Strong and weak current line intelligent layout and interference prediction analysis method and system

Abstract

The invention relates to the technical field of electrical engineering and artificial intelligence, and discloses a strong and weak current line intelligent layout and interference prediction analysis method and system.The strong and weak current line intelligent layout and interference prediction analysis method comprises the steps that a state representation and interference propagation model of electronic system layout is constructed; uncertainty quantification is carried out based on Bayesian reasoning and a particle filtering method; solving a Pareto optimal layout scheme set of the strong and weak current line by applying an improved multi-objective optimization algorithm; realizing robust reinforcement learning strategy optimization based on the distributed deep Q network; performing decision support through a preference learning model; according to the method, the problems of difficulty in multi-target balance, weak anti-interference capability and imperfect decision support in the traditional wiring technology are solved, the line layout design quality and efficiency are improved, and the electromagnetic interference risk is reduced.

Description

Technical Field

[0001] The present invention relates to the field of electrical engineering and artificial intelligence technology, and more specifically, to a method and system for intelligent layout and interference prediction analysis of strong and weak current lines. Background Art

[0002] As the complexity of modern electrical systems continues to increase, high-voltage and low-voltage lines must coexist within confined spaces, leading to increasingly prominent electromagnetic interference issues. However, existing technologies suffer from the following key shortcomings: traditional single-objective optimization methods struggle to balance multiple, conflicting objectives, such as interference suppression, space utilization, and cost control; most methods fail to account for environmental uncertainties, such as load variations and material property deviations, resulting in poor performance of theoretically optimal layouts in practical applications; existing methods suffer from low computational efficiency, making it difficult to handle large-scale layout problems within a reasonable timeframe; and there is a lack of effective mechanisms for systematically integrating expert experience into the optimization process, resulting in limited human-machine collaboration.

[0003] Therefore, a method for intelligent layout and interference prediction analysis of strong and weak current lines is needed that can simultaneously consider multi-objective balance, environmental uncertainty, computational efficiency and expert knowledge fusion to improve the quality, efficiency and robustness of line layout. Summary of the Invention

[0004] The present invention provides a method and system for intelligent layout and interference prediction analysis of strong and weak current lines, which solves the technical problems in related technologies that are difficult to simultaneously take into account multi-objective balance, environmental uncertainty, computational efficiency and expert knowledge integration.

[0005] The present invention provides a method for intelligent layout and interference prediction analysis of strong and weak current lines, including:

[0006] Construct state representation and interference propagation model of electronic system layout, including: establishing electromagnetic interference propagation model; forming interference impact matrix;

[0007] Using the constructed interference propagation model and interference impact matrix, uncertainty quantification is performed based on Bayesian reasoning and particle filtering methods, including: collecting interference data under different working conditions; obtaining the probability distribution of major uncertainty factors;

[0008] Based on state representation, interference propagation model and probability distribution of uncertainty factors, an improved multi-objective optimization algorithm is applied to solve the Pareto optimal layout solution set of power and weak current lines, including: setting multiple optimization objectives; generating a set of non-dominated solutions;

[0009] Using the set of non-dominated solutions as training data, robust reinforcement learning strategy optimization is implemented based on a distributed deep Q-network, including: defining the environment state space, action space, and reward function; optimizing the layout decision strategy; and performing robustness evaluation and optimization adjustments on the trained strategy.

[0010] Combining the set of non-dominated solutions and the optimal layout decision strategy, decision support is provided through a preference learning model, including: constructing a preference learning model; evaluating and ranking all non-dominated solutions; and recommending the optimal layout solution.

[0011] In a preferred embodiment, constructing the state representation and interference propagation model of the electronic system layout specifically includes:

[0012] Discretize the layout space into G = {g i，j，k}, where G represents the grid set of the entire layout space, i, j, and k represent the X-axis, Y-axis, and Z-axis coordinates of the space, respectively. i，j，k A grid cell representing a specific location in three-dimensional space;

[0013] For any two circuit elements e a and e b Potential interference between I(e a , e b ), and establish a calculation model based on electromagnetic field theory:

[0014]

[0015] Among them, I(e a , e b ) represents the element e a to element e b The electromagnetic interference intensity, P a Indicates the interference source power, S b Indicates the sensitivity of the interfered component, d ab Represents the distance between two elements, f(θ ab ) represents the direction factor, g(ω a ,ω b ) represents the frequency correlation function, α is the correction coefficient, and α is the comprehensive correction coefficient;

[0016] The calculation results are organized into the interference influence matrix M n×n , where n is the total number of circuit elements in the system.

[0017] In a preferred embodiment, the uncertainty quantification based on Bayesian reasoning and particle filtering method specifically includes:

[0018] Install multiple electromagnetic interference monitoring sensors {s1, s2, ..., s m}, collect interference data under different load conditions, where s1, s2, s m They represent the 1st, 2nd, and mth electromagnetic interference monitoring sensors deployed in the system, respectively, and m represents the total number of sensors;

[0019] Establish state transition model p(x t |x t-1 ) and the measurement model p(z t |x t ), where x t represents the interference state at time t, x t-1 represents the interference state at time t-1, z t Indicates the corresponding measurement value;

[0020] The particle filter algorithm is used to estimate the posterior probability of the interference distribution:

[0021]

[0022] Among them, p(x t |z 1：t ) indicates that in the observation sequence z 1：t Under the condition, the interference state x at time t t The posterior probability distribution of represents the particle set and the corresponding weight, δ is the Diracdelta function, represents the i-th particle, N represents the total number of particles, represents the sum of all N particles;

[0023] The probability distribution of the main uncertainty factors is obtained through sampling and importance resampling steps.

