Low-voltage ac-dc hybrid distribution network power quality collaborative treatment and optimal allocation method, system and device based on improved intelligent algorithm and medium

Abstract

The present application relates to the technical field of power quality analysis, and discloses a low-voltage AC-DC hybrid distribution network power quality collaborative treatment and optimal configuration method and system based on an improved intelligent algorithm, which comprises: constructing a low-voltage AC-DC hybrid distribution network topology model; determining the comprehensive weight coefficients of various power quality indicators by combining subjective and objective weighting methods, and performing power quality comprehensive scoring on the topology model according to the comprehensive weight coefficients; using an improved firefly algorithm to solve a double-objective optimization function constructed by the equipment investment cost and the power quality comprehensive score to obtain an offline planning layer configuration scheme, and constructing a Markov decision process model; in online operation, using a deep Q network to solve the Markov decision model to predict the treatment benefits, and selecting physical device control parameters. The present application can effectively suppress the frequent action of energy storage and other devices to prolong their service life, and enhance the adaptive adjustment capability and operation stability of the distribution network in high dynamic fluctuation scenarios.

Description

Technical Field

[0001] This invention relates to the field of power quality optimization technology, and in particular to a method, system, device and medium for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms. Background Technology

[0002] With the large-scale integration of new power sources such as distributed photovoltaic power and electric vehicle charging stations, traditional low-voltage AC distribution networks are transforming into low-voltage AC / DC hybrid distribution networks. This has brought about some power quality issues, and existing mitigation methods mainly focus on three levels: planning and configuration, real-time control, and hardware-based management. In terms of planning and configuration, an evaluation system is typically established using the analytic hierarchy process (AHP) combined with the entropy weight method, and heuristic algorithms are used to statically optimize the installation location and capacity of reactive power compensation devices and active power filters. Regarding real-time control, deep reinforcement learning algorithms are often introduced to address issues such as voltage exceeding limits and three-phase imbalance in real time, and power flow is adjusted by dynamically controlling flexible interconnection devices and reactive power compensation equipment. Furthermore, some weak power grids also utilize hardware methods combining static capacitors with dynamic reactive power compensation equipment to reduce network losses and improve the system power factor.

[0003] However, the aforementioned technologies have shortcomings when facing new AC / DC hybrid distribution networks. There is a disconnect between equipment configuration and real-time control. The planning phase typically only performs static optimization based on typical scenarios, neglecting the dynamic response requirements brought about by high-frequency fluctuating source loads. The control phase is often limited to fixed, known equipment layouts, lacking an overall consideration of layout rationality. Furthermore, existing research generally suffers from singular governance objectives or overly subjective evaluations. Some methods focus only on a specific power quality indicator, or rely solely on subjective weighting in multi-objective evaluations, lacking objectivity. With the continuous increase in source load penetration, voltage exceeding limits, harmonic distortion, and three-phase imbalance often occur concurrently. Traditional closed-loop planning plus open-loop operation modes are insufficient to flexibly address such complex concurrent problems. In addition, the system's utilization rate for AC / DC collaborative governance is low, failing to fully explore the proactive governance potential of voltage source converters in power mutual assistance between AC and DC sides, resulting in their ineffective release of proactive execution capabilities in power quality governance. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method and system for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms, which solves the problems of voltage exceeding limits, three-phase imbalance and concurrent harmonic distortion caused by the high proportion of new energy access in current new low-voltage AC / DC hybrid distribution networks.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on an improved intelligent algorithm, including: Acquire distribution network data and construct a low-voltage AC / DC hybrid distribution network topology model; The comprehensive weighting coefficients of each power quality indicator are determined by combining subjective and objective weighting methods, and the power quality of the topology model is comprehensively scored based on the comprehensive weighting coefficients. An improved firefly algorithm is used to solve a bi-objective optimization function constructed based on equipment investment cost and the comprehensive power quality score, in order to determine the installation location and capacity of power equipment and obtain an offline planning layer configuration scheme. Based on the network environment under the offline planning layer configuration scheme, a Markov decision process model is constructed. During online operation, the Markov decision process model is solved using a deep Q-network. The state vector converted from real-time physical operation data is input into the deep Q-network to predict the expected governance benefits corresponding to each power quality governance action in the action space. The target control action is selected, decoded into physical device control parameters, and the physical device control parameters are sent to the corresponding device for governance.

[0007] As a preferred embodiment of the collaborative governance and optimization configuration method for power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms described in this invention, the determination of the comprehensive weight coefficients for various power quality indicators using a subjective and objective weighting method includes: The subjective weights of each power quality indicator are calculated using the analytic hierarchy process (AHP), and the objective weights of each power quality indicator are calculated using the entropy weight method. The subjective weights and objective weights are fused using the least squares method to obtain the final comprehensive weight coefficients for each power quality indicator. The comprehensive power quality score is obtained by weighting and summing the comprehensive weight coefficients of each item with the normalized voltage deviation index, harmonic current content index, and three-phase imbalance index.

