Power quality optimization control method and system based on distributed collaborative decision

Abstract

The invention relates to the technical field of power system automation and electric energy quality control, in particular to an electric energy quality optimization control method and system based on distributed collaborative decision. The method comprises the following steps: deploying synchronous phasor measurement units and corresponding edge intelligent decision nodes at a plurality of nodes of a power grid, and constructing a distributed network; each node calculates an electric energy quality characteristic index based on local data; dynamic role negotiation is carried out through local communication, and a leading controller is determined; the dominant controller fuses strategies and confidence information of local and adjacent nodes to generate a cooperative control strategy; and finally, equipment such as a static var generator and an active power filter is controlled to perform coordinated execution. Through a decentralized collaborative decision-making mechanism, the defects that centralized control is slow in response and low in reliability are overcome, rapid self-adaptive treatment on voltage sag, harmonic waves and other problems is achieved, and the response speed, the treatment success rate and the system robustness are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of power system automation and power quality control technology, specifically to a power quality optimization control method and system based on distributed collaborative decision-making. Background Technology

[0002] With the large-scale grid connection of new energy power generation and the widespread integration of various power electronic devices, power quality problems in modern power grids (such as voltage sags, harmonic distortion, and frequency fluctuations) are becoming increasingly prominent and their dynamic characteristics are complex. Traditional centralized power quality optimization control systems typically rely on a central master station to collect global data, perform unified calculations, and issue control commands. This model has significant technical bottlenecks: First, the long links of data uploading, centralized processing, and command issuance lead to significant decision-making delays, making it difficult to respond promptly to rapid transient events at the millisecond level; second, the massive amount of synchronous phasor measurement data converges at the central node, placing enormous pressure on communication bandwidth and the computing power of the central server.

[0003] Synchronous phasor measurement units (PMUs) can provide high-precision, time-synchronized power grid operation data, providing a data foundation for refined power quality management. However, most existing PMU-based applications still follow a centralized processing paradigm, failing to fully utilize their distributed measurement characteristics and making it difficult to adapt to the wide-area, decentralized, and fast-response control requirements of modern power grids.

[0004] Therefore, there is an urgent need for a new power quality optimization control scheme that has a faster response speed, higher reliability, more efficient resource utilization, and can adapt to changes in the power grid's operating state. Summary of the Invention

[0005] To address the problems of decision delay, communication and computing bottlenecks, low reliability, and lack of adaptive capabilities inherent in centralized control as mentioned in the background, this invention proposes a power quality optimization control method and system based on distributed collaborative decision-making. The aim is to achieve rapid localized perception, collaborative decision-making, and precise management of power quality issues by constructing a decentralized edge intelligent decision-making network, thereby improving the overall resilience and recovery speed of the power grid against power quality disturbances.

[0006] The present invention adopts the following technical solution: In a first aspect, this application provides a power quality optimization control method based on distributed collaborative decision-making, comprising the following steps: S1: At multiple nodes of the power grid, synchronous phasor measurement units and corresponding edge intelligent decision-making nodes are deployed to form a distributed decision-making network. S2: Each edge intelligent decision node calculates power quality characteristic indicators based on local data collected by its associated synchronous phasor measurement unit. S3: Each edge intelligent decision node, based on the power quality characteristic index, conducts dynamic role negotiation with neighboring edge intelligent decision nodes through local communication to determine the dominant controller from the negotiating parties; S4: The dominant controller generates a collaborative control strategy based on the locally generated control strategy and by integrating the control strategies and corresponding confidence information of at least one adjacent edge intelligent decision-making node. S5: According to the aforementioned coordinated control strategy, control the power quality management equipment connected to the relevant power grid nodes to perform coordinated actions.

[0007] Optionally, step S0 is included before step S1: S0: Based on the improved multi-objective dung beetle optimization algorithm, the deployment location of the synchronous phasor measurement unit is optimized to determine the power grid node where the synchronous phasor measurement unit is deployed; wherein, the optimization configuration takes the overall network observation accuracy, equipment investment cost and communication reliability of the power grid node as optimization objectives.

[0008] Optionally, the objective function of the improved multi-objective dung beetle optimization algorithm is:

[0009] In the formula, x Represents a vector of decision variables; N Indicates the total number of nodes in the power grid; M Indicates the total number of lines; x i Indicates the first i The decision variables for whether to deploy a synchronous phasor measurement unit at a power grid node; c i Indicates the first i The cost of deploying a synchronous phasor measurement unit at a power grid node; w ij Indicates the first i The first power grid node j The weighting coefficient of each line; o ij Indicates the first i The first power grid node is related to the first j The observation capability coefficient of each line; r i : No. i Communication reliability indicators for each power grid node.

[0010] Optionally, in step S2, the power quality characteristic indicators include at least one of voltage deviation rate, total harmonic distortion rate, and frequency fluctuation rate.

[0011] Optionally, the dynamic role negotiation in step S3 specifically includes: S31: Each edge intelligent decision node calculates the power quality severity index of its respective power grid node; S32: Edge intelligent decision-making nodes exchange the power quality severity index and governance capability information; S33: Based on the exchanged information, the contribution of each edge intelligent decision-making node to the collaborative governance alliance is calculated by improving the Shapley value algorithm; S34: Based on the contribution level, dynamically allocate the roles of the dominant controller and the auxiliary actuator in each edge intelligent decision-making node.