[0024] In a preferred embodiment, the application of the improved multi-objective optimization algorithm specifically includes:

[0025] Set the optimization objective function set:

[0026] F(x)={f1(x), f2(x),..., f k (x)};

[0027] Among them, F(x) represents the set of objective functions in the multi-objective optimization problem, x represents the decision variable vector of the layout solution, f1(x) represents the interference suppression metric, f2(x) represents the space utilization, and f k (x) represents cost control, and k represents the total number of optimization objectives;

[0028] The improved NSGA-III algorithm is used for multi-objective optimization, including:

[0029] Using adaptive crossover probability p cand mutation probability p m :

[0030]

[0031] Among them, p c represents the crossover probability in the current iteration, p c0 represents the initial value of the crossover probability, p m Represents the mutation probability in the current iteration, p m0 represents the initial value of the mutation probability, G is the current generation, β1 and β2 represent the amplitude coefficients of the crossover probability and mutation probability, γ1 and γ2 represent the attenuation rates of the crossover probability and mutation probability respectively;

[0032] A reference point-based selection mechanism is used to ensure the diversity of the solution set in the target space;

[0033] Generate non-dominated solution set P * , represents the optimal compromise between multiple objectives.

[0034] In a preferred embodiment, the robust reinforcement learning strategy optimization based on the distributed deep Q network specifically includes:

[0035] Define the state space S, action space A and reward function R:

[0036] The state space contains the current layout state and uncertainty estimates;

[0037] The action space is defined as the possible placement and direction of the lines;

[0038] The reward function comprehensively considers interference suppression, space utilization, and cost control;

[0039] Training a distributed deep Q-network:

[0040] Use double Q learning strategy to reduce Q value estimation bias;

[0041] Apply the experience replay mechanism D to store the transformed samples (s t , a t , r t , s t+1 ), where s t represents the state at time t, a t represents the action executed at time t, r t Indicates execution of action a t The immediate reward after t+1 Indicates execution of action a t The new state to which it is transferred;

[0042] Minimize the loss function:

[0043] L(θ)=E (s,a,r,s')～D [(r+γmax a′ Q(s′,a′;θ - )–Q(s,a;0)) 2 ];

[0044] Among them, L(θ) represents the loss function of network training, E (s，a，r，s′)～D represents the expectation sampled from the experience replay buffer D, r represents the immediate reward, γ represents the discount factor, and max a′ Q(s′, a′; θ - ) indicates the use of target network parameters θ - The maximum Q value of the next state s′ is calculated. Q(s, a; θ) represents the Q value of the state-action pair (s, a) calculated using the current network parameters θ, θ and θ - Represent the current network and target network parameters respectively;

[0045] Through Monte Carlo simulation and scenario analysis, the robustness of the strategy in an uncertain environment is evaluated and adjustments and optimizations are made.

[0046] In a preferred embodiment, the decision support by the preference learning model specifically includes:

[0047] Collect expert evaluation dataset D p ={(x i , x j ,y ij )}, where D p represents the expert evaluation dataset, x i 、x j Represents the i-th and j-th layout schemes, y ij Indicates the expert's opinion on solution x i and x j degree of preference;

[0048] An enhanced twin neural network is constructed for the preference learning model, and its loss function is defined as:

[0049]

[0050] Among them, L p (θ) represents the loss function of the preference learning model, Denotes the expert evaluation dataset D p Sum all data samples in, f θ represents the preference learning model, y ij Indicates the expert's opinion on solution x i and x j The preference degree, λ is the regularization parameter, represents the regularization term;

[0051] Use the trained preference model to analyze the non-dominated solution set P * Score all layout options in:

[0052] score(x)=f θ (x)+α·diversity(x);

[0053] Among them, score(x) represents the comprehensive score of layout plan x, f θ (x) represents the utility score of the preference model for layout option x, diversity(x) is the diversity measure, and α is the balance parameter;

[0054] Recommend the optimal layout solution based on the comprehensive score.

[0055] In a preferred embodiment, the importance of different objectives is comprehensively considered through an adaptive weighting mechanism, where the objective weights are dynamically adjusted according to environmental conditions and user needs:

[0056]

[0057] Among them, W i (t) represents the weight of the i-th target at time t, β i (t) is a parameter that changes with environmental conditions, Indicates the exponential transformation of the importance parameter, represents the sum of the exponential transformations of all k objectives, where k represents the total number of optimization objectives.

[0058] In a preferred embodiment, the uncertainty quantification based on the Bayesian reasoning and particle filtering method further includes an uncertainty propagation analysis method for evaluating the propagation characteristics of uncertainty in the system:

[0059] Build an uncertainty propagation graph:

[0060] G u =(V, E, W);

[0061] Among them, G u represents the uncertainty propagation diagram, V represents the system components, E represents the association between components, and W represents the uncertainty propagation intensity;

[0062] The Fokker-Planck equation is used to describe the evolution of uncertainty in the system:

[0063]

[0064] in, represents the rate of change of the probability density function over time, p(x, t) represents the probability density of the system state x at time t, A(x) and B(x) represent the drift and diffusion coefficients respectively, represents the deterministic drift term, represents the random diffusion term.

