Quantum deep q-learning reconfiguration method and system for new energy access distribution network

Abstract

This invention belongs to the field of power grid optimization technology, and particularly relates to a quantum deep Q-learning reconstruction method and system for distribution networks with renewable energy integration. The method includes: S1: constructing a distribution network model based on a neural network; S2: constructing a quantum deep Q-learning reconstruction method; S3: training the quantum deep Q-learning algorithm. By precisely controlling the switching operations in the distribution network, this invention can maximize the utilization rate of renewable energy and reduce unnecessary energy losses while ensuring stable grid operation. Utilizing the combination of quantum computing and deep learning, this invention enables the distribution network to automatically optimize its reconstruction strategy in the face of renewable energy integration and load changes, enhancing the adaptability and intelligent management level of the distribution network.

Description

Technical Field

[0001] This invention belongs to the field of power grid optimization technology, and in particular relates to a quantum deep Q-learning reconstruction method and system for new energy access to distribution networks. Background Technology

[0002] In the field of distribution network management and optimization, traditional optimization methods typically rely on precise mathematical models to describe the operating state of the distribution network and the integration of new energy sources. While this approach may be effective for simple or static systems, it becomes extremely difficult, if not impossible, to construct a model that is both accurate and comprehensive when dealing with real-world distribution networks, especially considering the dynamic and uncertain nature of new energy integration. This is because the power generation of new energy sources is affected by various unpredictable factors, such as the direct impact of weather conditions on solar and wind power. These factors make the operating environment of the distribution network highly nonlinear and dynamically changing, making it difficult to accurately capture using traditional static models.

[0003] Furthermore, even if a relatively accurate model can be established, traditional optimization algorithms face problems such as high computational cost and slow solution speed when solving such complex problems, making it difficult to meet the requirements of real-time or near-real-time optimization. This not only affects the operating efficiency of the distribution network but also limits the flexibility and efficiency of new energy integration, thus hindering the wider application of renewable energy in the power grid.

[0004] In contrast, reinforcement learning, as a model-free approach, offers a completely new solution. It does not require a pre-established precise mathematical model of the distribution network; instead, it automatically discovers the optimal strategy through interactive learning with the environment. This method is particularly suitable for handling complex and dynamically changing systems like distribution networks. Reinforcement learning can learn directly from the operational data of the distribution network, adaptively adjusting strategies to cope with changes brought about by the integration of new energy sources, thereby achieving more efficient and flexible distribution network optimization and management.

[0005] However, despite the aforementioned advantages of reinforcement learning, how to effectively utilize quantum computing to accelerate the learning process and improve learning efficiency and policy quality in practical applications, especially with the increasing maturity of quantum computing, has become a worthwhile challenge. Quantum deep Q-learning, as an emerging technology combining the advantages of quantum computing and deep learning, offers a potential solution to the computational bottleneck of traditional reinforcement learning when dealing with high-dimensional and complex systems. By leveraging the parallelism and efficiency of quantum computing, quantum deep Q-learning can maintain or even enhance the quality of learning policies while improving computational speed, which is of great significance for achieving efficient and intelligent power distribution network management.

[0006] Based on the above analysis, the existing technologies suffer from the following problems and shortcomings: traditional optimization methods face challenges such as large computational load and slow solution speed, making it difficult to meet the needs of real-time optimization. These limitations restrict the flexibility and efficiency of new energy integration and hinder the widespread application of renewable energy in the power grid. While reinforcement learning provides a model-free solution, its network structure is typically large, requiring significant storage memory, which places higher demands on computing devices in complex real-world scenarios. Summary of the Invention

[0007] To address the problems existing in current technologies, this invention provides a quantum deep Q-learning reconfiguration method and system for distribution networks with renewable energy integration. It aims to achieve optimized reconfiguration strategies for distribution networks with renewable energy integration through quantum deep Q-learning (QDQL) technology. With the large-scale integration of renewable energy, distribution networks are facing significant changes in structure and operation. This invention, by constructing an advanced learning model, quickly and effectively optimizes the distribution network structure, reduces power losses, and improves overall system efficiency.

