China is developing a new operating model for its electrical grid that relies on a layered control system built around virtual power plants, distributed energy resources, and a blockchain network that dictates how thousands of small units negotiate their role in real time. The study behind this model describes a system that behaves almost like a coordinated organism. It receives targets, interprets them, and forces each connected resource to respond in a way that maximizes stability, economic efficiency, and environmental performance. The research explains how this architecture works, why it is being built, and what it changes inside a modern power network.
The challenge begins with the amount of distributed energy now present in China’s electricity supply. Wind farms, rooftop solar units, small thermal plants, and large energy storage facilities are increasing in number across many provinces. These resources expand the total capacity of the system, but they also increase volatility. Traditional grid scheduling was built around predictable output from large plants. The new pattern contains thousands of independent units that fluctuate in output or follow local incentives. This creates coordination problems that cannot be solved through a single market signal or a single scheduling command. The authors of the study argue that the market environment remains incomplete because China’s carbon trading market, green certificate system, and electricity market each function separately. There is no unified mechanism for assigning priority to renewables, pricing their benefits, or coordinating them across competing actors. Without a structure for allocating value across all these dimensions, distributed resources fail to synchronize. Virtual power plants have been introduced as a partial solution because they pool resources into aggregated entities that can respond to grid needs as a single unit. Yet virtual power plants introduce another problem. They must decide which resources to include, which to exclude, and how to settle rewards among participants. None of that can happen without trust, transparency, and verifiable decision rules.
To solve this, the researchers build a three tier system. The top tier is the grid dispatch center. The middle tier is the virtual power plants that receive instructions from the grid and pass them to distributed energy resources. The bottom tier is the full set of resources that can participate in aggregation. This environment is supported by a blockchain network that includes nodes for the grid, the virtual power plants, and the distributed resources. The blockchain is not used for currency. It is used as the control layer that establishes identity, guarantees that every participant sees the same scheduling information, and prevents any entity from accessing sensitive data that could distort the market. The chain uses the PBFT consensus algorithm, which limits malicious interference and maintains high throughput. The authors tested the design in a simulated environment based on the Ethereum network. The throughput exceeded eight hundred transactions per second and handled repeated exchanges of aggregation schemes, scheduling demands, and registration information without notable delays.
The grid uses the blockchain to publish a set of multidimensional value weights. These are numerical priorities assigned to three factors. The first is economic value. The second is risk, which measures exposure to imbalance penalties that arise when aggregated resources fail to meet scheduled output. The third is environmental value, which reflects green certificate income and carbon penalties. These weights are not fixed. The grid adjusts them according to the goals of the system. If the grid needs to increase clean energy penetration, environmental value receives a higher coefficient. If volatility in renewable output becomes difficult to manage, risk receives a higher coefficient. If supply and demand are tight, economic value receives a larger coefficient. Consistent value weights create a common understanding of what the grid wants.
Virtual power plants receive these weights and construct their aggregation strategies through an optimization model. That model calculates economic costs, risk exposure through conditional value at risk, and environmental income. It balances these three values to produce the highest total return under the given weights. Once the model decides which resources maximize the total return, the virtual power plant publishes an encrypted aggregation scheme to the grid. Encryption prevents competitors from seeing each other’s plans. Only the dispatch center can decrypt and verify them. The grid then desensitizes the scheme before sending it to distributed energy resources so they cannot identify individual competitors. This is designed to eliminate collusion between small resources that could distort prices or manipulate the aggregation process.
Distributed resources do not simply accept or reject offers. They must evaluate their expected revenue using their own benefit functions. Each resource calculates revenue from electricity sales, minus operating costs and penalties for deviations. If the resource is wind, it evaluates maintenance costs, forecast errors, and penalties for abandoning output. If it is solar, it accounts for tariff compensation and penalties for abandoned light. If it is energy storage, it evaluates charging and discharging revenue minus battery degradation and operation costs. If it is demand response, it evaluates payments for load reduction, minus the cost of shifting consumption. If it is a small thermal plant, it evaluates fuel costs, carbon penalties, and operation costs. Every type of unit has a detailed mathematical representation of profit and constraints. These constraints vary. Storage must satisfy a state of charge balance across each day. Thermal plants must abide by ramping limits. Wind and solar must remain within forecast ranges. Demand response must balance total curtailment and total rebound load.
The key mechanism connecting individual resources to the system is the comprehensive benefit function. This function includes the revenue of each resource and a portion of the virtual power plant’s integrated value. The share of integrated value each resource receives is determined by its contribution to the virtual power plant’s regulation service. Resources that contribute higher regulation value gain more from the shared portion. Resources that contribute less gain less. As a result, each unit’s decision aligns with both its internal profit and its influence on the overall aggregated value.