[0012] Optionally, the formula for calculating the power quality severity index is:

[0013] In the formula, S i Represents a power grid node i The power quality severity index; Δ U i Represents a power grid node i Voltage deviation; U n Indicates the rated voltage; T i Represents a power grid node i Total harmonic distortion; Δ f i Represents a power grid node i Frequency deviation; f n Indicates the rated frequency; a、b、c Represents the weighting coefficient (α) + b + c = 1).

[0014] Optionally, the final fusion control strategy is generated using the following formula:

[0015] In the formula, i * Indicates the final fusion control strategy; i i Indicates the first i The control strategy proposed by each node; C i For the first i Confidence of the strategy for each node; K The number of nodes participating in the fusion.

[0016] Optionally, the local control strategy of each edge intelligent decision node is generated by a deep reinforcement learning model, wherein the confidence level... C iThe value function and state access frequency are determined based on the corresponding deep reinforcement learning model.

[0017] Optionally, in step S5, the power quality management equipment includes a static var generator and an active power filter; The coordinated control strategy includes: issuing a reactive power compensation reference value command to the static var generator; and issuing a harmonic current compensation reference value command to the active power filter.

[0018] Secondly, this application provides a power quality optimization control system based on distributed collaborative decision-making, comprising: A distributed measurement module is used to deploy synchronized phasor measurement units at multiple power grid nodes to collect data; Multiple edge intelligent decision-making modules are deployed at each power grid node in a one-to-one correspondence with the synchronous phasor measurement unit, and are used to calculate power quality characteristic indicators based on local data; The role negotiation module is used to enable each edge intelligent decision-making module to dynamically negotiate its role with adjacent edge intelligent decision-making modules through local communication based on the power quality characteristic indicators, so as to determine the dominant control module. The strategy fusion module is built into the edge intelligent decision module that is determined to be the dominant control module. It is used to generate a collaborative control strategy based on the locally generated control strategy and by fusing the control strategies and corresponding confidence information of at least one adjacent edge intelligent decision module. The control execution module is used to control the operation of the power quality management equipment according to the collaborative control strategy.

[0019] Compared with the prior art, the present invention has the following significant advantages: Fast response speed: By leveraging local computation at edge intelligent decision nodes and collaboration with neighboring nodes, the latency of remote data transmission and centralized processing is avoided, reducing typical control response time from 100 seconds in traditional centralized systems. ms Level shortened to 15 ms Within a certain range, it can effectively suppress rapid transient power quality events.

[0020] High reliability: It adopts a decentralized dynamic role allocation mechanism with no single control center. When any node fails, the system can automatically reconstruct the control architecture through negotiation to ensure the continuity of the overall governance function. The governance efficiency retention rate exceeds 90% under single point of failure.

[0021] High resource utilization: The raw measurement data is processed and features are extracted at the edge. Only a small amount of decision information (such as strategy and confidence) needs to be exchanged between adjacent nodes. Compared with the full data upload mode, the backbone communication bandwidth requirement can be reduced by more than 60%, and the computing load is distributed.

[0022] Strong adaptability: Based on the dynamic role allocation of the real-time power quality severity index and the improved Shapley value algorithm, as well as the confidence-based strategy fusion, the control system can adaptively adjust the control structure and decision weights according to the spatiotemporal distribution and severity of power grid disturbances, so as to achieve the best or second-best collaborative governance effect globally. The success rate of managing events such as voltage sags can be increased from 85% to over 96%. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of the power quality optimization control method based on distributed collaborative decision-making provided in the embodiments of the present invention.

[0025] Figure 2 This is a flowchart of the power quality optimization control method based on distributed collaborative decision-making provided in the embodiments of the present invention.

[0026] Figure 3 This is a schematic diagram of the dynamic role allocation mechanism provided in an embodiment of the present invention.

[0027] Figure 4 This is a schematic diagram of the confidence-weighted fusion algorithm provided in the embodiments of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0029] refer to Figure 1 The power quality optimization control method based on distributed collaborative decision-making provided in this application includes the following steps: S1: At multiple power grid nodes, synchronous phasor measurement units and corresponding edge intelligent decision-making nodes are deployed to form a distributed decision-making network. This step forms the physical foundation for system deployment and networking, aiming to establish a decentralized, locally intelligent collaborative control architecture.

[0030] The implementation method is as follows: Based on factors such as the power grid topology, the location of critical loads, and areas known to have frequent power quality problems, a group of representative power grid nodes is selected. These nodes are typically electrical connection points such as substation busbars, important feeder connection points, or distributed generation grid connection points.

[0031] At each selected power grid node, a synchronous phasor measurement unit and an edge intelligent decision node that is hardware connected or logically bound to it are deployed.

[0032] Synchronous phasor measurement unit: usually refers to synchronous phasor measurement device, which is responsible for collecting data such as voltage, current amplitude, phase, and frequency at the node in real time at a high sampling rate (such as 50 frames per second), and stamping them with a unified timestamp to ensure data synchronization across the entire network.