[0065] In a preferred embodiment, a system for implementing a method for intelligent layout and interference prediction analysis of strong and weak current lines includes:

[0066] State representation and interference modeling module, used to build state representation and interference propagation models of electronic system layout;

[0067] Uncertainty quantification module, used for uncertainty quantification based on Bayesian reasoning and particle filtering methods;

[0068] Multi-objective optimization module, used to apply an improved multi-objective optimization algorithm to solve the Pareto optimal layout solution set for power and weak current lines;

[0069] Reinforcement learning strategy module, used to implement robust reinforcement learning strategy optimization based on distributed deep Q network;

[0070] Preference learning and decision support module, used to provide decision support through preference learning models;

[0071] Human-computer interaction interface for visualizing layout solutions and receiving user feedback.

[0072] In a preferred embodiment, a computer-readable medium having a computer program recorded thereon implements a method for intelligent layout and interference prediction analysis of strong and weak current lines when the computer program is executed by a processor.

[0073] The beneficial effects of the present invention are:

[0074] Multi-objective balancing capability: This method simultaneously considers multiple conflicting objectives, such as interference suppression, space utilization, and cost control, forming a Pareto front representing different trade-offs and providing decision makers with a variety of compromise options. Experimental results have shown that in complex engineering environments, layout solution quality improves by approximately 35% compared to traditional methods.

[0075] Interference robustness: By leveraging Bayesian reasoning and uncertainty quantification techniques, this invention effectively handles the uncertainty of environmental perception and the randomness of noise interference, predicting and analyzing various interference scenarios and generating robust layout solutions. Field tests have shown that the reliability of the layout solution under complex operating conditions has increased by approximately 50%, and its adaptability to actual environments has been enhanced by approximately 40%.

[0076] Improved Optimization Efficiency: By combining deep reinforcement learning with an improved genetic algorithm, this method overcomes the computational bottlenecks of traditional optimization methods and significantly improves layout optimization efficiency. For large-scale layout problems, optimization speed increases by approximately 80%, and convergence speed increases by approximately 125% under complex constraints.

[0077] Enhanced Adaptability: Through partially observable Markov decision process modeling and distributed deep Q-network learning, this system can sense environmental changes and automatically adjust layout strategies to effectively address dynamic factors such as load variations and material aging. In practical applications, the system's adaptability to new scenarios has increased by approximately 40%, and operational stability has improved by approximately 30%.

[0078] Improved Decision Support: Through preference learning and interactive visualization techniques, this invention provides designers with intuitive decision support tools, effectively integrating expert knowledge into the system and rapidly identifying the layout solution that best meets actual needs from among numerous Pareto-optimal solutions. This technology has been shown to reduce solution evaluation time by approximately 70% and the number of optimization iterations by approximately 40%. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 This is a flow chart of a method for intelligent layout and interference prediction analysis of strong and weak current lines of the present invention;

[0080] Figure 2 is a detailed flow chart of the state representation and interference propagation model for constructing an electronic system layout according to the present invention;

[0081] Figure 3 It is a detailed flow chart of uncertainty quantification based on Bayesian reasoning and particle filtering method of the present invention;

[0082] Figure 4 It is a detailed flow chart of the application of the improved multi-objective optimization algorithm of the present invention to solve the Pareto optimal layout solution set of strong and weak power lines;

[0083] Figure 5 This is a detailed flow chart of the present invention's robust reinforcement learning strategy optimization based on a distributed deep Q network;

[0084] Figure 6 It is a detailed flow chart of decision support through preference learning model of the present invention. DETAILED DESCRIPTION

[0085] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that these embodiments are discussed solely to enable those skilled in the art to better understand and implement the subject matter described herein, and that the functions and arrangements of the elements discussed may be varied without departing from the scope of this specification. Various examples may omit, substitute, or add various processes or components as needed. Furthermore, features described in some examples may be combined in other examples.

[0086] At least one embodiment of the present invention discloses a method for intelligent layout and interference prediction analysis of strong and weak power lines, such as Figures 1 to 6 As shown, the following steps are included:

[0087] Step 1: construct a state representation and interference propagation model of the electronic system layout, including: establishing an electromagnetic interference propagation model; forming an interference impact matrix;

[0088] The following sub-steps are included:

[0089] Step 1.1, line system status characterization;

[0090] Discretize the routing space into a 3D grid:

[0091] G grid ={g i，j,k |1≤i≤I max ,1≤j≤J max , 1≤k≤K max};

[0092] Among them, G grid Represents the grid set of the entire layout space, i, j, k represent the X-axis, Y-axis, and Z-axis coordinates of the space respectively, g i，j，k Represents a grid cell at a specific location in three-dimensional space, I max 、J max , K max Represents the maximum number of grids in the X-axis, Y-axis, and Z-axis directions respectively.