[0008] As a preferred embodiment of the low-voltage AC / DC hybrid distribution network power quality collaborative governance and optimization configuration method based on improved intelligent algorithms described in this invention, the method comprises: using an improved firefly algorithm to solve a bi-objective optimization function constructed based on equipment investment cost and the comprehensive power quality score, to determine the installation location and capacity of power equipment, and obtaining an offline planning layer configuration scheme, including: In the low-voltage AC / DC hybrid distribution network topology model, the governance equipment adopts a dynamic-static combined configuration strategy; the governance equipment includes static reactive power compensation devices and dynamic reactive power compensation devices. The equipment investment cost in the bi-objective optimization function includes the capacity cost function of the static reactive power compensation device and the dynamic reactive power compensation device. The improved firefly algorithm is optimized by using the position and attraction update formula to minimize the comprehensive power quality evaluation index corresponding to the calculated comprehensive power quality score.

[0009] As a preferred embodiment of the collaborative governance and optimized configuration method for power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms described in this invention, a Markov decision process model is constructed based on the network environment under the offline planning layer configuration scheme, including: The Markov decision process model includes a state space, an action space, and a reward function. The reward function includes: power quality related limit violation penalty items, which specifically include AC side voltage limit violation penalty items, DC side voltage limit violation penalty items, and AC side three-phase unbalance limit violation penalty items. The reward function also includes: a penalty for solar curtailment and a penalty for frequent energy storage operations; Each penalty item has a corresponding weight coefficient set in the reward function, and the weight coefficient is preset according to the governance priority of different power quality problems.

[0010] As a preferred embodiment of the collaborative governance and optimal allocation method for power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms described in this invention, the method includes: solving the Markov decision process model using a deep Q-network, including: By utilizing measurement equipment deployed at key nodes of the distribution network, real-time data are collected on voltage amplitude, branch power flow, and real-time status of controllable flexible equipment, including AC and DC side nodes. The collected data are normalized and then concatenated into a current state vector in the form of a long vector according to the state space format defined in the Markov decision process model. The current state vector is input into the input layer of the deep Q-network.

[0011] As a preferred embodiment of the low-voltage AC / DC hybrid distribution network power quality collaborative governance and optimization configuration method based on improved intelligent algorithms described in this invention, the method involves: converting real-time physical operation data into a state vector and inputting it into the deep Q-network to predict the expected governance benefits corresponding to each power quality governance action in the action space; and selecting target control actions to decode into physical equipment control parameters, including: The deep Q-network is used to predict and output the expected governance benefits corresponding to all control actions in the action space; A greedy strategy is adopted to balance the relationship between exploration and utilization. An action is randomly selected in the action space with a first preset probability for exploration, and the control action with the greatest expected governance benefit is selected as the target control action with a second preset probability. The physical device control parameters formed after the target control action is decoded include: the active and reactive power transmission power setting values ​​of the voltage source converter, the reactive current reference value of the static var generator, and the discrete switching signals of the static capacitor bank.

[0012] As a preferred embodiment of the method for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms described in this invention, the method includes: sending the control parameters of the physical devices to the corresponding devices for governance, including: In response to the real-time monitoring of the AC side voltage exceeding the limit, the voltage source converter is controlled to backfeed active power on both the AC and DC sides, and the static var generator is controlled to output capacitive reactive current, so as to adjust the node voltage to the qualified range. In response to real-time monitoring of excessive three-phase imbalance or harmonic distortion, the static var generator or voltage source converter is controlled to output asymmetrical compensation current to achieve phase-to-phase load balance, and a reverse harmonic current is emitted to cancel the harmonics generated by the nonlinear load. After each device executes the corresponding physical device control parameters, the distribution network environment status at the next moment is obtained and the reward function is updated. The sample consisting of the current state, the target control action, the calculated reward value, and the state at the next moment is stored in the experience replay pool. Batch samples are periodically extracted and the weight parameters of the deep Q network are iteratively updated online using the gradient descent method.

[0013] Secondly, this invention provides a low-voltage AC / DC hybrid distribution network power quality collaborative governance and optimization configuration system based on an improved intelligent algorithm, comprising: The first construction module is used to acquire distribution network data and construct a low-voltage AC / DC hybrid distribution network topology model. The scoring module is used to determine the comprehensive weight coefficients of various power quality indicators by combining subjective and objective weighting methods, and to perform a comprehensive power quality score on the topology model based on the comprehensive weight coefficients. The optimization module is used to solve the bi-objective optimization function constructed by the equipment investment cost and the comprehensive power quality score using the improved firefly algorithm, so as to determine the installation location and capacity of the power equipment and obtain the offline planning layer configuration scheme. The second construction module is used to construct a Markov decision process model based on the network environment under the offline planning layer configuration scheme. The solution module is used to solve the Markov decision process model using a deep Q-network during online operation. It converts real-time physical operation data into state vectors and inputs them into the deep Q-network to predict the expected governance benefits corresponding to each power quality governance action in the action space. It also selects target control actions, decodes them into physical device control parameters, and sends the physical device control parameters to the corresponding devices for governance.

