-power or dormant state, ensuring that unnecessary operations are avoided to conserve computational and control resources. However, when the average power output of the system lies within the predefined operational thresholds i.e., between the minimum and maximum allowable power limits the system actively monitors the power generation and consumption in real time as displayed in Fig. 12. If this condition is met, the system dynamically assigns the available energy value to a control variable that governs further decision-making processes, such as price adjustment, load management, or energy dispatch. This mechanism ensures that energy availability is accurately tracked and utilized, promoting efficient energy balancing while avoiding overestimation or wastage of resources.

Fig. 12.

Generation of electricity.

The supply-to-demand ratio is given by:

The ratio, ranging from 0.5 to 2.2, plays a crucial role in managing power distribution and potential outages. When supply exceeds demand, prices decrease; conversely, when supply is lower than demand, prices increase which is demonstrated in Fig. 15. The following section provides a detailed breakdown of the energy consumption pricing structure.

Fig. 15.

Response of price based on variations of demand and supply.

The proposed dynamic pricing model introduces a real-time approach to adjusting energy prices based on supply and demand fluctuations indicated in Fig. 13. When supply exceeds demand, lower prices signal an energy surplus, encouraging increased consumption, reducing waste, and optimizing grid utilization. This strategy ensures a smoother, more predictable energy system, benefiting both consumers and producers. Additionally, during periods of low supply, the pricing mechanism helps stabilize the market by promoting responsible consumption, ultimately fostering a more resilient and efficient energy ecosystem (Figs. 14, 15, 16, 17).

Fig. 13.

Variations of demand and supply.

Machine learning integration

Machine learning (ML) has been considered as a powerful approach for reducing the complexities of energy management, especially in implementing renewable energy sources for dynamic pricing. A detailed examination of machine learning applications in energy systems demonstrates its adaptability by looking deeper into its use in enhancing grid operations, demand-side management, and renewable energy sources. Machine learning algorithms can use real-time data to predict demand spikes, energy storage demands, and renewable energy output. This meshes flawlessly with the objectives of the research, which involve pricing systems that deals with unpredictable energy generation and consumption⁴⁶. By utilizing these predictive abilities, micro grids can enhance their load management capabilities and enable seamless real-time modifications to dynamic pricing based on energy consumption trends⁴⁷. To ensure efficient energy generation, micro grids aims to achieve energy balance in circumstances with various power sources rely on these models. As renewable energy sources have become more prevalent in the energy landscape, ML has become increasingly important in variable energy systems. New breakthroughs in machine learning demonstrate how state-of-the-art algorithms are revolutionizing variable energy management by enabling energy systems to make judgments in real-time. Since renewable energy sources like solar are infamously unreliable, these advancements are particularly significant for dynamic pricing since they require solutions that can adapt and anticipate. By incorporating machine learning (ML)-based forecasting models into dynamic pricing systems as presented in Fig. 18, grid stability and consumer satisfaction can be enhanced through precise price adjustments in response to fluctuations in renewable energy⁴⁸. Unlike conventional studies that applied machine learning frameworks primarily for addressing managerial challenges within the grid, many of them failed to incorporate adaptive pricing mechanisms that respond to fluctuating supply and demand conditions.

Fig. 18.

Machine learning process.

This oversight limited the effectiveness of their models in integrating renewable energy resources within energy systems, especially in real-time operations. For instance, some studies have focused on short-term renewable forecasting using CNN and LSTM models to improve generation prediction accuracy⁵⁰, while others compared deep learning models for electricity price prediction without dynamically linking price with battery SOC or demand–supply ratios⁵¹. Similarly, broader reviews of AI in microgrid optimization emphasized system scheduling and control but did not provide direct strategies for price forecasting tied to system states like SOC^49,52. To overcome these limitations, the proposed model in this study introduces a dynamic pricing strategy powered by artificial intelligence specifically, an ANN-driven approach integrating battery SOC and supply–demand ratio as core factors. This not only enhances the adaptability of price prediction but also contributes significantly to real-time energy optimization in microgrid operations. By leveraging the algorithms, an AI-based predictive framework trained on a structured dataset generated through the simulation of MATLAB platform. It introduces a paradigm shift by embedding dynamic pricing mechanisms into real-time decision-making, ensuring greater adaptability and efficiency in energy distribution. To ensure robust validation, the sample dataset is strategically separated as 70% and 30% for training as well as testing. Through iterative fine-tuning and training, the model is validated the prices using percentage error analysis to assess its accuracy. To further establish the superiority of our approach, a comparative evaluation is conducted using algorithms as outlined in Table 6, which include Support Vector Regression (SVR), Random Forest (RF) and Artificial Neural Network (ANN). Because of their application in regression tasks, it was assumed that these models would make good comparisons for predicting SOC-based dynamic power prices. Unlike traditional methods that focus primarily on demand prediction or large-scale grid management, this proposed model extends the static price forecasting by dynamically adjusting pricing in real time based on fluctuations in parameters. This adaptive pricing mechanism optimizes energy distribution within the Micro grid, ensuring both economic efficiency and operational resilience. By utilizing the AI driven algorithms, our framework not only enhances the cost-effectiveness of RES integration but also represents a transformative leap in modern energy management systems particularly while pricing is considered.