[0033] Edge intelligent decision-making node: This can be an embedded industrial control computer, an industrial server, or a smart gateway with powerful computing capabilities. It has a built-in processor, memory, and dedicated software, enabling data reception, computational analysis, and intelligent decision-making. It corresponds one-to-one with a synchronous phasor measurement unit, forming a basic "sensor-decision" unit pair.

[0034] All deployed "measurement unit-decision node" configurations are connected via a collaborative communication network (such as 5G-U, industrial fiber optic Ethernet, etc.). The key networking principle is that each edge intelligent decision node only needs to establish direct communication links with a few nodes that are electrically or topologically adjacent to it, rather than connecting to all nodes in the network. This forms a distributed decision-making network where control is dispersed among the nodes, relying on local communication for collaboration, thus eliminating a centralized central node.

[0035] S2: Each edge intelligent decision node calculates power quality characteristic indicators based on local data collected by its associated synchronous phasor measurement unit. The core of this step is to transform massive amounts of raw data into lightweight, information-rich features that can be used for intelligent decision-making, thereby greatly reducing the amount of data transmitted and the computational burden.

[0036] The specific implementation method is as follows: Data reception: Each edge intelligent decision node receives the raw synchronous phasor data stream uploaded by its associated synchronous phasor measurement unit in real time.

[0037] Feature Calculation: The algorithm software within the node rapidly processes the raw data to calculate key power quality characteristic indicators. These indicators are core parameters reflecting power quality issues and typically include: Voltage deviation rate: (Measured voltage - Rated voltage) / Rated voltage 100%, reflects the degree of deviation of voltage amplitude.

[0038] Total Harmonic Distortion (THD): The ratio of the effective values ​​of all harmonic components to the effective value of the fundamental component, which measures the severity of waveform distortion.

[0039] Frequency fluctuation rate: (measured frequency - rated frequency) / rated frequency × 100%, reflects the stability of the system frequency.

[0040] Localized processing: All calculations are performed locally on the edge intelligent decision-making node. The generated power quality characteristic index data volume is much smaller than the original phasor data, creating conditions for low-bandwidth, low-latency information exchange between subsequent nodes.

[0041] S3: Each edge intelligent decision node, based on the power quality characteristic index, conducts dynamic role negotiation with neighboring edge intelligent decision nodes through local communication to determine the dominant controller from the negotiating parties; This step is the "brain" of distributed collaboration. Based on real-time power grid conditions, it dynamically determines the control core of a local area through democratic consultation, thus achieving self-adaptation of the control structure.

[0042] The specific implementation method is as follows: Information broadcasting: Each edge intelligent decision node periodically broadcasts the power quality characteristic index (or the "severity index" obtained by further integration, see background technology) calculated in step S2, as well as its available control resource information (such as the remaining capacity of the connected SVG) to its neighboring nodes through local communication.

[0043] Negotiation-based decision-making: After receiving information from neighboring nodes, each node runs a pre-defined dynamic role allocation algorithm within a small local area (e.g., all nodes that communicate directly with each other form a "negotiation group").

[0044] Determining the dominant controller: The algorithm assesses each node's potential contribution to addressing the current local power quality event based on the severity of the problems reported by each node and their governance capabilities. The node with the highest contribution is elected as the dominant controller for that negotiation cycle. The remaining nodes act as auxiliary executors. This process is "dynamic," meaning that the dominant controller's role may shift among nodes as the grid state changes.

[0045] This ensures that in the event of events such as voltage dips, the node closest to the source of the problem or with the greatest potential for resolution automatically assumes the responsibility of organization and coordination, thus achieving the optimal dynamic allocation of control resources.

[0046] S4: The dominant controller generates a collaborative control strategy based on the locally generated control strategy and by integrating the control strategies and corresponding confidence information of at least one adjacent edge intelligent decision-making node. This step is the core of collaborative decision-making, where the leading controller makes more robust and better decisions based on a combination of local intelligence and neighboring opinions.

[0047] The specific implementation method is as follows: Local policy generation: The dominant controller invokes its local intelligent decision-making model (such as a model trained based on deep reinforcement learning) to generate a preliminary local control policy (e.g., a suggested reactive power compensation amount) based on its own feature indicators and state history. Simultaneously, the model evaluates the reliability of this policy and outputs a confidence score.

[0048] Collect neighbor policies: The dominant controller requests the control policies and corresponding confidence levels generated by its neighboring auxiliary actuator nodes based on their local perspectives.

[0049] Policy Fusion: After receiving all information, the dominant controller employs a confidence-weighted fusion algorithm. This algorithm weights and averages the policies proposed by each node (including itself) according to their confidence levels. Formulated as: Final Policy = Σ(Nodes) i Strategy × Node i (Confidence of the node) / Σ(Confidence of all nodes).

[0050] Generate a cooperative control strategy: Through the above fusion, a cooperative control strategy is obtained that comprehensively considers the local perspectives of multiple nodes and the reliability of their individual judgments. This strategy is theoretically superior to the independent decision-making of any single node and is better adapted to the interconnected characteristics of the power grid.

[0051] S5: According to the aforementioned coordinated control strategy, control the power quality management equipment connected to the relevant power grid nodes to perform coordinated actions.

[0052] This step translates digital world decisions into physical world actions, completing the closed loop of power quality governance.