[0014] Thirdly, the present invention provides a computer device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the method for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms.

[0015] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the method for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention establishes a normalized index system through a fusion of subjective and objective weighting, providing a more comprehensive and objective basis for power quality evaluation; and utilizes an improved firefly algorithm to solve the offline bi-objective optimization function, achieving dynamic and static optimization of the installation location and capacity of power management equipment, thereby effectively improving the economy and rationality of equipment configuration during the planning stage; at the online operation level, by transforming the network environment constructed by the offline configuration scheme into a Markov decision process model and using a deep Q-network for solution, the limitations of the traditional open-loop operation mode can be overcome. This approach addresses real-time, adaptive, and closed-loop concurrent governance of multi-dimensional power quality issues such as voltage overshoot, three-phase imbalance, and harmonic distortion under high-frequency random source-load fluctuations. By integrating AC / DC active power support, renewable energy consumption penalties, and equipment operation frequency penalties into the reward mechanism, it explores the proactive governance potential of flexible converters in bidirectional power mutual assistance on both AC and DC sides. It can also effectively suppress frequent operation of energy storage and other equipment to extend their service life. As a result, it achieves overall synergy between offline planning and configuration and online operation control, enhancing the adaptive adjustment capability and operational stability of the distribution network under high dynamic fluctuation scenarios. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. 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.

[0018] Figure 1 This is a schematic diagram of the overall process of a method for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on an improved intelligent algorithm, as described in one embodiment of the present invention.

[0019] Figure 2 This is the overall system architecture and topology diagram of the application of the low-voltage AC / DC hybrid distribution network power quality collaborative governance and optimization configuration method based on improved intelligent algorithm according to an embodiment of the present invention.

[0020] Figure 3 This is a schematic diagram of the specific structure of the method for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms, as described in one embodiment of the present invention. Detailed Implementation

[0021] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0022] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on an improved intelligent algorithm is provided, including: S100: Acquire distribution network data and construct a low-voltage AC / DC hybrid distribution network topology model; S200: Combine subjective and objective weighting methods to determine the comprehensive weighting coefficients of various power quality indicators, and perform a comprehensive power quality score on the topology model based on the comprehensive weighting coefficients; S300: The improved firefly algorithm is used to solve the bi-objective optimization function constructed with equipment investment cost and the comprehensive power quality score, so as to determine the installation location and capacity of power equipment and obtain the offline planning layer configuration scheme. S400: Based on the network environment under the offline planning layer configuration scheme, construct a Markov decision process model; S500: During online operation, the Markov decision process model is solved using a deep Q-network. The state vector converted from real-time physical operation data is input into the deep Q-network to predict the expected governance benefits corresponding to each power quality governance action in the action space. The target control action is selected, decoded into physical device control parameters, and the physical device control parameters are sent to the corresponding device for governance.

[0023] Currently, the configuration phase typically relies on static optimization based on typical scenarios, neglecting the dynamic response capabilities of equipment in real-time operation when facing high-frequency fluctuating source loads. The control phase often assumes the equipment is fixed, lacking consideration for the rationality of equipment layout. Governance objectives often focus on single power quality indicators such as voltage or harmonics, or rely solely on subjective weighting in multi-objective evaluations, lacking objectivity. With increasing source load penetration, the concurrent occurrence of voltage exceeding limits, harmonic distortion, and three-phase imbalance is difficult to flexibly address in a closed-loop planning system combined with open-loop operation. Furthermore, it fails to fully exploit the potential of VSCs (Voltage Source Converters) in power exchange between AC and DC sides, treating them merely as connecting devices rather than proactive power quality governance execution units.

[0024] To address the specific issues mentioned above, steps S100-S500 were established, employing a collaborative governance approach that first optimizes configuration and then controls in real time. This approach ensures both the economic viability and rational layout of governance equipment (static / dynamic compensation devices, VSC, etc.) and guarantees that the system can achieve real-time optimal control of power quality through autonomous decision-making by intelligent agents during operation.

[0025] Example 2, based on the above implementation method, referring to Figures 1-3 As an embodiment of the present invention, a method for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on an improved intelligent algorithm is provided.

[0026] Before describing the specific solution, let's first explain the specific architecture to which this method can be applied. For example, refer to... Figure 2 The overall system architecture and topology diagram.