[0093] Each grid point g i，j，k The following information may be brought:

[0094] Is it occupied: Boolean value O i，j，k ∈{0, 1};

[0095] Occupancy type: T i，j，k ∈{empty, strong current line, weak current line, obstacle};

[0096] Line attribute: P i，j，k = {current magnitude, frequency, signal type} (if applicable);

[0097] The overall state of the system is expressed as:

[0098] S svstem ={G grid , C connect};

[0099] Among them, S system Represents the overall state of the system, G grid Represents the grid set of the entire layout space, C connect Represents a set of line connectivity constraints, defined as pairs of endpoints that must be connected:

[0100] C connect ={(g start , g end ) l |l=1,2,...,L total};

[0101] Among them, C connect represents the line connection constraint set, g start 、g end They represent the starting and ending grid points of the lth line, l represents the line number, L total The total number of lines that need to be routed.

[0102] Step 1.2, electromagnetic interference model creation;

[0103] Based on physical electromagnetic field theory, an electromagnetic interference propagation model is created for each line.

[0104] For any grid point g i，j，k The surrounding electromagnetic field strength is calculated as follows:

[0105] Calculation of electric field strength:

[0106]

[0107] Magnetic field strength calculation:

[0108]

[0109] in, Represents the grid point g i，j，k The electric field strength vector at total Indicates the total number of all lines in the system, C l represents the path of the lth line, Represents the grid point g i，j，k The position vector of represents the position vector of the infinitesimal element dl′ on line l, dl′ represents the tiny vector line element on the line, ρ(l′) represents the charge density at position l′ on line l, Represents the grid point gi，j，k The magnetic field strength vector at current，l represents the current in line l, ε0 is the vacuum dielectric constant, μ0 is the vacuum permeability, and × represents the vector cross product operation.

[0110] Step 1.3, multi-objective optimization function definition;

[0111] Define the multi-objective optimization function for circuit layout:

[0112] F obj (x layout )=(f1(x layout ),f2(x layout ),f3(x layout ));

[0113] Among them, F obj (x layout ) represents the multi-objective optimization function of the line layout, x layout Represents a layout plan, including the path information of all lines. f1, f2, and f3 represent the optimization sub-objective functions of three different dimensions: electromagnetic interference suppression, space utilization, and cost control. The sub-objective functions are defined as follows:

[0114] Interference suppression goals:

[0115]

[0116] Space utilization goals:

[0117]

[0118] Cost control objectives:

[0119]

[0120] Among them, f1(x layout ) represents the interference suppression target, f2(x layout ) represents the space utilization target, f3(x layout ) represents the cost control target, Indicates the sum of all grid points in the entire three-dimensional grid space, α i，j，k is the grid point g i，j，k The weight coefficient reflects the sensitivity of the point to interference; c balance is the electromagnetic field weight balance coefficient; length(l) represents the length of line l; cost(l) represents the unit length cost of line l, Indicates the sum of all lines in the system, L total It represents the total number of lines that need to be wired in the system, and cost(l) represents the unit length cost of line l.

[0121] Through the above sub-steps, the state representation and interference modeling of the line layout problem are completed, providing a mathematical basis for subsequent optimization and learning algorithms.

[0122] Step 2: Using the constructed interference propagation model and interference impact matrix, uncertainty quantification is performed based on Bayesian reasoning and particle filtering methods, including: collecting interference data under different working conditions; obtaining the probability distribution of major uncertainty factors;

[0123] The following sub-steps are included:

[0124] Step 2.1, identification and modeling of environmental uncertainty factors;

[0125] Identify and model the main uncertainties in the circuit layout environment:

[0126] Measurement uncertainty: The sensor measurement noise is modeled as a Gaussian distribution:

[0127]

[0128] Among them, p(z|s real ) means that the real state is s real Under the condition, the conditional probability density of the measured value is z, Indicates that the mean is s real , the covariance matrix is ∑ z The multivariate Gaussian distribution of .

[0129] Environmental parameter uncertainty: Parameter uncertainty caused by factors such as material properties and ambient temperature is modeled as a priori probability distribution p(θ env ), where θ env Represents a set of environmental parameters, such as θ env ={ε r , μ r , σ conduct ,...}, where ε r 、μ r , σ conduct They represent relative permittivity, relative magnetic permeability, and electrical conductivity, respectively.

[0130] Load variation uncertainty: Line load variation is modeled as a conditional probability p(I current，t |I current，t-1 ,ω), where I current，t Represents the load current at time t, I current，t-1 represents the load current value at time t-1, and ω represents the external factors that affect the load change.

[0131] Step 2.2, Bayesian inference model implementation;

[0132] Based on the above uncertainty model, a Bayesian inference framework is implemented to estimate the posterior distribution of the true state of the environment:

[0133] State transition model:

[0134] p(s t |s t-1 ,a t-1 )=∫p(s t |s t-1 ,a t-1 ,θ env )p(θ env )dθ env ;

[0135] Observation model:

[0136] p(o t |s t )=∫p(o t |s t ,θ env )p(θ eny )dθ env ;

[0137] Posterior distribution calculation: Apply Bayes' rule to calculate the posterior distribution of the state:

[0138] p(s t |o 1：t )∝p(o t |s t )∫p(s t |s t-1 ,a t-1 )p(s t-1 |o 1:t-1 )ds t-1 ;

[0139] Among them, p(s t |s t-1 , a t-1 ) represents the conditional probability of the current state, s t represents the system state at time t, s t-1 represents the system state at time t-1, a t-1 represents the layout adjustment action performed at time t-1, θ env represents the set of environmental parameters, d represents differential calculation, p(o t |s t ) represents the conditional probability of a specific observation data, o t represents the observation data at time t, p(s t |o 1：t ) represents the posterior probability distribution of the state at time t under the condition of observing all observation data from time 1 to time t, p(st-1 |o 1：t-1 ) represents the posterior distribution of the state at the previous moment, and ∝ represents proportional to .