[0053] The specific implementation method is as follows: Command decomposition and issuance: The master controller parses the fused collaborative control strategy into a set of specific, executable device control commands. For example, for a static var generator, the command might be "output reactive power". Q =+5MVar”; For an active power filter, the instruction might be “Inject 5th harmonic compensation current, amplitude 50A, phase opposite”.

[0054] Coordinated Actions: These instructions are precisely distributed to power quality management devices connected to relevant grid nodes via the control network. Upon receiving the instructions, each device executes them synchronously within a very short time (milliseconds). Because the instructions originate from a collaborative strategy that integrates information from multiple nodes, the actions of management devices at different locations are mutually coordinated and consistent, thereby achieving rapid and coordinated management of voltage fluctuations, harmonics, and other issues.

[0055] In some embodiments, step S0 is included before step S1: S0: Based on the improved multi-objective dung beetle optimization algorithm, the deployment location of the synchronous phasor measurement unit is optimized to determine the power grid node where the synchronous phasor measurement unit is deployed; wherein, the optimization configuration takes the overall network observation accuracy, equipment investment cost and communication reliability of the power grid node as optimization objectives.

[0056] The objective function of the improved multi-objective dung beetle optimization algorithm is:

[0057] In the formula, x Represents a vector of decision variables; N Indicates the total number of nodes in the power grid; M Indicates the total number of lines; x i Indicates the first i The decision variables for whether to deploy a synchronous phasor measurement unit at a power grid node; c i Indicates the first i The cost of deploying a synchronous phasor measurement unit at a power grid node; w ij Indicates the first i The first power grid node j The weighting coefficient of each line; o ij Indicates the first i The first power grid node is related to the first j The observation capability coefficient of each line; r i : No. i Communication reliability indicators for each power grid node.

[0058] This step aims to solve the "first mile" problem of distributed decision networks—that is, how to scientifically and rationally determine the deployment location of synchronous phasor measurement units (PMUs) in a more economical and reliable manner while ensuring the overall observation performance of the system. This is a typical multi-objective optimization problem that requires a balance between technical performance, economic cost, and engineering reliability.

[0059] The specific implementation method is as follows: Define a binary decision vectorx = [ x 1, x 2, ... , x N ],in N This represents the total number of nodes in the power grid. If... x i = 1, then it means the first time i PMUs are deployed at each power grid node; if x i = 0, then no deployment.

[0060] Optimization objectives (forming a multi-objective optimization function): Objective 1: Maximize overall network observation accuracy. Ensure that the deployed PMUs can observe the dynamic state of the entire power grid as completely and accurately as possible. This typically translates to maximizing the observability of critical lines and nodes, mathematically expressed as maximizing the weighted sum of the observation capabilities, as shown in the following formula: .

[0061] Objective 2: Minimize equipment investment costs. The PMU and its associated equipment are expensive, necessitating cost control. The cost objective function is as follows: c i In the first i The overall cost formula for deploying a PMU on one node is as follows: .

[0062] Objective 3: Maximize the communication reliability of deployed nodes. To ensure the real-time performance and reliability of subsequent collaborative decisions, nodes with well-developed communication infrastructure, good channel quality, and high redundancy should be prioritized. This objective can be expressed as maximizing the communication reliability index of the selected nodes. r i weighted sum .

[0063] Constraints include, but are not limited to: constraints to ensure full observability of the entire network state (i.e., the voltage phasor of any node can be calculated directly or indirectly by the deployed PMU), constraints on special network topologies such as zero-injection nodes, and budget limits.

[0064] To solve the aforementioned complex high-dimensional, discrete, multi-objective optimization problem, this embodiment employs an improved multi-objective dung beetle optimization algorithm.

[0065] The scheme provided in this application dynamically adjusts based on the population's evolutionary stage and individual fitness, conducting large-scale exploration in the early stages of iteration and fine-tuning in the later stages, effectively balancing global and local search capabilities. In each generation of evolution, superior individuals (elite solutions) in the non-dominated solution set are forcibly retained to directly enter the next generation, preventing the loss of high-quality solutions during evolution, accelerating algorithm convergence, and improving the quality of the solution set.

[0066] After the algorithm runs, it outputs a Pareto optimal solution set. Each solution in this set represents a PMU deployment scheme in which it is impossible to improve any one objective without weakening at least one other objective.

[0067] From the obtained Pareto optimal solution set, a multi-attribute decision-making method such as entropy weight-TOPSIS is used to select a final optimal compromise solution based on the preference weights of the three objectives of observation accuracy, cost, and reliability in actual engineering. This solution clarifies the specific power grid nodes where PMUs need to be deployed.

[0068] Step S0, as an optional but highly preferred preparatory step, ensures that the "distributed decision-making network" constructed in the subsequent step S1 is itself an optimized network that achieves the best balance between performance, cost, and reliability, thereby improving the overall practical value and economy of the invention.

[0069] In some embodiments, the power quality characteristics include at least one of voltage deviation rate, total harmonic distortion rate, and frequency fluctuation rate.

[0070] The raw synchronization phasor data stream is extracted into key features that can directly and quantitatively characterize the core issues of power quality. The selected indicators cover the three most important aspects of power quality: voltage quality, waveform quality, and frequency quality, providing accurate and concise input for subsequent intelligent decision-making.