[0008] This invention is implemented as follows: a quantum deep Q-learning reconstruction method for new energy access to distribution networks, comprising:

[0009] S1: Construct a distribution network model based on neural networks;

[0010] S2: Constructing a quantum deep Q-learning reconstruction method;

[0011] S3: Training quantum deep Q-learning.

[0012] Furthermore, S1 specifically includes: constructing a mathematical model of the distribution network, which includes detailed descriptions of distribution network nodes, sectionalizing switches, tie switches, and renewable energy access points, mainly including node power balance constraints:

[0013]

[0014] Where n represents the total number of nodes in the distribution network system, P gen,i This represents the power output of the generator at the i-th node; if there is no generator at that node, it is 0. load,i It is the load power at the i-th node; P renew_energy,i P is the power generation capacity of the new energy source at the i-th node; loss,lines This represents the power loss on different transmission lines.

[0015] Furthermore, S1 also includes:

[0016] Voltage constraint:

[0017] V i,min ≤V i |≤V i,max

[0018] Among them, V i,min and V i,max V represents the minimum and maximum voltage limits for node i, respectively. i Let it be represented as the voltage at node i.

[0019] Current constraint:

[0020] |I ij |≤I ij,max

[0021] Among them, I ij,max This indicates the maximum current carrying capacity of the line between nodes i and j.

[0022] Ohm's Law constraint:

[0023] ΔV ij =I ij ·Z ij

[0024]

[0025] Among them, I ij Z is the current in the line between nodes i and j. ij R is the impedance of the line between nodes i and j. ij It is the resistance portion of the line between nodes i and j, ΔV ij P is the voltage drop of the line between nodes i and j. loss,ij It represents the power loss of the line between nodes i and j.

[0026] Power generation constraints:

[0027] P renew_energy,min,i ≤P renew_energy,i ≤P renew_energy,max,i

[0028] P gen,min,i ≤P gen,i ≤P gen,max,i

[0029] Among them, P renew_energy,min,i and P renew_energy,max,i P represents the minimum and maximum power generation limits for the i-th new energy power generation unit, respectively; gen,min,i and P gen,max,i These represent the minimum and maximum power generation limits for the i-th traditional power generation unit, respectively.

[0030] Integrating the above constraints into the optimal power flow problem, the overall mathematical model includes an objective function, power balance constraints, voltage constraints, current constraints, generation constraints, and Ohm's law constraints. The objective function is the total power loss.

[0031] Minimize∑P loss,lines

[0032] Based on the model established above, data on different switching actions under different loads are collected to construct the optimal power flow dataset for the model.

[0033] Furthermore, a fully connected neural network is constructed to learn the dataset. The training process can be accomplished by minimizing the difference between the predicted output and the actual distribution network state.

[0034] L=∑(P pred -∑P loss,lines ) 2

[0035] Where L is the loss function, P pred It is the sum of power loss in the distribution network predicted by the neural network (i.e., the objective function of the model just constructed).

[0036] Furthermore, S2 specifically includes: after the distribution network model is learned, the neural network model can be used as a quantum deep Q-learning model. Next, an encoder is used to encode the state of the distribution network, thereby compressing and encoding the state information output by the environment into a data size acceptable to qubits; then the encoded state is input into the PQC, the PQC processes these quantum states and outputs the quantum measurement results of the action values; the output of the quantum deep Q-learning model, that is, the switching operation strategy of the distribution network, is then decoded by a traditional neural network and mapped to the specific switching action of the distribution network.