[0140] Since the above integral is usually difficult to solve analytically in high-dimensional state space, the particle filter algorithm is used for approximate calculation:

[0141] Sampling M samples particles From the proposed distribution middle;

[0142] Calculate particle weights:

[0143]

[0144] in, represents the weight of the i-th particle at time t, represents the weight of the particle at the previous moment, Indicates that the status is Under the conditions o t The probability of represents the proposed distribution;

[0145] Normalized weights:

[0146]

[0147] in, represents the normalized particle weight, represents the weight of the i-th particle at time t, represents the sum of all particle weights, which serves as a normalization factor;

[0148] Resample according to the effective number of particles to obtain the particle set representing the posterior distribution

[0149] Step 2.3, Monte Carlo simulation scenario generation and robustness assessment;

[0150] Based on the implemented Bayesian model, multiple simulation scenarios are generated and the robustness of the layout solutions is evaluated:

[0151] From the posterior distribution p(s t |o 1：t ) in the sample M samples Environmental status

[0152] For each sampled state, calculate the given layout solution x layout The objective function value of

[0153] Construct robustness evaluation indicators:

[0154] Expected performance:

[0155] Performance variance:

[0156]

[0157] Worst-case performance:

[0158]

[0159] Robustness indicators:

[0160]

[0161] in, represents the average performance of the layout scheme in all sampled environments, represents the weight factor for the average calculation, represents the sum of the performance of all sampling states, M samples Indicates the number of samples taken, represents the environmental state obtained by the jth sampling, x layout represents the circuit layout scheme to be evaluated, Indicates the state of the environment at the jth sampling Under these conditions, the layout scheme x layout Performance rating, Indicates the degree of fluctuation of the performance of the layout scheme in different environments, F worst (x layout ) represents the worst performance of the layout scheme in all sampled environments, Indicates taking the minimum value among all sampling states, R robust (x layout ) represents the comprehensive robustness score of the layout scheme, and λ1, λ2, and λ3 represent the weight coefficients of average performance, performance variance, and worst-case performance, respectively.

[0162] Through the above sub-steps, the quantification of environmental uncertainty and the implementation of the Bayesian inference model are completed, providing a theoretical basis for robust optimization.

[0163] Step 3: Based on the state representation, interference propagation model, and probability distribution of uncertainty factors, an improved multi-objective optimization algorithm is applied to solve the Pareto optimal layout solution set for power and weak current lines, including: setting multiple optimization objectives; generating a set of non-dominated solutions;

[0164] The following sub-steps are included:

[0165] Step 3.1: layout encoding and initial population generation;

[0166] Encode the line layout scheme into a chromosome structure suitable for genetic algorithm processing:

[0167] Path encoding: For each line l, its path consists of a series of continuous grid points and is represented by relative direction encoding, where each step of movement is represented by a vector (dx, dy, dz), dx, dy, dz∈{-1, 0, 1}, where dx, dy, and dz represent movements in the X-axis, Y-axis, and Z-axis directions, respectively.

[0168] In actual implementation, the upper limit of the coding length of each line is set to 200, and lines with more than 200 grid points are segmented and coded;

[0169] Complete layout planx layout Represented as the set of all line paths: Among them, x layout Represents a complete circuit layout solution, Respectively represent the 1st, 2nd, and L total The complete path of the line, L total Represents the total number of lines in the system; for typical line layout problems, L total Generally, there are 10 to 50 lines;

[0170] Initial population generation: Combine heuristic rules and random perturbations to generate N pop Initial layout scheme. The heuristic rules include:

[0171] Shortest Path First: Use the A* algorithm to generate the approximate shortest path for each route, the heuristic function is:

[0172] h heuristic (n) = α heuristic Manhattan(n,goal)+β heuristic EMI(n);

[0173] Among them, h heuristic (n) represents the estimated cost from node n to the target node, α heuristic Represents the weight coefficient of the Manhattan distance term, Manhattan(n,goal) represents the Manhattan distance from node n to the target node, β heuristic represents the weight coefficient of the electromagnetic interference term, EMI(n) is the electromagnetic interference intensity at node n;

[0174] Obstacle avoidance: Ensures that the path avoids pre-set obstacles by setting the cost of obstacle nodes to infinity in the A* algorithm;

[0175] Principle of separation of strong and weak current: In the initial layout, strong and weak current lines maintain a preset minimum distance. In the specific implementation, this distance is set to 3 grid units;

[0176] Random perturbation: Add a certain degree of random variation to the generated path by randomly selecting 20% of the points on the path and replacing them with nearby feasible points;

[0177] In practical applications, the population size N pop Set to 100.

