Virtual power plant - another type of intelligence at the edge of the new power system

I personally believe that at the edge of the new power system, there are opportunities for innovation, even disruptive innovation in the field of intelligence, and virtual power plants are one of the trends.

GPT's definition of disruptive innovation is:

Disruptive Innovation is a concept proposed by Harvard Business School professor Clayton Christensen to describe an innovative approach that can change the market landscape and disrupt existing industries. Unlike traditional incremental innovation, disruptive innovation typically enters the market by providing simpler, cheaper, or more convenient products or services, thereby attracting a previously unmet customer base and ultimately triggering a transformation in the entire industry.

The edge of the new power system

The edge of the new power system lies in the public distribution system of medium and low voltage, as well as within the user distribution system.

1. Traditionally, public distribution systems are constructed and operated by DSO (a wholly-owned power supply company of the power grid), a distribution company holding a special operation license (electricity business license). Based on power supply business services, the business technology trend is towards the integration of operation, distribution, and commissioning.

2. The user power distribution system did not previously become a "system". Whether it was popular concepts such as "plant power", "civil building power supply and distribution", "building electrical", etc., they were mostly viewed from the perspective of equipment and design, because as the end of the power distribution system, the user power distribution system is only based on a simple topology structure (non mesh), one-way, and almost unmanaged final distribution of electrical energy.

Evolutionary Trends at the Edge of Power Systems

Due to new social trends and changes in business models, especially the rapid penetration of distributed new energy, there are two impacts:

1. Impact at the market and business levels:

(1) The boundary between the property rights of public distribution and private distribution for users is gradually becoming blurred. For example, user specific transformer and distribution rooms and cable facilities are invested, constructed, and maintained by public distribution companies, and leased to users for use; For example, in a large number of urban villages, commercial complexes, and industrial real estate projects, the internal public power distribution facilities form a "power conversion"; There are also new energy investors who are renovating distribution facilities for homeowners and adding photovoltaic storage grid connected cabinets. The internal definition of property rights has become more ambiguous and diversified.

(2) New market entities have emerged in user specific power distribution systems. In the past, the main body of the power system was relatively clear, with four identities: generation, transmission, distribution, and use, and power users were only the main users of electricity. However, at the edge of the new power system, distributed generation companies, equipment maintenance companies, lower level power users (such as tenants), distributed energy storage companies, microgrid operators, charging station operators, power sales companies, and virtual power plant operators have emerged. They have formed a large number of "many to many" transactions on the edge, and even many commercial disputes have arisen.

2. Technical impact

(1) The emergence of silicon-based power systems. Whether it is solid-state switches, power electronic transformers/converters, distributed photovoltaic inverters, energy storage PCS, charging piles, frequency converters, DC air conditioners, LED lighting, electric boilers, electric kilns, induction cookers, most new power equipment are power components of power electronics, and the embryonic form of "silicon-based power systems" has emerged at the edge.

(2) The complexity on the edge side increases by at least one order of magnitude

The traditional low-voltage distribution system is characterized by unidirectional power flow, simple topology, minimal sensing and control, and almost no need for management. For example, low-voltage switches are designed to be almost maintenance free.

However, at the edge of the new power system, there are bidirectional variable power flows, complex network topologies, and a large number of intelligent devices that require "four capabilities" to be connected. Both the number of intelligent nodes and the multi-layered interactive relationship between distribution network microgrid group microgrid are 1-2 orders of magnitude more complex than traditional power distribution systems.

The edge evolution of the new power system will bring a new direction of intelligence.

Edge intelligence, another type of intelligence

In October 2024, a paper titled "Intelligence at the Edge of Chaos" was published on arXiv by Shiyang Zhang et al. from Yale University,

There is an interesting viewpoint inside: traditionally, it is believed that AI intelligence must be trained on a large amount of intelligent data, even artificially synthesized data, in order to emerge. But we have found another possibility:

Intelligence can be generated through modeling simple systems, as long as these systems exhibit complex behavior, even if the process of generating this data itself does not possess inherent intelligence.

They trained AI to recognize and predict based on data from cellular automata with different rules, and found that the chaotic rule-based cellular automata dataset can train prediction algorithms with high AI intelligence.

The dataset they chose is data from cellular automata, which iteratively run based on very simple rules to produce different data results.

Selecting chaotic system data in an edge state for training can result in intelligent AI; If it is regular data, dead data, or completely chaotic data, it is impossible to train AI with sufficient intelligence.

The conclusion of this paper is:

In a sense, intelligence can be generated in a sophisticated "dessert" dimension based on a data set of complex systems with simple rules, in a dissipative structure (chaos) far from equilibrium.

New power system, edge intelligence brought by physical edge

The edge of the new power system may be in the early stage of the emergence of intelligence:

1. A large number of grid connected photovoltaic storage and source grid load storage microgrid systems have emerged at the edge of the power grid, gradually surpassing the critical quantity. Just like cellular automata, they must be of sufficient length and undergo enough iterations to exhibit a chaotic dataset.

2. These edge nodes have relatively simple rules. Microgrids or photovoltaic storage systems in networked form, operating according to PQ nodes, only need to focus on power balance.

3. Simple interactions are formed between nodes, such as wall mounted electricity sales and microgrid group coordination, with certain and simple perception, communication, and control collaboration functions.

4. The marketization of electricity and technological advancements have driven the advancement and iteration of intelligence and collaboration.

The physical edge of the new power system, combined with the "four capabilities" and marketization, will form new edge intelligence and collaborate with high-voltage distribution and transmission networks.

In my opinion, there is not only one path for power system AIGC, that is, follow the semantic model in the Internet field, stack computing power and form a centralized AI algorithm. Because different levels of power systems face different physical limitations, information structures, intelligent rules, and application scenarios.

At the level of transmission network, the rules are complex, the data is huge, and the operation is centralized, so it is suitable for centralized intelligent large models such as LLM to do centralized prediction and AI scheduling.

In the edge system of the power grid, namely the aggregation of public distribution networks, user distribution networks, microgrids, and even microgrids, the rules of individual nodes are simple, data is limited, and the number of nodes is numerous. After the interaction between Hai Lian nodes, new edge intelligent states will emerge. The above technological and commercial trends are also driving the evolution and iteration of this edge intelligence to some extent:

This is another path of intelligent emergence, and also a more disruptive innovation path.

Virtual power plant, an emerging form of edge intelligence

Virtual power plants are essentially the edge of a new type of power system, an abstract and upper level application of intelligence that emerges based on massive nodes and simple rules. With enough nodes, a power plant can be virtually created.

Just like how there are enough small fish or birds to simulate a large organism and intimidate natural enemies, but each fish/bird has only about 2-3 rules of operation.

Here is a photo of a flock of starlings flying:

Disruptive innovation mostly begins to emerge on the edge.

Another intelligent future brought by distributed energy is worth exploring and looking forward to.