[0071] Voltage deviation rate: This measures the degree of deviation between the actual operating voltage of a node and the rated voltage of the system. It is a core indicator for evaluating voltage stability and is used to quickly detect voltage drops, surges, or prolonged over / undervoltage events. This indicator is a key input for evaluating voltage stability and triggering reactive power compensation control.

[0072] Total Harmonic Distortion (THD) is the ratio of the total effective value of harmonic components to the effective value of the fundamental component in an AC signal. It quantifies the degree to which a waveform deviates from a standard sine wave and is used to detect the level of harmonic pollution generated by nonlinear loads, inverters, etc., in the power grid. This indicator is the direct basis for controlling active power filters for harmonic mitigation.

[0073] Frequency fluctuation rate: measures the relative instantaneous deviation between the actual frequency and the rated frequency of the system. It reflects the real-time balance between power generation and load, as well as the system's inertial response. It is used to monitor system frequency stability and detect frequency events caused by high-power load switching, new energy power fluctuations, etc. It is an important state variable for assessing the safe and stable operation of the system.

[0074] By compressing raw phasor data—which consists of tens of frames per second, each containing multi-channel amplitude and phase information—into a few key scalar indicators, data simplification is achieved by hundreds of times, completely resolving the bandwidth bottleneck problem of transmitting massive amounts of data to the central node. The extracted indicators directly correspond to core power quality issues defined in international and national standards, allowing subsequent decision-making algorithms to focus directly on the essence of the problem without processing redundant information, thus improving decision-making efficiency. The calculations are simple and fast, and can be completed in real-time on the limited computing resources of edge intelligent decision-making nodes, meeting the millisecond-level response speed requirements of distributed systems. These indicators are clearly defined and calculated in a standardized manner, ensuring the comparability of feature values ​​generated by different nodes and devices from different manufacturers, providing a unified "language" foundation for inter-node collaborative negotiation and strategy fusion.

[0075] In some embodiments, refer to Figure 2 , Figure 3 The dynamic role negotiation in step S3 specifically includes: S31: Each edge intelligent decision node calculates the power quality severity index of its respective power grid node; Multiple independent power quality characteristics (such as voltage deviation rate, THD, and frequency fluctuation rate) are combined into a unified scalar to quantify the overall severity of power quality problems at a given node. Each node calculates its power quality severity index using the following formula.

[0076] In the formula, S i Represents a power grid node i The power quality severity index; Δ U i Represents a power grid node i Voltage deviation; U n Indicates the rated voltage; T i Represents a power grid node iTotal harmonic distortion; Δ f i Represents a power grid node i Frequency deviation; f n Indicates the rated frequency; α , β , c Indicates the weighting coefficient ( α + β + c = 1).

[0077] S32: Edge intelligent decision-making nodes exchange the power quality severity index and governance capability information; Each node broadcasts a data packet to all its neighboring nodes via a low-latency local communication network. This packet contains at least the following: Power quality severity index for this node S i .

[0078] Information on the governance capabilities of this node, such as the current available reactive power capacity of the static var generator connected to it, and the rated filter current of the active power filter.

[0079] Communication range: The exchange only takes place between adjacent nodes connected by direct communication links, forming local and overlapping "information sensing domains" rather than flooding the entire network, thus ensuring communication efficiency.

[0080] S33: Based on the exchanged information, the contribution of each edge intelligent decision-making node to the collaborative governance alliance is calculated by improving the Shapley value algorithm; After collecting information about its neighboring nodes, each node considers itself and these neighboring nodes as a potential "cooperative governance alliance".

[0081] Contribution calculation: An improved Shapley value algorithm is used to calculate the contribution of each node in the consortium. i Contribution φ i ( v The Shapley value is a classic method in cooperative game theory for fairly distributing the total payoff of an alliance among its members. In this invention, "payoff" is defined as governance effectiveness.

[0082] Governance effectiveness function v ( S ): For any subset (alliance) of the alliance S ,function v ( SThis estimates the overall effectiveness of the power quality problems that the nodes within the subset can solve if they work together. The calculation takes into account the sum of the severity indices of the nodes within the subset (reflecting the total number of problems) and the sum of their governance capabilities (reflecting their problem-solving capabilities).

[0083] Improved Shapley Value Calculation: Traditional Shapley value calculation requires calculating the marginal contribution of a node to all possible sub-alliances, and the computational complexity increases exponentially with the number of nodes. This invention improves upon this: Monte Carlo sampling approximation: Since the number of local federation nodes is usually small (e.g., 3-7), but still to reduce computational overhead, the Monte Carlo method can be used to randomly generate a partial federation sequence to approximate the Shapley value, instead of traversing all possibilities.

[0084] Electrical distance weighting: When calculating marginal contributions, not only the absolute capabilities of nodes are considered, but electrical distance is also introduced as a weight. Nodes with closer electrical distances have stronger coupling and mutual influence in their governance behaviors. Therefore, when evaluating the contribution of a node joining a coalition, higher weight is given to its synergistic effects with electrically neighboring nodes within the coalition.