[0178] Step 3.2, non-dominated sorting and congestion calculation;

[0179] Perform multi-objective evaluation and non-dominated sorting on the layout solutions in the population:

[0180] For each layout solution x i , calculate its function value on each target:

[0181] F(x i )=(f1(x i ),f2(x i ),...,f k (x i ));

[0182] Among them, F(x i ) represents the layout scheme x i The multi-objective function value vector, f1(x i )、f2(x i ),f k (x i ) represent the layout scheme x i Function values on the 1st, 2nd, and kth objectives, where k represents the total number of optimization objectives;

[0183] In the strong and weak power line layout problem, k = 3, corresponding to the interference suppression, space utilization and cost control goals respectively;

[0184] Based on the Pareto dominance relationship, the population is divided into different levels of non-dominated frontiers {F1, F2, ...};

[0185] If for all j∈{1,2,...,k}, f j (x m )≥f j (x n ), and there exists at least one j such that f j (x m )>f j (x n ), then it is called plan x m Control Plan x n , denoted as x m >x n ;

[0186] The first frontier F1 consists of solutions in the population that are not dominated by any other solution;

[0187] For r>1, the frontier F r By removing the frontier F1, ..., F r-1 After all the options are selected, the remaining options are those that are not dominated by any other options;

[0188] The complexity of the sorting algorithm is O(MN 2 ), where M is the number of targets and N is the population size;

[0189] Crowding calculation based on reference points:

[0190] Regularize in the target space and generate a uniformly distributed reference point set W = {w1, w2, ..., w H}, where W represents the reference point set, w1, w2, W H They represent the 1st, 2nd, and Hth reference points respectively, and H represents the total number of reference points;

[0191] For the case of three targets, a two-layer grid method was used to generate 91 reference points;

[0192] For each solution x i , calculate its vertical distance to each reference point and assign it to the nearest reference point;

[0193] Calculate the local crowding degree of each reference point as the number of solutions in the area;

[0194] The implementation uses the KD tree data structure to accelerate the search for the nearest reference point, reducing the time complexity from O(NH) to O(NlogH), where N is the number of solutions and H is the number of reference points.

[0195] Step 3.3, selection, crossover and mutation operations;

[0196] Based on non-dominated sorting and crowding, perform genetic operations to generate a new population:

[0197] Selection operation: tournament selection is used, randomly selecting T individuals from the population each time, and selecting the best one based on the following rules:

[0198] If individuals belong to different non-dominated fronts, the individual with the lower (better) front rank is selected;

[0199] If individuals belong to the same frontier, the individual with the smaller crowding degree at the reference point is selected;

[0200] Crossover operation: To ensure that the path after crossing still connects the start and end points, a segment-based crossover operation is used:

[0201] Layout plan x for two parentsa and x b ,For each line, a random intersection point is selected;

[0202] Swap the path segments before and after the selected intersection and add connecting segments as needed to ensure path continuity;

[0203] Check whether the path after intersection meets the constraints. If not, adjust it through the local repair algorithm;

[0204] Mutation operation: Based on the characteristics of the line layout, the following mutation operations are used:

[0205] Path node movement: randomly select a node on the path and move it to the adjacent position while satisfying the constraints;

[0206] Path local rerouting: randomly select a section on the path and use the A* algorithm to recalculate the route of this section;

[0207] Path smoothing: reduce unnecessary turning points to make the path smoother;

[0208] Step 3.4, Pareto front generation and update;

[0209] Iteratively perform the above genetic operations and continuously update the Pareto frontier:

[0210] The parent population and the offspring population are merged to form a population of size 2N pop mixed populations;

[0211] Perform non-dominated sorting on the mixed population to obtain the frontier set {F1, F2, ...};

[0212] Add the frontiers F1, F2, ... to the new population in sequence until the new population size is close to but not larger than N pop ;

[0213] For the last frontier F1 that cannot be fully included, based on the reference point association and crowding, select some solutions to be included in the new population so that the size of the new population is exactly N pop ;

[0214] Repeat the above steps until the termination condition is met (such as reaching the maximum number of iterations or the Pareto front remains stable);

[0215] Through the above sub-steps, the execution of the multi-objective Pareto optimization algorithm is completed, and a series of non-dominated layout schemes representing different trade-off relationships are obtained.

[0216] Step 4: Using the non-dominated solution set as training data, robust reinforcement learning strategy optimization is implemented based on the distributed deep Q network, including: defining the environment state space, action space, and reward function; optimizing the layout decision strategy; and performing robustness evaluation and optimization adjustments on the trained strategy.

[0217] The following sub-steps are included:

[0218] Step 4.1, partially observable Markov decision process modeling;

[0219] The line placement problem is formalized as a partially observable Markov decision process (POMDP):

[0220] State space S: the complete environmental state, including circuit layout, electromagnetic field distribution, material properties, etc., which cannot be directly and completely observed;

[0221] Observation space O: observable data such as electromagnetic field intensity and temperature obtained by sensors;

[0222] Action space A: Layout adjustment actions, including adding new lines, moving lines, adjusting shielding measures, etc.

[0223] Transition function T(s′|s,a): the probability that the state transitions from s to s′ after executing action a;

[0224] Observation function Z(o|s′,a): the probability of observing o after performing action a in state s′;

[0225] Reward function R(s, a): The reward obtained by performing action a in state s, taking into account multiple objectives:

[0226] R(s,a)=w1·R interference (s,a)+w2·R space (s,a)+w3·R cost (s,a);

[0227] Among them, R(s, a) represents the comprehensive reward value obtained after executing action a in state s, R interference (s, a) represents the interference suppression reward component, R space (s, a) represents the space utilization reward component, R cost (s, a) represents the cost control reward component, w1, w2, and w3 represent the weight coefficients of interference suppression, space utilization, and cost control reward components, respectively.