[0037] Furthermore, S3 specifically includes: after completing the encoding of the distribution network state and establishing the parameterized quantum circuit, the next step is to train the model using the quantum deep Q-learning framework. This training process includes the following key parts:

[0038] (1) Initialize parameters: Initialize the parameters θ in PQC, as well as the weights and biases of the encoder and decoder networks;

[0039] (2) Environment interaction: Encode the current distribution network state S as S encoded PQC according to S encoded The quantum measurement result of the output action value is then converted into a specific switching operation A by the decoder. action Execute A action Then, the environment returns a new state S' and a reward R, while simultaneously storing the interaction array (S, A) action (R,S') are stored in the playback pool;

[0040] (3) Update and optimization: Based on the data in the replay pool, the mean squared error loss function is used to measure the difference between the predicted Q value and the target Q value, and the time difference method is used to calculate the loss function.

[0041] Another objective of this invention is to provide a quantum deep Q-learning reconstruction system for new energy-integrated distribution networks, which implements the aforementioned quantum deep Q-learning reconstruction method for new energy-integrated distribution networks, comprising:

[0042] Distribution network model building module: used to build distribution network models based on neural networks;

[0043] Quantum Deep Q-Learning Reconstruction Method Building Module: Used to build quantum deep Q-learning reconstruction methods;

[0044] Quantum Deep Q-Learning Training Module: Used for training quantum deep Q-learning.

[0045] Another object of the present invention is to provide a computer device, the computer device including a memory and a processor, the memory storing a computer program, and when the computer program is executed by the processor, causing the processor to perform the steps of the quantum deep Q-learning reconstruction method for new energy access to the distribution network.

[0046] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the quantum deep Q-learning reconstruction method for new energy access to the distribution network.

[0047] Another objective of this invention is to provide an information data processing terminal for implementing the aforementioned quantum deep Q-learning reconstruction system for new energy access to the power distribution network.

[0048] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows:

[0049] First, regarding the technical problems existing in the aforementioned prior art, the creative technical effects resulting from solving these problems are described in detail below:

[0050] 1. By precisely controlling the switching operations in the power distribution network, this invention can maximize the utilization rate of new energy sources and reduce unnecessary energy losses while ensuring the stable operation of the power grid;

[0051] 2. By combining quantum computing and deep learning, this invention enables the distribution network to automatically optimize its reconfiguration strategy when facing the access of new energy sources and load changes, thereby enhancing the adaptability and intelligent management level of the distribution network.

[0052] 3. Considering the characteristics of current quantum computers as noisy medium-scale quantum (NISQ) devices, this invention ensures the application of quantum computing in power distribution network management even when the number of qubits is limited through efficient encoding and decoding strategies, providing a new approach for the application of quantum technology in practical power grid systems.

[0053] Secondly, this invention proposes an innovative solution combining quantum computing and deep learning technologies, specifically designed for the integration of new energy sources into distribution networks, aiming to improve the operational efficiency and intelligent management level of distribution networks. Through a meticulously designed quantum deep Q-learning (QDQL) framework, this invention can efficiently optimize and reconstruct distribution networks. The specific technical effects and advantages are summarized below:

[0054] High-efficiency energy utilization and reduced losses: This technical solution can precisely control the switching operations in the distribution network, optimizing the access and utilization of new energy sources such as solar and wind power. Through intelligent management, it maximizes the utilization rate of new energy sources while reducing energy losses during transmission and conversion, achieving a more environmentally friendly and economical power grid operation mode.

[0055] Adaptive Optimization and Reconfiguration Strategy: Combining the high-efficiency processing power of quantum computing with the intelligent decision-making capabilities of deep learning, this invention can respond in real time to changes in the distribution network state and automatically optimize and adjust the reconfiguration strategy. This not only enhances the adaptability of the distribution network to new energy access and load fluctuations, but also improves the stability and reliability of the power grid.

[0056] Overcoming the limitations of quantum computing applications: Considering the physical limitations and noise characteristics of current quantum computers, this invention overcomes the challenge of the limited number of qubits by designing an efficient state encoding and decoding strategy. This enables the application of quantum deep Q-learning in power distribution network management even under current quantum computing technology conditions, demonstrating the potential for widespread application of quantum technology in the energy sector.