[0085] The specific calculation formula is as follows:

[0086] In the formula, f i ( v ) represents a node i Shapley value; N The set of all nodes; S express N The middle does not contain nodes i any subset; | S |for sets S The number of elements; | N |for sets N The number of elements; v ( S ) represents the governance effectiveness function value of alliance S.

[0087] S34: Based on the contribution level, dynamically allocate the roles of the dominant controller and the auxiliary actuator in each edge intelligent decision-making node.

[0088] Within the current local alliance, all nodes are sorted from highest to lowest according to their contribution.

[0089] Dominant Controller: The node with the highest contribution is dynamically elected as the dominant controller for this control cycle. It is responsible for executing subsequent policy fusion and instruction generation tasks.

[0090] Auxiliary Executors: All other nodes act as auxiliary executors. Their primary responsibility is to execute the final issued collaborative control commands and continue to provide local information and strategies in the next round of negotiation.

[0091] Dynamic nature: This allocation is not fixed. Whenever the power grid state changes (such as the occurrence of a new disturbance or the mitigation of an existing disturbance), the severity index and mitigation capability information of each node are updated accordingly. Steps S31-S34 are re-executed periodically or by event triggering, which may cause the identity of the dominant controller to dynamically transfer between different nodes.

[0092] The scheme provided in this application theoretically guarantees fairness in role allocation based on the Shapley value allocation mechanism, with nodes contributing more assuming more core responsibilities, which aligns with the principle of optimal overall system efficiency. The control core dynamically migrates with the power grid status, enabling the system to automatically adapt to changes such as faults and disturbances. If the current dominant controller fails, a new replacement will be automatically selected in the next round of negotiation, greatly improving the system's fault tolerance and survivability. The entire negotiation process is completed autonomously among local nodes without any assignment or intervention from any upper-level central node, achieving truly decentralized and self-organizing intelligent control.

[0093] In some embodiments, refer to Figure 4 The final fusion control strategy is generated using the following formula:

[0094] In the formula, i * Indicates the final fusion control strategy; i i Indicates the first i The control strategy proposed by each node; C i For the first i Confidence of the strategy for each node; K The number of nodes participating in the fusion.

[0095] This fusion method integrates the local perspectives and information of multiple nodes, effectively overcoming decision-making biases or errors caused by the limited observation range or model uncertainty of individual nodes. Even if individual nodes provide suboptimal or even incorrect strategies, their negative impact on the final decision will be reduced as long as their confidence level is low. By introducing confidence level as a dynamic weight, the system can automatically identify and trust nodes that are "more knowledgeable" in specific situations. For example, when harmonic disturbances occur, nodes closer to the harmonic source may naturally have higher confidence in their models regarding harmonic compensation strategies, thus gaining greater weight in the fusion. This formula is the mathematical core of distributed collaboration, moving from "each fighting their own battle" to "collective wisdom." It makes the final collaborative control strategy no longer a command from a single node, but a weighted consensus of the collective wisdom of the entire local node group, which is more in line with the physical characteristics of interconnected power grids, thereby producing a governance effect of "1+1>2".

[0096] In some embodiments, the local control strategy of each edge intelligent decision node is generated by a deep reinforcement learning model, wherein the confidence level... C i The value function and state access frequency are determined based on the corresponding deep reinforcement learning model.

[0097] In some embodiments, the power quality management device includes a static var generator (SVA) and an active power filter (APF); the coordinated control strategy includes: a reactive power compensation reference value command sent to the SVA; and a harmonic current compensation reference value command sent to the APF.

[0098] The dominant controller generates a collaborative control strategy based on the fusion. i * This is interpreted as precise control commands for different treatment devices.

[0099] Static var generators dynamically adjust the voltage at the connection point by rapidly generating or absorbing reactive power. They are mainly used to control voltage deviation, flicker, and improve system stability.

[0100] Strategy i * The portion corresponding to the SVG directly manifests as the reactive power compensation reference value command. The calculation of this command follows the basic principles of voltage-reactive power droop control to ensure both rapid response and stability. In a preferred embodiment, the instruction is determined by the following formula:

[0101] In the formula, Q ref This indicates the reference value for reactive power compensation. K qThis represents the reactive power control coefficient; it is a key parameter that can be dynamically adjusted, and its magnitude determines the system's response strength to voltage deviations. It can be adaptively tuned based on the system's short-circuit capacity, the SVG's own capacity, and the current operating mode.

[0102] Δ U Indicates voltage deviation, according to strategy i * The target voltage deviation calculated from real-time data may be the local voltage deviation or the "equivalent cooperative voltage deviation" after incorporating information from neighboring nodes.

[0103] Active power filters detect load harmonic currents in real time and inject a compensation current with equal amplitude but opposite phase to cancel harmonics and purify the power grid current waveform.

[0104] Strategy i * The part corresponding to the APF (Advanced Persistent Factor) is the harmonic current compensation reference value instruction. The core of this instruction is to determine the amplitude and phase of the harmonic current that needs to be compensated.

[0105] In a preferred embodiment, the amplitude reference value of the command is determined by the following formula:

[0106] I h_ref This indicates the reference value for harmonic current compensation; K h This represents the harmonic control coefficient; it is a dynamically adjustable gain coefficient used to regulate the intensity and speed of harmonic mitigation. Through distributed collaborative decision-making, K h Coordination can be carried out between adjacent APFs to avoid overcompensation or governance conflicts.