[0027] Figure 2This description should explain the hardware and information interaction relationships of a low-voltage AC / DC hybrid distribution network, including at least the AC subnetwork, DC subnetwork, Flexible Interconnect Interface (VSC), governance equipment layer, measurement and acquisition devices, and intelligent control center. The AC subnetwork may include a 380V AC bus, AC loads, AC-side distributed photovoltaic (PV) systems, and weak nodes on the AC side; the DC subnetwork may include a 750V DC bus, DC loads, Energy Storage System (ESS), and Photovoltaic Power Generation Unit (PV); the VSC connects the AC and DC buses to achieve bidirectional active / reactive power balance between AC and DC; the governance equipment layer includes Static Var Compensator (SC) and Dynamic Var Compensator (SVG / APF), used for voltage support, reactive power compensation, harmonic suppression, and three-phase imbalance management; the measurement and acquisition devices collect node voltage, current, harmonics, power, energy storage state of charge (SOC), and equipment status; and the intelligent control center is used for operation evaluation, optimized configuration, and Deep Q-learning online control algorithms.

[0028] A collaborative governance and optimization method for power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms includes: S100: Acquire distribution network data and construct a low-voltage AC / DC hybrid distribution network topology model.

[0029] Specifically, in S100, the topology of a low-voltage AC / DC hybrid distribution network is first established as the basic environment for subsequent calculations. The model specifically includes: AC subgrid: Rated voltage 380V, including AC bus, AC load and distributed photovoltaic.

[0030] DC subgrid: Rated voltage 750V (or set according to actual standards), including DC bus, DC load, energy storage system (ESS) and photovoltaic power generation unit (PV).

[0031] Flexible interconnection interface: A voltage source converter (VSC) is used to connect the AC bus and the DC bus, serving as an execution unit for bidirectional power flow.

[0032] Governance equipment layer: Static var compensator (SC) and dynamic var compensator (SVG / APF) are deployed at weak nodes on the AC side to achieve a combination of static and dynamic governance.

[0033] S200: Combine subjective and objective weighting methods to determine the comprehensive weight coefficients of each power quality indicator, and perform a comprehensive power quality score on the topology model based on the comprehensive weight coefficients.

[0034] In the embodiments of this application, reference is made to Figure 1 and Figure 3 The method of combining subjective and objective weighting to determine the comprehensive weighting coefficients of various power quality indicators, as described in S200, includes the following steps A1-A3: A1: The subjective weights of each power quality indicator are calculated using the analytic hierarchy process (AHP), and the objective weights of each power quality indicator are calculated using the entropy weight method. A2: The subjective weights and objective weights are fused using the least squares method to obtain the final comprehensive weight coefficients for each power quality indicator; A3: The comprehensive power quality score is obtained by weighting and summing the comprehensive weight coefficients of each item with the voltage deviation index, harmonic current content index and three-phase unbalance index after normalization.

[0035] Specifically, in sections A1-A3, to accurately quantify the operating status of the distribution network, an evaluation index system is constructed that includes three dimensions: voltage deviation, harmonics, and three-phase imbalance. Voltage deviation is further analyzed... Harmonic current content Three-phase imbalance Normalization is performed.

[0036] Among them, voltage deviation index : In the formula, The total number of sampling points within the evaluation period; For the first One sampling point; For nodes The actual measured voltage value; For nodes The rated voltage value.

[0037] Among them, the harmonic current content index : In the formula, For harmonic order; For nodes The RMS value of subharmonic current; For nodes The effective value of the fundamental current is calculated. The total harmonic distortion (THD) is also calculated as a constraint.

[0038] Among them, the three-phase imbalance index : In the formula, This represents the maximum phase current value among the three-phase currents at the node. This represents the minimum phase current value among the three-phase currents at the node.

[0039] Subsequently, the subjective weights were calculated using the Analytic Hierarchy Process (AHP). The objective weights are calculated using the entropy weight method. Finally, the comprehensive weights are calculated using the least squares method. This yields the final overall power quality score. : In the formula, This is a comprehensive power quality assessment index; the smaller the value, the better the power quality. This represents the overall weighting coefficient for each indicator; These are the normalized values ​​of each component index (voltage, harmonics, unbalance, etc.).

[0040] S300: The improved firefly algorithm is used to solve the bi-objective optimization function constructed with equipment investment cost and the comprehensive power quality score, so as to determine the installation location and capacity of power equipment and obtain the offline planning layer configuration scheme.

[0041] In the embodiments of this application, reference is made to Figure 1 and Figure 3 S300 utilizes an improved firefly algorithm to solve a bi-objective optimization function constructed based on equipment investment cost and the comprehensive power quality score, in order to determine the installation location and capacity of power equipment and obtain an offline planning layer configuration scheme, including the following steps B1-B3: B1: The governance equipment in the low-voltage AC / DC hybrid distribution network topology model adopts a dynamic-static combination configuration strategy; wherein, the governance equipment includes static reactive power compensation devices and dynamic reactive power compensation devices; Specifically, static var compensators use grouped switching capacitor banks to compensate for basic reactive power demand. The response time is in the range of seconds to minutes. The cost is low, but it cannot be continuously adjusted. The dynamic reactive power compensation device uses a static var generator (SVG) to compensate for rapidly fluctuating reactive power components. Its response time is in the millisecond range, and it can achieve continuous and smooth adjustment. Although it is more expensive, it has a better governance effect. The static-dynamic coordination mechanism refers to the static device bearing the basic reactive power component and the low-frequency component of the daily fluctuation curve, while the dynamic device bears the remaining high-frequency fluctuation component. The capacity ratio between the two is determined based on the spectral analysis results of the reactive power time-series fluctuation at the nodes.