Once all individual benefit functions are defined, the system enters an evolutionary process. Distributed resources do not have perfect information and do not operate as perfect optimizers. Instead they adjust their aggregation choice over time based on observed payoffs. Evolutionary game theory is used to represent this adjustment. In each iteration, distributed resources increase the probability of selecting the virtual power plant that offers higher expected returns. Over repeated iterations the system converges toward a stable equilibrium. The equilibrium represents a configuration where no resource can improve its return by switching. This creates a natural alignment between individual incentives and the global optimization target set by the grid.
The researchers simulated a region containing two virtual power plants, two thermal units, two wind farms, and an energy storage plant. They assigned each plant realistic capacities and risk preferences. They generated wind and solar output scenarios using meteorological data, Latin Hypercube Sampling, and clustering algorithms. These scenarios capture the uncertainty of renewable output and provide the basis for calculating risk. Under initial weights of 0.4 for economic, 0.2 for risk, and 0.4 for environmental value, the equilibrium allocation placed wind farm one, thermal unit two, and the energy storage unit into virtual power plant one. Thermal unit one and wind farm two were placed into virtual power plant two. The pattern reflects strategic complementarity. Virtual power plant one faced wider fluctuations in demand and benefited more from energy storage’s ability to absorb surplus power and release it during peaks. Virtual power plant two, which adopted a risk avoidance posture, favored units with stable generation. At the end of the simulation both virtual power plants achieved complete absorption of wind power, something that is difficult to achieve under typical scheduling practices.
The researchers then varied the multidimensional value weights to measure how the system adjusts. When the economic weight increased and the environmental weight decreased, returns for all distributed resources fell. The main reason is that environmental value carries strong incentives for clean energy and significantly offsets imbalance penalties. When that value is reduced, the total income of wind and solar units declines and the system yields lower total benefit. Energy storage, however, exhibited stable returns regardless of weight shifts. It profits from price differentials, not green income. Virtual power plants can shape these differentials inside their aggregated tariff structures. If a virtual power plant wants more storage participation, it can increase the gap between peak and valley prices inside its scheme.
The study also analyzed the gain and loss of the integrated value of each virtual power plant under different weight combinations. A clear pattern appeared. If environmental weight is increased beyond reasonable limits, the system gains environmental income but begins to struggle with stability because flexible resources cannot absorb all renewable surges. If economic weight is increased too far, the system reduces renewable absorption and drives up imbalance penalties. The balanced point in their test scenario occurred at the initial mix of 0.4 economic, 0.2 risk, and 0.4 environmental value. At that point the magnitude of gains and losses on each dimension was minimized and the system produced the highest coordinated return.
They also tested the sensitivity of the evolutionary process. They constructed a scenario containing three distributed resources competing between two virtual power plants. They ran the evolutionary model with fixed probability for one resource and observed the trajectories of the others. The system converged to the same equilibrium regardless of the fixed probability. They then randomized all initial probabilities and ran another series of tests. Again the system converged to the same equilibrium. This demonstrates that the model is robust across varied starting positions. The distributed resources always reached the configuration that maximized the global benefit function.
Finally the authors tested supply and demand imbalances. In a scenario where available distributed resources exceeded demand, the system successfully aggregated the required units and assigned roles to each resource through the evolutionary process. When demand exceeded supply, the system evaluated each resource according to its contribution to global benefit. Resources with low contribution were removed from the final aggregation state. This produced a higher total system benefit than a configuration that attempted to include every available unit. The model therefore acts as a filter that removes resources that degrade stability or produce negative integrated value.
The system described in the study creates a new method for controlling a modern grid. It blends decentralized negotiation and centralized planning. It allows the grid to adjust long term development goals by altering the value weights. It forces virtual power plants to internalize environmental and risk dimensions that were previously ignored. It allows individual resources to behave according to economic self interest while still contributing to a coordinated regional objective. The blockchain network guarantees that all participants see the same data and prevents manipulation or premature disclosure of aggregation schemes. The PBFT consensus mechanism ensures that malicious nodes cannot disrupt the process. The evolutionary game mechanism ensures convergence even when participants lack full information.
The research demonstrates that this integrated framework can manage distributed resources in a way that improves renewable absorption, improves stability, and supports coordinated scheduling across all actors. As distributed energy continues to expand, traditional scheduling methods will become increasingly difficult to maintain. The model presented here provides a pathway for grids that need to integrate complex mixes of intermittent generation, flexible loads, and storage units without losing control of system stability. It offers an architecture in which value is not simply measured by price, but by a balanced evaluation of economic returns, operational risk, and environmental contribution. The result is a system that can adapt to changing conditions, interpret long term planning objectives, and allocate resources with more precision than any traditional market platform can achieve.
Source
Dong, F., Wang, X., Yang, L., and Shi, M. Multi dimensional value diversion and aggregation of DER based on alliance chain under the grid VPP DER three level synergy framework optimization. Journal of Renewable and Sustainable Energy, 18, 015306 (2026).
https://doi.org/10.1063/5.0297338