[0107] T h Total harmonic distortion (THD); Strategy i * The identified total harmonic distortion rate or specified subharmonic content that needs to be addressed. This value incorporates assessments of the harmonic situation from multiple nodes.

[0108] I 0 represents the fundamental current, the currently detected effective value of the fundamental current, which is used to convert the per-unit value of the distortion rate into the actual current compensation amount.

[0109] Transforming high-level strategies into precise mathematical instructions tailored to the physical characteristics of the equipment enables a seamless and accurate transition from "decision-making" to "action." Control coefficients can be based on... i *Dynamic adjustments enable the system to flexibly adapt to different grid operating modes and disturbance types, achieving adaptive optimization control. This ensures that, under the distributed decision-making framework, the final physical actions executed are coordinated and unified, producing globally optimal governance results and effectively solving complex power quality problems such as voltage sags and harmonic amplification.

[0110] This application also provides a power quality optimization control system based on distributed collaborative decision-making, including: A distributed measurement module is used to deploy synchronized phasor measurement units at multiple power grid nodes to collect data; Multiple edge intelligent decision-making modules are deployed at each power grid node in a one-to-one correspondence with the synchronous phasor measurement unit, and are used to calculate power quality characteristic indicators based on local data; The role negotiation module is used to enable each edge intelligent decision-making module to dynamically negotiate its role with adjacent edge intelligent decision-making modules through local communication based on the power quality characteristic indicators, so as to determine the dominant control module. The strategy fusion module is built into the edge intelligent decision module that is determined to be the dominant control module. It is used to generate a collaborative control strategy based on the locally generated control strategy and by fusing the control strategies and corresponding confidence information of at least one adjacent edge intelligent decision module. The control execution module is used to control the operation of the power quality management equipment according to the collaborative control strategy.

[0111] The power quality optimization control system based on distributed collaborative decision-making provided in this application will be described below with reference to specific embodiments: The distributed synchronous phasor measurement unit group is deployed at key nodes of the power grid to collect synchronous measurement data of voltage, current, frequency, and phase angle in real time. The edge intelligent decision-making node is configured in a one-to-one correspondence with the synchronous phasor measurement unit, and has local computing and decision-making capabilities; The collaborative communication network adopts the 5G-U lightweight communication protocol to enable direct communication between adjacent edge intelligent decision-making nodes; The power quality management equipment includes a static var generator and an active power filter, and accepts control commands from edge intelligent decision nodes.

[0112] In practical applications, power quality optimization control systems based on distributed collaborative decision-making are used to execute the following methods: This paper proposes an optimized PMU (Personal Utility Unit) configuration based on an improved multi-objective dung beetle optimization algorithm. It comprehensively considers multiple objectives, including observation accuracy, investment cost, and communication reliability, to determine the optimal PMU deployment scheme. This method can significantly reduce equipment investment costs and improve system economy while ensuring network-wide observability.

[0113] Each PMU node collects real-time power grid operation data, and edge intelligent decision nodes extract key power quality characteristic indicators. Through local feature extraction, the amount of data transmission is significantly reduced, lowering communication bandwidth requirements.

[0114] Based on a dynamic role allocation mechanism, each node determines its role as the dominant controller and auxiliary actuator through local communication negotiation. This mechanism ensures that the system can adaptively adjust its control structure when the power grid's operating state changes.

[0115] The dominant controller generates an initial control strategy based on a local reinforcement learning model, and integrates information from neighboring nodes through a confidence-weighted fusion algorithm to form the optimal cooperative control strategy.

[0116] By implementing a coordinated control strategy, each power quality management device coordinates its actions according to the assigned instructions, achieving rapid and accurate power quality management.

[0117] Evaluate the control effectiveness and dynamically adjust the role allocation weights and control parameters based on performance indicators to achieve continuous system optimization.

[0118] Specifically, determining the optimal deployment scheme for the PMU includes: Establish a power grid topology model to identify key observation nodes and important lines; Initialize the parameters of the improved multi-objective dung beetle optimization algorithm, including population size, maximum number of iterations, and rolling factor; Construct constraints for PMU configuration, including zero-injection node constraints and communication reliability constraints; Execute the optimization algorithm to find the Pareto optimal solution set; The final configuration scheme is selected from the Pareto solution set based on the entropy weight-TOPSIS method.

[0119] Among them, the improved multi-objective dung beetle optimization algorithm introduces an adaptive rolling mechanism and an elite retention strategy on the basis of the traditional dung beetle algorithm, which significantly improves the convergence speed and solution quality.

[0120] Furthermore, the specific implementation of the dynamic role allocation mechanism includes: Construct a local set of neighboring nodes to determine the communication range; Periodically calculate the power quality severity index for each node; The contribution of each node to collaborative governance is evaluated based on the improved Shapley value algorithm; The dominant controller role is dynamically assigned based on contribution ranking.

[0121] Key technical points of the confidence-weighted fusion algorithm: Each node’s local deep reinforcement learning model uses the DDPG algorithm to adapt to the control requirements of continuous action space; Confidence calculation is based on the policy's value function and state access frequency; The integration process takes into account the differences in electrical distance and governance capabilities between nodes.