[0042] B2: The equipment investment cost in the bi-objective optimization function includes the capacity cost function of the static reactive power compensation device and the dynamic reactive power compensation device; Specifically, the overall score obtained from S200 The dual objectives are optimality and minimum equipment investment cost, determining the optimal installation nodes and capacities for both static and dynamic compensation devices. A dual-objective function is established, namely, a total objective function encompassing both economic cost and governance effectiveness. , can be represented as: Among them, equipment investment costs It consists of two parts: In the formula, For nodes Reactive power capacity to be compensated; The rated capacity benchmark for a single compensation device; This represents the total number of installed nodes.

[0043] It should be noted that a similar capacity-cost function is used for active filters. .

[0044] B3: The improved firefly algorithm is optimized by updating the position and attraction formula to minimize the power quality comprehensive evaluation index corresponding to the calculated power quality comprehensive score. Specifically, in B3, the firefly algorithm is used to solve the above objective function, and the core update formula is as follows: Firefly brightness The corresponding objective function value Higher brightness indicates a better solution.

[0045] attraction It depends on the distance between the two fireflies. The closer the distance, the greater the attraction.

[0046] Location update, fireflies Fireflies with higher brightness Attracted and moved, thereby updating the equipment configuration scheme (location and capacity).

[0047] It should be noted that the S300 is mainly configured for offline planning. It uses an improved firefly algorithm to solve the optimization function with the dual objectives of equipment investment cost and power quality comprehensive score. This determines the installation nodes and capacity of static and dynamic reactive power compensation devices and active filters, forming a basic deployment scheme that combines static and dynamic elements, providing the optimal initial network environment for subsequent online control.

[0048] S400: Based on the network environment under the offline planning layer configuration scheme, construct a Markov decision process model.

[0049] In the embodiments of this application, reference is made to Figure 1 and Figure 3 Based on the network environment under the offline planning layer configuration scheme, S400 constructs a Markov decision process model, including: The Markov decision process model includes a state space, an action space, and a reward function. Specifically, after the equipment is installed, a Markov decision process (MDP) is established to address random source load fluctuations that occur during operation. Wherein, the state space is defined at time. environmental conditions : In the formula, Let P be the active power matrix and Q be the reactive power matrix for each node. This is the voltage magnitude matrix for each node; This refers to the current switching status of the static var compensator. This indicates the real-time power output status of distributed photovoltaic systems. This refers to the state of charge (SOC) and charging / discharging power of the energy storage system. This represents the current transmission power status of the flexible interconnect device.

[0050] The action space defines the control actions of the agent. : In the formula, Action commands representing various controllable devices may include: active / reactive power transmission settings of VSC (continuous variable); switching positions of static capacitors (discrete variable); and charging / discharging commands of energy storage systems (continuous variable).

[0051] The reward function includes: power quality related limit violation penalty items, which specifically include AC side voltage limit violation penalty items, DC side voltage limit violation penalty items, and AC side three-phase unbalance limit violation penalty items. The reward function also includes: a penalty for solar curtailment and a penalty for frequent energy storage operations; Each penalty item has a corresponding weight coefficient set in the reward function, and the weight coefficient is preset according to the governance priority of different power quality problems.

[0052] Specifically, based on the above reward function, in order to achieve multi-objective governance, an immediate reward is defined. If the power quality exceeds the limit, a negative reward (penalty) will be given, which can be represented as: In the formula, The total number of nodes with AC side voltage exceeding the limit (exceeding the limit means the voltage exceeds the limit). scope); This represents the total number of nodes whose DC-side voltage exceeds the limit. The total number of nodes where the three-phase imbalance on the AC side exceeds the limit, such as exceeding 2%; The number of distributed photovoltaic (PV) units that have been cut off or abandoned is used to promote grid integration; there will be penalties for cutting off PV units. The number of energy storage units that operate frequently is used to extend lifespan and suppress frequent operation; The weighting coefficients for each penalty item are set according to the governance priority; for example, prioritizing ensuring voltage compliance, then... Set a larger value.

[0053] It should be noted that the purpose of S400 is to abstract the distribution network environment determined by offline planning into a Markov decision process by constructing a sequential decision model, defining the state vector (real-time electrical quantities), action space (adjustment instructions of each governance device), and reward function (governance benefits), and transforming the online power quality control problem into a sequential decision framework solvable by reinforcement learning.

[0054] S500: During online operation, the Markov decision process model is solved using a deep Q-network. The state vector converted from real-time physical operation data is input into the deep Q-network to predict the expected governance benefits corresponding to each power quality governance action in the action space. The target control action is selected, decoded into physical device control parameters, and the physical device control parameters are sent to the corresponding device for governance.