[0057] Improving the level of intelligent power grid management: By realizing intelligent optimization and management of the distribution network, this invention provides strong technical support for power grid operation and maintenance. This includes, but is not limited to, functions such as real-time monitoring of power grid status, prediction and response to load changes, and automatic adjustment of power grid structure, thereby improving the operating efficiency and service quality of the power grid.

[0058] Promoting the integration of new and traditional energy sources: This invention not only optimizes the access and utilization of new energy sources but also considers their synergy with traditional energy generation methods. Through intelligent strategy adjustments, it ensures the efficient operation of the power grid and the stability of energy supply. This provides important technical support for achieving the green transformation of the energy structure and sustainable development goals.

[0059] This invention, through its innovative technical solution, ensures the stable operation of the power grid while improving the utilization rate of new energy sources and the level of intelligent management of the power grid. It demonstrates the prospects of quantum computing technology in the energy field and provides new ideas and technical paths for the future development of the power grid.

[0060] Third, the expected benefits and commercial value of the technical solution of this invention after transformation are as follows:

[0061] This invention provides an optimized reconfiguration solution for integrating renewable energy into power distribution networks by efficiently integrating quantum computing and deep learning technologies. This not only improves the energy efficiency and operational efficiency of the power distribution network but also significantly reduces energy loss, thereby lowering operating costs. From a commercial perspective, this optimization can promote the widespread use of renewable energy, support the development of the green energy industry, and possess enormous market potential and commercial value. Furthermore, with the advancement of quantum technology, this invention is expected to lead a new round of technological innovation and industrial upgrading, creating new growth points for enterprises and investors.

[0062] The technical solution of this invention fills a technological gap in the industry both domestically and internationally: combining quantum computing with deep learning for the optimization and reconfiguration of new energy access to the power distribution network is a new field. Currently, the application of quantum computing in energy management and optimization is still in its early stages. This invention not only realizes the specific application of this concept, but also successfully demonstrates the application potential of quantum technology in practical energy systems, filling a technological gap in this field both domestically and internationally.

[0063] The technical solution of this invention solves a long-standing but unresolved technical problem: the efficient integration of new energy sources and the intelligent optimization management of power distribution networks have always been challenging issues in the energy sector. Especially with the large-scale integration of new energy sources, how to efficiently and stably integrate these energy sources has become a significant challenge. This invention provides an effective solution that not only improves the utilization rate of new energy sources but also ensures the stability and reliability of the power grid, thus resolving a long-standing technical problem.

[0064] The technical solution of this invention overcomes technological bias: before quantum computing was applied to practical industrial and energy management fields, there was a view that quantum computing technology was not yet mature and difficult to apply to the optimization and management of complex systems in the short term. This invention, through the design of efficient encoding and decoding strategies, successfully applies quantum deep Q-learning to the optimization and reconfiguration of power distribution networks. This not only proves the feasibility of quantum technology in the energy field but also overcomes technological biases regarding the scope of quantum computing applications, opening up new paths for the application of quantum technology in industrial and energy sectors.

[0065] Fourth, the quantum deep Q-learning reconstruction method for new energy grid integration proposed in this invention has achieved significant technological progress in solving existing technical problems. The following are the technical problems solved by this invention and the significant technological advancements achieved in industrial applications:

[0066] Technical problems to be solved:

[0067] 1. Handling the Complexity of Distribution Network Models: Traditional distribution network management often relies on simple mathematical models and rules, making it difficult to handle the complexity and uncertainty brought about by the integration of new energy sources. This invention constructs a distribution network model based on neural networks, which can more accurately describe the dynamic characteristics of the distribution network and the impact of new energy integration.

[0068] 2. Optimization of High-Dimensional State Space: With the expansion of power distribution network scale and the increase in new energy access points, the dimensionality of the state space grows dramatically, making it difficult for traditional optimization methods to handle. This invention utilizes the parallelism and exponential state space processing capabilities of quantum computing to achieve optimization of high-dimensional state space through quantum deep Q-learning.