[0228] Discount factor γ∈[0,1]: controls the importance of future rewards;

[0229] The strategy optimization goal is to find the optimal strategy π * , which maximizes the expected cumulative discounted reward:

[0230]

[0231] Among them, π * represents the optimal strategy, π(b t ) is in belief state b t Next, the action selected by strategy π, E represents the mathematical expectation, represents the cumulative discounted reward obtained by executing strategy π starting from the initial state, γ t represents the discount factor of the reward at step t, b t is the belief state at time t, that is, the probability distribution b of the real environment state t =p(s t |o 1∶t ,a 0∶t-1 ), where p(s t |o 1∶t ,a 0∶t-1 ) indicates that when the historical observation sequence o is observed 1：t and execute historical action sequence a 0：t-1 Under the condition of t probability estimate.

[0232] Step 4.2, distributed deep Q network implementation and training;

[0233] Based on the partially observable characteristics, the distributed deep Q network (D-DQN) structure is realized:

[0234] Belief state encoder: transforms the observation history o 1：t and action history a 0：t-1 Encoded as belief state representation b t ;

[0235] Adopting recurrent neural network structure: h t =f RNN (h t-1 ,[o t , a t-1 ]), where h t represents the hidden vector at the current time t, h t-1 represents the hidden state vector at the previous moment t-1, o t represents the observation data at the current time t, a t-1 represents the action performed at the previous time t-1, f RNN Represents the state update function of the recurrent neural network;

[0236] In the specific implementation, a two-layer LSTM network is used, each layer contains 256 hidden units, and the input is the observation data o t(a vector containing environmental indicators such as the current electromagnetic field strength and temperature, with a dimension of 16) and the previous action a t-1 The output of LSTM passes through a 256×128 fully connected layer to generate the belief state representation b t ;

[0237] Output belief state representation: b t =f b (h t ), where b t Represents the belief state vector at time t, f b represents the mapping function from the LSTM hidden state to the belief state, h t Represents the hidden state of the LSTM network;

[0238] Distributed Q function: predicts not only the expectation of Q value but also its distribution to characterize uncertainty;

[0239] Model the Q-values as a set of different quantiles: in, Respectively represent the belief state b t The 1st, 2nd, and Nth quantiles of the Q-value distribution of action a are executed under [1]. N represents the total number of quantiles used.

[0240] In the specific implementation, N = 51 quantiles are selected, which are evenly distributed in the interval [0, 1];

[0241] The network structure adopts a three-layer fully connected network with layer sizes of [128, 256, 51×|A|], where |A| is the size of the action space, and the last layer outputs the value of each action at each quantile;

[0242] where τ i ∈[0, 1] is the preset quantile level, and N is the number of quantiles;

[0243] Dual network architecture:

[0244] Main network Q(b, a; θ): used for action selection;

[0245] Target network Q(b, a; θ - ): used for target value calculation, parameter θ - Regularly replicate from the main network;

[0246] In the actual implementation, after every 10,000 steps of interaction, the main network parameters are completely copied to the target network;

[0247] Distributed Q-network training:

[0248] Sample batches from the experience replay buffer (o t , a t , r t , o t+1 ), the batch size is set to 64, where o t represents the observation at time t, a t represents the action executed at time t, r t Represents the reward obtained after executing the action, o t+1 represents the observation at the next moment;

[0249] Calculate belief state: b t =f b (f RNN (h t-1 ,[o t , a t ])), where b t represents the belief state, f RNN Represents the state update function of the recurrent neural network, f b Represents the mapping function from the RNN hidden state to the belief state;

[0250] For each quantile τ i , calculate the target value:

[0251]

[0252] in, represents the target Q value of the i-th quantile, r t represents the immediate reward, γ represents the discount factor, argmax a represents the action with the highest value under the evaluation of the main network, θ represents the parameter set of the main network, θ - Represents the parameter set of the target network;

[0253] Update the network using quantile regression loss:

[0254]

[0255] Among them, L(θ) represents the loss function, represents the number of quantiles, and B represents the batch size. represents the quantile regression loss function, represents the i-th quantile target value of the j-th sample in the batch, Represents the main network's estimate of the i-th quantile Q value of the j-th sample;

[0256] The Adam optimizer is used, the learning rate is set to 0.0005, and the training process is carried out for 1 million steps of interaction in the circuit layout task.

[0257] Step 4.3: Progressive strategy exploration and experience replay optimization;

[0258] Implement a progressive strategy exploration mechanism and optimize the experience replay strategy:

[0259] Progressive policy exploration: using a parameterized noise policy that provides more structured exploration than a simple ∈ greedy policy;

[0260] Add noise to the network parameters: θ noisy =θ+σ⊙∈, where θ noisy represents the network parameters after adding noise, θ represents the original network parameters, σ is the learnable noise scale parameter, ⊙ represents the element-wise multiplication operator, and ∈ is the random noise sampled from the standard normal distribution; the noise scale gradually decreases with the training process, achieving a smooth transition from exploration to utilization.