[0055] In the embodiments of this application, reference is made to Figure 1 and Figure 3 The S500 uses a deep Q-network to solve the Markov decision process model, including the following steps C1-C3: C1: Using measurement equipment deployed at key nodes of the distribution network, real-time data are collected on the voltage amplitude of AC and DC side nodes, branch power flow, and real-time status data of controllable flexible equipment. Specifically, after completing the S400 MDP model definition in C1, it enters the online operation phase. Through the process of sensing, decision-making, execution, and feedback, it performs millisecond / second-level real-time control of the low-voltage AC / DC hybrid distribution network; that is, at any control moment... By deploying measurement equipment at key nodes of the power distribution network, the physical operation data of the system is collected in real time, realizing environmental status perception and vectorization; The data collected includes the voltage amplitude of each node on the AC side mentioned above. DC side node voltage Branch power flow and the real-time status of each flexible device, such as the state of charge of energy storage. SVG current output current, VSC current transmission power, etc.

[0056] C2: Normalize the collected data and concatenate them into a current state vector in the form of a long vector according to the state space format defined in the Markov decision process model. C3: Input the current state vector into the input layer of the deep Q-network.

[0057] Specifically, normalization is used to eliminate the influence of dimensions, and the vectors are concatenated into the current state vector according to the state-space format defined above. The input is fed into a deep Q-network.

[0058] In this embodiment of the application, the state vector converted from real-time physical operation data in S500 is input into the deep Q-network to predict the expected governance benefits corresponding to each power quality governance action in the action space, and the target control action is selected and decoded into physical device control parameters, including the following steps D1-D3: D1: Utilize the deep Q-network to predict and output the expected governance benefits corresponding to all control actions in the action space; Specifically, the state Input a Q-network, and the network outputs the Q-value (i.e., expected long-term reward) for all possible actions in the action space. Select the action with the largest Q-value as the current optimal policy. .

[0059] D2: A greedy strategy is adopted to balance the relationship between exploration and utilization. An action is randomly selected in the action space with a first preset probability for exploration, and the control action with the greatest expected governance benefit is selected as the target control action with a second preset probability. Specifically, adopt Greedy strategy ( -greedy), with a first preset probability Randomly select actions to explore unknown strategies, with a second preset probability. Choose the best action at present to make use of existing experience, and balance the relationship between exploration and utilization.

[0060] D3: The physical equipment control parameters formed after the target control action is decoded by the instruction include: the active and reactive power transmission power setting values ​​of the voltage source converter, the reactive current reference value of the static var generator, and the discrete switching signal of the static capacitor bank. Specifically, D3 will select the abstract action Decoded into specific physical device control parameters, including: VSC command, namely: active and reactive power transmission setting values ​​of voltage source converter, sets the direction and magnitude of active power transmission between AC and DC. ); SVG command, which stands for Static Var Generator Reactive Current Reference Value, sets the reactive current reference value that it needs to generate. Switching commands, namely: discrete switching signals for static capacitor banks, setting the operation signals for static capacitor banks and on-load tap changers.

[0061] After receiving the above control commands, each treatment device works in concert to resolve and generate power quality issues.

[0062] In this embodiment of the application, S500, sending the physical device control parameters to the corresponding device for management, includes: In response to the real-time monitoring of the AC side voltage exceeding the limit, the voltage source converter is controlled to backfeed active power on both the AC and DC sides, and the static var generator is controlled to output capacitive reactive current, so as to adjust the node voltage to the qualified range. For example, when a low-voltage node is detected on the AC side, the agent instructs the VSC to feed excess photovoltaic power from the DC side back to the AC side (active power support), and simultaneously instructs the SVG to output capacitive reactive power (reactive power support), using active / reactive power coordination to raise the node voltage to a qualified range. ).

[0063] In response to real-time monitoring of excessive three-phase imbalance or harmonic distortion, the static var generator or voltage source converter is controlled to output asymmetrical compensation current to achieve phase-to-phase load balance, and a reverse harmonic current is emitted to cancel the harmonics generated by the nonlinear load. For example, when the three-phase imbalance is detected to exceed the standard (>2%), the intelligent agent instructs the SVG or VSC to output an asymmetrical compensation current to transfer the power of the heavily loaded phase to the lightly loaded phase, thereby achieving phase-to-phase load balancing; at the same time, it emits a reverse harmonic current to cancel the harmonics generated by the nonlinear load.

[0064] After each device executes the corresponding physical device control parameters, the distribution network environment status at the next moment is obtained and the reward function is updated. The sample consisting of the current moment status, the target control action, the calculated reward value and the next moment status is stored in the experience playback pool. Batch samples are periodically extracted and the weight parameters of the deep Q network are iteratively updated online using the gradient descent method. Specifically, the aforementioned reward feedback and online model iteration translate into action execution, after which the power grid state changes from... Transition to the next state The system updates itself based on the governance results.