[0069] 3. Real-time Decision Making and Adaptive Adjustment: After new energy sources are integrated into the distribution network, switching operations need to be adjusted in real time to optimize network operation. This invention utilizes quantum deep Q-learning to achieve real-time perception of the distribution network status and adaptive adjustment of switching operation strategies.

[0070] Significant technological advancements achieved in industrial applications:

[0071] 1. Improve energy efficiency: By precisely controlling the switching operations of the distribution network, optimize the access and utilization of new energy sources, reduce energy loss, and improve energy efficiency.

[0072] 2. Enhanced distribution network stability: This invention can sense the status of the distribution network in real time and adjust the switching operation according to the status, thereby effectively avoiding problems such as overload and voltage fluctuation, and enhancing the stability of the distribution network.

[0073] 3. Reduce operation and maintenance costs: Through automated and intelligent distribution network management, reduce manual intervention and operation and maintenance costs, and improve the operating efficiency and economy of the distribution network.

[0074] 4. Promoting the application of quantum computing in the industrial field: This invention combines quantum computing with power distribution network management, providing new ideas and directions for the application of quantum computing in the industrial field, and promoting the practical application and industrialization of quantum computing technology.

[0075] This invention constructs a distribution network model based on neural networks and combines it with a quantum deep Q-learning reconstruction method to achieve optimized management of new energy access to the distribution network. It solves the problems existing in the prior art and has achieved significant technological progress in industrial applications. Attached Figure Description

[0076] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments of the present invention 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.

[0077] Figure 1 This is a schematic diagram of a quantum deep Q-learning reconstruction method for new energy access to the distribution network provided in an embodiment of the present invention;

[0078] Figure 2 This is a diagram of the quantum deep Q-learning structure provided in an embodiment of the present invention;

[0079] Figure 3 This is a structural diagram of a quantum deep Q-learning reconstruction system for new energy access to the distribution network provided in an embodiment of the present invention;

[0080] Figure 4 This is a schematic diagram of the IEEE 33-node power distribution network structure provided in an embodiment of the present invention;

[0081] Figure 5 This is a schematic diagram illustrating the training effects of traditional neural networks and quantum deep Q-learning provided in an embodiment of the present invention. Detailed Implementation

[0082] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0083] To address the problems existing in the prior art, this invention provides a quantum deep Q-learning reconstruction method and system for new energy access to the power distribution network. The invention will be described in detail below with reference to the accompanying drawings.

[0084] like Figure 1 As shown, the quantum deep Q-learning reconstruction method for new energy access to distribution networks provided in this embodiment of the invention includes:

[0085] S1: Construct a distribution network model based on neural networks;

[0086] S2: Constructing a quantum deep Q-learning reconstruction method;

[0087] S3: Training quantum deep Q-learning.

[0088] S1 specifically includes: constructing a mathematical model of the distribution network, which includes detailed descriptions of distribution network nodes, sectionalizing switches, tie switches, and renewable energy access points, mainly including node power balance constraints:

[0089]

[0090] Where n represents the total number of nodes in the distribution network system, P gen,i This represents the power output of the generator at the i-th node; if there is no generator at that node, it is 0. load,i It is the load power at the i-th node; P renew_energy,i P is the power generation capacity of the new energy source at the i-th node; loss,lines This refers to the power loss on different transmission lines.

[0091] S1 also includes:

[0092] Voltage constraint:

[0093] V i,min ≤|V i |≤V i,max

[0094] Among them, V i,min and V i,max These represent the minimum and maximum voltage limits for node i, respectively.

[0095] Current constraint:

[0096] |I ij |≤I ij,max

[0097] Among them, I ij,max This indicates the maximum current carrying capacity of the line.