Table 6.

Performance analysis of ANN compared with different algorithms.

Supply to demand ratio	SOC	Price	Random forest		SVR		ANN
			Predicted price	% Error	Predicted price	% error	Predicted price	% Error
1.51	60.18	6.82	6.947	1.873	6.870	0.741	6.826	0.089
0.73	40	9.3	8.642	7.073	9.337	0.407	9.301	0.011
0.71	85.48	8.11	7.755	4.376	8.203	1.158	8.110	0.007
0.6	84.66	8.41	8.035	4.448	8.392	0.214	8.418	0.098
1.32	61.03	7.24	7.172	0.938	7.230	0.133	7.235	0.069
1.4	52.93	7.23	7.018	2.918	7.268	0.534	7.230	0.011
0.76	89.43	7.89	7.719	2.166	8.002	1.432	7.885	0.061
0.56	88.59	8.42	7.999	4.994	8.384	0.418	8.420	0.001

The results clearly demonstrate that our proposed ANN-based approach achieves a percentage error consistently below 20%, with an exceptionally low RMSE of 0.010. In contrast, both LSTM and CNN-LSTM exhibit significantly higher errors, exceeding 50%, with average RMSE values of 2.03 and 1.62, respectively, under random SOC and ratio conditions as compared in Fig. 20. These findings confirm that ANN offers superior accuracy and reliability for the proposed pricing method.

Fig. 20.

Performance Analysis of ANN compared with LSTM and CNN-LSTM algorithms.

Conclusion

This research presents a novel AI-driven dynamic pricing strategy tailored for modern micro grid systems, incorporating two critical real-time operational parameters: the supply-to-demand (S/D) ratio and the Battery State of Charge (SOC). Unlike conventional static pricing schemes, the proposed model employs an exponential pricing function that continuously adjusts electricity prices in real time based on fluctuations in both energy availability and storage levels. This ensures that the pricing structure remains responsive, adaptive, and equitable under varying operating conditions. Extensive simulation results validate the effectiveness of the proposed model and it determines the price at intervals of 0.1 s during the simulation, achieving a minimum price of 4.8 INR/kWh when the S/D ratio is 2.2 and SOC reaches 100%, representing a surplus energy scenario. Conversely, the maximum price of 10 INR/kWh is observed when the S/D ratio falls to 0.5 and SOC declines to 40%, indicating high demand with limited storage. These outcomes demonstrate the model’s ability to self-regulate pricing while adhering to defined system constraints, thus promoting cost fairness for consumers and profit stability for suppliers. To enable predictive intelligence, the pricing model is integrated with a machine learning layer using Artificial Neural Networks (ANN). This ANN-based model successfully forecasts prices with a percentage error of less than 10%, within a fraction of 30 s. Such high-frequency accuracy makes it well-suited for real-time applications in micro grids and energy trading systems. Furthermore, a comparative evaluation with other machine learning techniques, namely Support Vector Regression (SVR) and Random Forest (RF), highlights the superior performance of ANN. Across various simulation scenarios as displayed in Fig. 19, ANN consistently delivered higher forecasting precision, showcasing its robustness in modelling nonlinear and complex relationships inherent in energy systems. As evidenced in the comparative analysis table, traditional approaches fail to incorporate dynamic elements like SOC and supply–demand fluctuations, limiting their applicability in modern energy systems and also previous studies have made progress in renewable forecasting, demand response modelling, or static optimization, they often neglect the integration of real-time pricing mechanisms tied to dynamic parameters like SOC or supply–demand variability. Most rely on predefined or elastic price models without predictive adaptability, limiting their effectiveness in fast-changing micro grid environments. In conclusion, the proposed ANN-enhanced dynamic pricing framework offers a significant advancement in intelligent micro grid management. It not only delivers accurate and responsive price forecasting but also contributes to the economic sustainability and operational resilience of decentralized energy systems. By aligning energy cost with real-time system behaviour, this approach fosters energy efficiency, grid stability, and consumer trust, marking a pivotal step toward future-ready smart micro grids (Fig. 20).

Fig. 19.

Comparison of Predicted price with machine learning algorithms for all cases.

Author contributions

I.Balakumar: Conceptualization, Design and Development, Methodology, Software, Data curation, Examined, Writing original Draft preparation, Visualization, Investigation. C. Vaithilingam: Supervision, Software, Validation, Resources, Writing-Reviewing and Editing, Data curation. All authors read and approved the final Manuscript.

Funding

Open access funding provided by Vellore Institute of Technology.

Data availability

The datasets used during the current study available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.