[0065] The reward calculation is based on the reward function defined in S400 to calculate the reward value at the current moment. If the voltage returns to normal, the three phases are balanced, and the photovoltaic absorption rate is high, a positive reward will be given; if there are still devices that exceed limits or frequently operate, a negative reward (penalty) will be given.

[0066] Experience playback and updating involve using quadruple samples. The samples are stored in an experience replay pool, and batches of samples are randomly drawn from the pool periodically. The weight parameters of the Q network are updated using the gradient descent method, so that the agent can continuously adapt to the source load fluctuation characteristics and improve the accuracy of the control strategy.

[0067] It should be noted that the S500 is used for online real-time governance. The deep Q-network solves the MDP in real time, maps the physical operation data collected by the sensors into a state input network, predicts the expected benefits of each governance action, selects the optimal action, decodes it into specific control parameters and sends them to the equipment, thereby realizing dynamic adaptive closed-loop regulation.

[0068] Example 3 illustrates a schematic scheme for a collaborative governance and optimization configuration method for power quality in low-voltage AC / DC hybrid distribution networks based on an improved intelligent algorithm. It should be noted that the technical solution of this system for collaborative governance and optimization configuration of power quality in low-voltage AC / DC hybrid distribution networks based on an improved intelligent algorithm is based on the same concept as the aforementioned method for collaborative governance and optimization configuration of power quality in low-voltage AC / DC hybrid distribution networks. Details not described in detail in this embodiment can be found in the description of the aforementioned method for collaborative governance and optimization configuration of power quality in low-voltage AC / DC hybrid distribution networks based on an improved intelligent algorithm.

[0069] This embodiment provides a collaborative governance and optimization configuration system for power quality in low-voltage AC / DC hybrid distribution networks based on an improved intelligent algorithm, including: The first construction module is used to acquire distribution network data and construct a low-voltage AC / DC hybrid distribution network topology model. The scoring module is used to determine the comprehensive weight coefficients of various power quality indicators by combining subjective and objective weighting methods, and to perform a comprehensive power quality score on the topology model based on the comprehensive weight coefficients. The optimization module is used to solve the bi-objective optimization function constructed by the equipment investment cost and the comprehensive power quality score using the improved firefly algorithm, so as to determine the installation location and capacity of the power equipment and obtain the offline planning layer configuration scheme. The second construction module is used to construct a Markov decision process model based on the network environment under the offline planning layer configuration scheme. The solution module is used to solve the Markov decision process model using a deep Q-network during online operation. It converts real-time physical operation data into state vectors and inputs them into the deep Q-network to predict the expected governance benefits corresponding to each power quality governance action in the action space. It also selects target control actions, decodes them into physical device control parameters, and sends the physical device control parameters to the corresponding devices for governance.

[0070] This embodiment also provides a computer device applicable to the collaborative governance and optimization configuration of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the method for collaborative governance and optimization configuration of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms as proposed in the above embodiment.

[0071] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the method for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms, as proposed in the above embodiments.

[0072] The storage medium proposed in this embodiment and the method for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0073] From the above description of the implementation methods, those skilled in the art will clearly understand that the present invention can be implemented using software and necessary general-purpose hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0074] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for collaborative governance and optimized allocation of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms, characterized in that: include: Acquire distribution network data and construct a low-voltage AC / DC hybrid distribution network topology model; The comprehensive weighting coefficients of each power quality indicator are determined by combining subjective and objective weighting methods, and the power quality of the topology model is comprehensively scored based on the comprehensive weighting coefficients. An improved firefly algorithm is used to solve a bi-objective optimization function constructed based on equipment investment cost and the comprehensive power quality score, in order to determine the installation location and capacity of power equipment and obtain an offline planning layer configuration scheme. Based on the network environment under the offline planning layer configuration scheme, a Markov decision process model is constructed. During online operation, the Markov decision process model is solved using a deep Q-network. The state vector converted from real-time physical operation data is input into the deep Q-network to predict the expected governance benefits corresponding to each power quality governance action in the action space. The target control action is selected, decoded into physical device control parameters, and the physical device control parameters are sent to the corresponding device for governance.

2. The method for collaborative governance and optimized allocation of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms as described in claim 1, characterized in that, The comprehensive weighting coefficients for determining various power quality indicators by combining subjective and objective weighting methods include: The subjective weights of each power quality indicator are calculated using the analytic hierarchy process (AHP), and the objective weights of each power quality indicator are calculated using the entropy weight method. The subjective weights and objective weights are fused using the least squares method to obtain the final comprehensive weight coefficients for each power quality indicator. The comprehensive power quality score is obtained by weighting and summing the comprehensive weight coefficients of each item with the normalized voltage deviation index, harmonic current content index, and three-phase imbalance index.