[0098] Ohm's Law constraint:

[0099] ΔV ij =I ij ·Z ij

[0100]

[0101] Among them, I ij Z is the current in the line between nodes i and j. ij R is the impedance of the line between nodes i and j. ij It is the resistance portion of the line between nodes i and j, ΔV ij P is the voltage drop of the line between nodes i and j. loss,ij It represents the power loss of the line between nodes i and j.

[0102] Power generation constraints:

[0103] P renew_energy,min,i ≤P renew_energy,i ≤P renew_energy,max,i

[0104] P gen,min,i ≤P gen,i ≤P gen,max,i

[0105] Among them, P renew_energy,min,i and P renew_energy,max,i P represents the minimum and maximum power generation limits for the i-th new energy power generation unit, respectively; gen,min,i and P gen,max,i These represent the minimum and maximum power generation limits for the i-th traditional power generation unit, respectively.

[0106] Integrating the above constraints into the optimal power flow problem, the overall mathematical model includes an objective function, power balance constraints, voltage constraints, current constraints, generation constraints, and Ohm's law constraints. The objective function is the total power loss.

[0107] Minimize∑P loss,lines

[0108] Based on the model established above, data on different switching actions under different loads are collected to construct the optimal power flow dataset for the model.

[0109] A fully connected neural network is constructed to learn the dataset. The training process can be accomplished by minimizing the difference between the predicted output and the actual distribution network state.

[0110] L=∑(P pred -∑P loss,lines ) 2

[0111] Where L is the loss function, P pred It is the sum of power loss in the distribution network predicted by the neural network (i.e., the objective function of the model just constructed).

[0112] S2 specifically includes: After the distribution network model is learned, the neural network model can be used as a quantum deep Q-learning model. Next, the encoder is used to encode the state of the distribution network, thereby compressing and encoding the state information output by the environment into a data size acceptable to qubits; then the encoded state is input into the PQC, the PQC processes these quantum states and outputs the quantum measurement results of the action values; the output of the quantum deep Q-learning model, that is, the switching operation strategy of the distribution network, is then decoded by a traditional neural network and mapped to the specific switching action of the distribution network.

[0113] S3 specifically includes: After completing the encoding of the distribution network state and establishing the parameterized quantum circuit, the next step is to train the model using the quantum deep Q-learning framework. This training process includes the following key parts:

[0114] (1) Initialize parameters: Initialize the parameters θ in PQC, as well as the weights and biases of the encoder and decoder networks;

[0115] (2) Environment interaction: Encode the current distribution network state S as S encoded PQC according to S encoded The quantum measurement result of the output action value is then converted into a specific switching operation A by the decoder. action Execute A action Then, the environment returns a new state S' and a reward R, while simultaneously storing the interaction array (S, A) action (R,S') are stored in the playback pool;

[0116] (3) Update and optimization: Based on the data in the replay pool, the mean squared error loss function is used to measure the difference between the predicted Q value and the target Q value, and the time difference method is used to calculate the loss function.

[0117] like Figure 3 As shown in the figure, the quantum deep Q-learning reconstruction system for new energy access to the distribution network provided in this embodiment of the invention includes:

[0118] Distribution network model building module: used to build distribution network models based on neural networks;

[0119] Quantum Deep Q-Learning Reconstruction Method Building Module: Used to build quantum deep Q-learning reconstruction methods;

[0120] Quantum Deep Q-Learning Training Module: Used for training quantum deep Q-learning.

[0121] An application embodiment of the present invention provides a computer device, which includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor performs the steps of a quantum deep Q-learning reconstruction method for new energy access to the power distribution network.

[0122] An application embodiment of the present invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor performs the steps of a quantum deep Q-learning reconstruction method for new energy access to the distribution network.

[0123] An application embodiment of the present invention provides an information data processing terminal, which is used to realize a quantum deep Q-learning reconstruction system for new energy access to the power distribution network.

[0124] II. Application Examples. To demonstrate the inventiveness and technical value of the technical solution of this invention, this section provides application examples of the technical solution of the claims on specific products or related technologies.