[0122] It is understood that the same or similar parts in the above embodiments can be referred to each other, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0123] It should be noted that in the description of this invention, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, in the description of this invention, unless otherwise stated, "a plurality of" means at least two.

[0124] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0125] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0126] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0127] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0128] The storage media mentioned above can be read-only memory, disk, or optical disk, etc.

[0129] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0130] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A power quality optimization control method based on distributed collaborative decision-making, characterized in that, Includes the following steps: S 1: At multiple nodes of the power grid, synchronous phasor measurement units and corresponding edge intelligent decision-making nodes are deployed to form a distributed decision-making network; S 2: Each edge intelligent decision node calculates power quality characteristic indicators based on local data collected by its associated synchronous phasor measurement unit; S 3: Each edge intelligent decision-making node, based on the power quality characteristic indicators, conducts dynamic role negotiation with neighboring edge intelligent decision-making nodes through local communication to determine the dominant controller from the negotiating parties; S 4: The dominant controller generates a collaborative control strategy based on the locally generated control strategy and by integrating the control strategies and corresponding confidence information of at least one adjacent edge intelligent decision-making node. S 5: According to the aforementioned coordinated control strategy, control the power quality management equipment connected to relevant power grid nodes to perform coordinated actions.

2. The method according to claim 1, characterized in that, In the steps S Before step 1, there are also steps. S 0: S 0: Based on the improved multi-objective dung beetle optimization algorithm, the deployment location of the synchronous phasor measurement unit is optimized to determine the power grid nodes where the synchronous phasor measurement unit is deployed; wherein, the optimization configuration takes the overall network observation accuracy, equipment investment cost and communication reliability of the power grid nodes as optimization objectives.

3. The method according to claim 2, characterized in that, The objective function of the improved multi-objective dung beetle optimization algorithm is: In the formula, x Represents a vector of decision variables; N Indicates the total number of nodes in the power grid; M Indicates the total number of lines; x i Indicates the first i The decision variables for whether to deploy a synchronous phasor measurement unit at a power grid node; c i Indicates the first i The cost of deploying a synchronous phasor measurement unit at a power grid node; w ij Indicates the first i The first power grid node j The weighting coefficient of each line; o ij Indicates the first i The first power grid node is related to the first j The observation capability coefficient of each line; r i : No. i Communication reliability indicators for each power grid node.

4. The method according to claim 1, characterized in that, step S In step 2, the power quality characteristic indicators include at least one of voltage deviation rate, total harmonic distortion rate, and frequency fluctuation rate.

5. The method according to claim 1, characterized in that, step S The dynamic role negotiation described in section 3 specifically includes: S 31: Each edge intelligent decision node calculates the power quality severity index of its respective power grid node; S 32: Edge intelligent decision-making nodes exchange the power quality severity index and governance capability information; S 33: Based on the exchanged information, the contribution of each edge intelligent decision-making node in the collaborative governance alliance is calculated by improving the Shapley value algorithm; S 34: Based on the contribution level, dynamically allocate the roles of the dominant controller and the auxiliary executor in each edge intelligent decision-making node.

6. The method according to claim 5, characterized in that, The formula for calculating the power quality severity index is as follows: In the formula, S i Represents a power grid node i The power quality severity index; Δ U i Represents a power grid node i Voltage deviation; U n Indicates the rated voltage; T i Represents a power grid node i Total harmonic distortion; Δ f i Represents a power grid node i Frequency deviation; f n Indicates the rated frequency; α , β , γ Indicates the weighting coefficient ( α + β + γ = 1).

7. The method according to claim 1, characterized in that, The final fusion control strategy is generated using the following formula: In the formula, θ * This indicates the final fusion control strategy; θ i Indicates the first i The control strategy proposed by each node; C i For the first i Confidence of the strategy for each node; K This refers to the number of nodes participating in the fusion.

8. The method according to claim 1, characterized in that, The local control strategies of each edge intelligent decision-making node are generated by a deep reinforcement learning model, and the confidence level... C i The value function and state access frequency are determined based on the corresponding deep reinforcement learning model.

9. The method according to claim 1, characterized in that, step S In section 5, the power quality management equipment includes a static var generator and an active power filter; The coordinated control strategy includes: issuing a reactive power compensation reference value command to the static var generator; and issuing a harmonic current compensation reference value command to the active power filter.

10. A power quality optimization control system based on distributed collaborative decision-making, characterized in that, include: A distributed measurement module is used to deploy synchronized phasor measurement units at multiple power grid nodes to collect data; Multiple edge intelligent decision-making modules are deployed at each power grid node in a one-to-one correspondence with the synchronous phasor measurement unit, and are used to calculate power quality characteristic indicators based on local data; The role negotiation module is used to enable each edge intelligent decision-making module to dynamically negotiate its role with adjacent edge intelligent decision-making modules through local communication based on the power quality characteristic indicators, so as to determine the dominant control module. The strategy fusion module is built into the edge intelligent decision module that is determined to be the dominant control module. It is used to generate a collaborative control strategy based on the locally generated control strategy and by fusing the control strategies and corresponding confidence information of at least one adjacent edge intelligent decision module. The control execution module is used to control the operation of the power quality management equipment according to the collaborative control strategy.

Images