3. The method for collaborative governance and optimized allocation of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms as described in claim 2, characterized in that, An improved firefly algorithm is used to solve a bi-objective optimization function constructed based on equipment investment cost and the comprehensive power quality score, in order to determine the installation location and capacity of power equipment and obtain an offline planning layer configuration scheme, including: In the low-voltage AC / DC hybrid distribution network topology model, the governance equipment adopts a dynamic-static combined configuration strategy; the governance equipment includes static reactive power compensation devices and dynamic reactive power compensation devices. The equipment investment cost in the bi-objective optimization function includes the capacity cost function of the static reactive power compensation device and the dynamic reactive power compensation device. The improved firefly algorithm is optimized by using the position and attraction update formula to minimize the comprehensive power quality evaluation index corresponding to the calculated comprehensive power quality score.

4. The method for collaborative governance and optimized allocation of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms as described in claim 3, characterized in that, Based on the network environment under the aforementioned offline planning layer configuration scheme, a Markov decision process model is constructed, including: The Markov decision process model includes a state space, an action space, and a reward function. The reward function includes: power quality related limit violation penalty items, which specifically include AC side voltage limit violation penalty items, DC side voltage limit violation penalty items, and AC side three-phase unbalance limit violation penalty items. The reward function also includes: a penalty for photovoltaic curtailment and a penalty for frequent energy storage operations; Each penalty item has a corresponding weighting coefficient set in the reward function, and the weighting coefficient is preset according to the governance priority of different power quality problems.

5. The method for collaborative governance and optimized allocation of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms as described in claim 4, characterized in that, The Markov decision process model is solved using a deep Q-network, including: By utilizing measurement equipment deployed at key nodes of the distribution network, real-time data are collected on voltage amplitude, branch power flow, and real-time status of controllable flexible equipment, including AC and DC side nodes. The collected data are normalized and then concatenated into a current state vector in the form of a long vector according to the state space format defined in the Markov decision process model. The current state vector is input into the input layer of the deep Q-network.

6. The method for collaborative governance and optimized allocation of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms as described in claim 5, characterized in that, The state vector, converted from real-time physical operation data, is input into the deep Q-network to predict the expected governance benefits corresponding to various power quality governance actions in the action space, and the target control actions are selected and decoded into physical equipment control parameters, including: The deep Q-network is used to predict and output the expected governance benefits corresponding to all control actions in the action space. A greedy strategy is adopted to balance the relationship between exploration and utilization. An action is randomly selected in the action space with a first preset probability for exploration, and the control action with the greatest expected governance benefit is selected as the target control action with a second preset probability. The physical device control parameters formed after the target control action is decoded include: the active and reactive power transmission power setting values ​​of the voltage source converter, the reactive current reference value of the static var generator, and the discrete switching signals of the static capacitor bank.

7. The method for collaborative governance and optimized allocation of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms as described in claim 6, characterized in that, Sending the physical device control parameters to the corresponding device for management includes: In response to the real-time monitoring of the AC side voltage exceeding the limit, the voltage source converter is controlled to backfeed active power on both the AC and DC sides, and the static var generator is controlled to output capacitive reactive current, so as to adjust the node voltage to the qualified range. In response to real-time monitoring of excessive three-phase imbalance or harmonic distortion, the static var generator or voltage source converter is controlled to output asymmetrical compensation current to achieve phase-to-phase load balance, and a reverse harmonic current is emitted to cancel the harmonics generated by the nonlinear load. After each device executes the corresponding physical device control parameters, the distribution network environment status at the next moment is obtained and the reward function is updated. The sample consisting of the current state, the target control action, the calculated reward value, and the state at the next moment is stored in the experience replay pool. Batch samples are periodically extracted and the weight parameters of the deep Q network are iteratively updated online using the gradient descent method.

8. A low-voltage AC / DC hybrid distribution network power quality collaborative governance and optimization configuration system based on improved intelligent algorithms, using the method described in any one of claims 1-7, characterized in that, include: The first construction module is used to acquire distribution network data and construct a low-voltage AC / DC hybrid distribution network topology model. The scoring module is used to determine the comprehensive weight coefficients of various power quality indicators by combining subjective and objective weighting methods, and to perform a comprehensive power quality score on the topology model based on the comprehensive weight coefficients. The optimization module is used to solve the bi-objective optimization function constructed by the equipment investment cost and the comprehensive power quality score using the improved firefly algorithm, so as to determine the installation location and capacity of the power equipment and obtain the offline planning layer configuration scheme. The second construction module is used to construct a Markov decision process model based on the network environment under the offline planning layer configuration scheme. The solution module is used to solve the Markov decision process model using a deep Q-network during online operation. It converts real-time physical operation data into state vectors and inputs them into the deep Q-network to predict the expected governance benefits corresponding to each power quality governance action in the action space. It also selects target control actions, decodes them into physical device control parameters, and sends the physical device control parameters to the corresponding devices for governance.

9. A computer device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the method for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores computer-executable instructions, which, when executed by a processor, implement the steps of the method for collaborative governance and optimized configuration of power quality in low-voltage AC / DC hybrid distribution networks based on improved intelligent algorithms as described in any one of claims 1 to 7.

Images