[0125] To fully verify the inventiveness and technical value of the present invention, this invention uses the IEEE 33-node power distribution network as an example to demonstrate the application and implementation of this technology in specific products and technologies. For example... Figure 4 As shown, the IEEE 33-node distribution network includes three generators and one photovoltaic power station. Each generator has a maximum power output of 0.7 megawatts (MW), and the photovoltaic power station has a maximum power output of 0.2 megawatts (MW). Within this network framework, this invention designs a quantum deep Q-learning (QDQL) network with four qubits and equips the output side with a 4*52 action decoder. This decoder is responsible for converting the output of the quantum deep Q-learning network into specific distribution network switching operation strategies.

[0126] In this embodiment, the input data are the total load demand and total generation power of the distribution network, which serve as a quantitative representation of the network state. The quantum deep Q-learning network utilizes this input information, leveraging the parallel processing of quantum computing and the pattern recognition capabilities of deep learning to optimize the structural reconstruction of the distribution network and minimize losses.

[0127] The present invention employs an efficient encoding strategy to process the quantization representation of power distribution network status, adapting to the reality of a limited number of qubits. This strategy covers key parameters of the power grid status, such as node voltage and line current, enabling accurate reflection of the real-time state of the power grid even with a limited number of qubits.

[0128] Employing a 4-qubit quantum deep Q-learning network ensures the algorithm operates with limited quantum resources while maintaining performance and computational efficiency. The 4*52 action decoder on the output side maps the output of the quantum deep Q-learning network to 52 specific possible distribution network switching actions, thus providing decision support for the optimized operation of the distribution network.

[0129] Through the application of this embodiment, the present invention has demonstrated its effectiveness in practical power grid systems, including improving the energy utilization efficiency of the power grid, optimizing the power grid operation structure, and realizing intelligent management. The present invention creatively integrates quantum computing and deep learning technologies into existing distribution network technologies, providing an innovative optimization strategy for the integration of new energy sources into the distribution network, fully demonstrating its commercial feasibility and technological value.

[0130] The embodiments of this invention have achieved significant positive results during the research and development and testing process, particularly demonstrating excellent performance in the task of distribution network optimization and reconfiguration. For example... Figure 5As shown, by comparing the training effects of traditional neural networks and quantum deep Q-learning (QDQL), the following conclusions are drawn:

[0131] In traditional neural networks, the system performance is represented by the red and blue lines in the graph when the number of neurons in the hidden layer is set to 4 and 9, respectively. Figure 5 As can be seen, quantum deep Q-learning exhibits better performance and faster convergence speed throughout the training process, and its robustness and optimization effect are superior to traditional neural networks. In the final test, the optimal action selection rate of quantum deep Q-learning reached 100%, significantly higher than the 55.21% and 71.88% of traditional neural networks. This comparison not only highlights the significant advantage of quantum deep Q-learning in action decision accuracy, but also verifies its practical application value in power grid optimization.

[0132] These experimental results demonstrate that the present invention offers a significant performance improvement over existing technologies in distribution network optimization and reconfiguration. It not only improves the operational efficiency of the power grid but also enhances the management capabilities for the integration of new energy sources. By combining quantum computing and deep learning, the embodiments of the present invention overcome the limitations of traditional algorithms in handling complex power grid structures and dynamic changes, providing strong technical support for the intelligent management and sustainable development of distribution networks.

[0133] It should be noted that embodiments of the present invention can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented using hardware circuitry such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., or using software executed by various types of processors, or using a combination of the above-described hardware circuitry and software, such as firmware.

[0134] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A quantum deep Q-learning reconstruction method for new energy access to distribution networks, characterized in that, include: S1: Construct a distribution network model based on neural networks; S2: Constructing a quantum deep Q-learning reconstruction method; S3: Training quantum deep Q-learning; S3 specifically includes: After completing the encoding of the distribution network state and establishing the parameterized quantum circuit, the next step is to train the model using the quantum deep Q-learning framework. This training process includes the following key parts: (1) Initialize parameters: Initialize the parameters in PQC , as well as the weights and biases of the encoder and decoder networks; (2) Environment interaction: The current distribution network status Encoded as PQC according to The quantum measurement result of the output action value is then converted into a specific switching operation by the decoder. ,implement Afterwards, the environment returns to a new state. and rewards At the same time, the interactive array Store in the playback pool; (3) Update and optimization: Based on the data in the replay pool, the mean squared error loss function is used to measure the difference between the predicted Q value and the target Q value, and the time difference method is used to calculate the loss function; S1 specifically includes: constructing a mathematical model of the distribution network, which includes detailed descriptions of distribution network nodes, sectionalizing switches, tie switches, and renewable energy access points, mainly including node power balance constraints: ， in, This means the distribution network system has a total of n nodes. It is the power output of the generator at the i-th node; if there is no generator at that node, it is 0. It is the load power at the i-th node; It represents the power generation capacity of the new energy source at the i-th node; This refers to the power loss on different transmission lines; S1 also includes: Voltage constraint: ， in, and These represent the minimum and maximum voltage limits for node i, respectively; Current constraint: ， in, This indicates the maximum current carrying capacity of the line; Ohm's Law constraint: ， in, It is the current in the line between nodes i and j. It is the impedance of the line between nodes i and j. It is the resistance portion of the line between nodes i and j. It is the voltage drop of the line between nodes i and j. It is the power loss of the line between nodes i and j; Power generation constraints: ， in, and These represent the minimum and maximum power generation limits for the i-th new energy power generation unit, respectively. and These represent the minimum and maximum power generation limits for the i-th traditional power generation unit, respectively. Integrating the above constraints into the optimal power flow problem, the overall mathematical model includes an objective function, power balance constraints, voltage constraints, current constraints, generation constraints, and Ohm's law constraints. The objective function is the total power loss. ， Based on the model established above, data on different switching actions under different loads are collected to construct the optimal power flow dataset for the model.

2. The quantum deep Q-learning reconstruction method for new energy access to distribution networks as described in claim 1, characterized in that, A fully connected neural network is constructed to learn the dataset. The training process can be accomplished by minimizing the difference between the predicted output and the actual distribution network state. ， in, It is a loss function. It is the sum of power loss in the distribution network predicted by the neural network.

3. The quantum deep Q-learning reconstruction method for new energy access to distribution networks as described in claim 1, characterized in that, S2 specifically includes: After the distribution network model is learned, the neural network model can be used as a quantum deep Q-learning model. Next, the encoder is used to encode the state of the distribution network, thereby compressing and encoding the state information output by the environment into a data size acceptable to qubits; then the encoded state is input into the PQC, the PQC processes these quantum states and outputs the quantum measurement results of the action values; the output of the quantum deep Q-learning model, that is, the switching operation strategy of the distribution network, is then decoded by a traditional neural network and mapped to the specific switching action of the distribution network.

4. A quantum deep Q-learning reconstruction system for new energy-integrated distribution networks, implementing the quantum deep Q-learning reconstruction method for new energy-integrated distribution networks as described in any one of claims 1 to 3, characterized in that, include: Distribution network model building module: used to build distribution network models based on neural networks; Quantum Deep Q-Learning Reconstruction Method Building Module: Used to build quantum deep Q-learning reconstruction methods; Quantum Deep Q-Learning Training Module: Used for training quantum deep Q-learning.

5. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the quantum deep Q-learning reconstruction method for new energy access to the distribution network as described in any one of claims 1 to 3.

6. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor performs the steps of the quantum deep Q-learning reconstruction method for new energy access to distribution networks as described in any one of claims 1 to 3.

7. An information data processing terminal, the information data processing terminal being used to implement the quantum deep Q-learning reconstruction system for new energy access to the distribution network as described in claim 4.

Images