[0261] Priority experience replay: Set priorities based on the temporal difference error of samples to improve learning efficiency;

[0262] Sample priority calculation: p i =|δ i |+∈ small ; Among them, p i represents the priority of the i-th sample, |δ i | represents the absolute value of the time series difference error of the i-th sample, ∈ small Represents a small positive number;

[0263] Sampling probability calculation: Among them, P(i) represents the probability of the i-th sample being sampled, represents the power of the priority of the i-th sample, represents the sum of all sample priority powers, and α controls the degree of priority usage;

[0264] Importance weight calculation: Among them, w i represents the importance weight of the i-th sample, N is the playback buffer size, P(i) represents the probability that the i-th sample is sampled, and β controls the degree of importance sampling;

[0265] Multi-timescale experience replay: Maintain replay buffers at different time scales to capture short-term and long-term dependencies;

[0266] Short-term buffer: stores recent interaction experiences, has a relatively small capacity, and is updated frequently;

[0267] Long-term buffer: stores historical representative experience, has a large capacity, and is updated infrequently;

[0268] Sample proportionally from both buffers during interactive training;

[0269] Step 4.4, strategy evaluation and adaptive deployment;

[0270] Evaluate the performance of learning strategies under different environmental conditions and implement adaptive deployment mechanisms:

[0271] Robustness evaluation: Evaluate policy performance under various simulated environment conditions;

[0272] For the environmental state sample set generated in step 2 Evaluate the performance of strategy π on each sample;

[0273] Calculate robustness metrics: expected return E[R(π)], performance variance Var[R(π)], and worst-case performance

[0274] Strategy adaptive adjustment: Dynamically adjust strategy parameters according to environmental changes;

[0275] Maintain a set of policies {π1, π2, ..., π K}, each strategy is suitable for different environmental conditions, where π1, π2, π K They represent the 1st, 2nd, and Kth strategies respectively, and K represents the total number of strategies in the strategy library;

[0276] Based on the current observation t and historical information to estimate current environmental conditions and select the most appropriate strategy;

[0277] Monitor strategy performance in real time and trigger strategy switching when environmental conditions change beyond a threshold;

[0278] Through the above sub-steps, the robust reinforcement learning strategy optimization is completed, so that the layout solution can adapt to the uncertainty and changes of the environment.

[0279] Step 5: Combine the non-dominated solution set and the optimized layout decision strategy to provide decision support through the preference learning model, including: building a preference learning model; evaluating and ranking all non-dominated solutions; and recommending the optimal layout solution.

[0280] The following sub-steps are included:

[0281] Step 5.1, expert preference data collection and representation;

[0282] Collect and represent expert preference data for subsequent preference learning:

[0283] Preference query system implementation:

[0284] Select a pair of layout options (x i , x j ) presented to an expert;

[0285] Record expert preference feedback i，j ∈{1, 0, -1}, respectively representing preference x i , no obvious preference, preference x j ;

[0286] Active learning strategies are used to select the most informative query pairs, thus reducing the number of required expert feedbacks.

[0287] Preference data indicates:

[0288] Create a preference dataset D = {(x i , x j ,y i，j )}, where D represents the preference dataset, (x i , x j ) is a pair of layout schemes, y i，j is the preferred label of the experts;

[0289] Each layout scheme X is represented by its performance vector F(x)=(f1(x), f2(x), ..., f k (x)), where F(x) is used as the input of the utility function to learn the implicit preferences of the expert, f1(x), f2(x), f k (x) represents the performance score of layout scheme x on the 1st, 2nd, and kth different objectives, respectively, and k represents the total number of optimization objectives.

[0290] Step 5.2, implicit utility function learning;

[0291] Based on the collected preference data, learn the utility function implicit in the expert decision:

[0292] Utility function parameterization:

[0293] The expert's implicit utility function is parameterized as:

[0294] u(x;ω)=g ω (F(x));

[0295] Among them, u(x;ω) represents the parameterized utility function, g ω It is a neural network model controlled by the parameter ω. The input is the performance vector of the scheme on each objective. F(x) is the input of the utility function, which is used to learn the implicit preferences of experts.

[0296] In the specific implementation, g ω A four-layer fully connected neural network structure is used, and the specific configuration is:

[0297] Input layer: The dimension is the target number (3 in this example, corresponding to interference suppression, space utilization, and cost control);

[0298] First hidden layer: 32 neurons, ReLU activation function;

[0299] Second hidden layer: 16 neurons, ReLU activation function;

[0300] The third hidden layer: 8 neurons, ReLU activation function;

[0301] Output layer: 1 neuron, linear activation function;

[0302] To improve the expressiveness of the model, a batch normalization layer and a dropout layer (with a dropout rate of 0.2) are added after each hidden layer.

[0303] Bayesian Preference Learning:

[0304] Definition of preference probability model: When the expert compares the options (x i , x j ), prefer x i The probability is:

[0305] P(y i，j =1|ω)=σ(u(x i ;ω)-u(x j ;ω));

[0306] Among them, P(y i，j =1|ω) means that under the given parameter ω, the expert prefers the solution x i Rather than x j The probability of u(x i ;ω),u(x j ; ω) represent the scheme x i and x j The utility value of , σ is the sigmoid function;

[0307] Based on the preference data set D, calculate the posterior distribution of the utility function parameters:

[0308]

[0309] Among them, p(ω) represents the posterior probability distribution of parameter ω after a given preference dataset D. Represents the likelihood function product of all preference data, p(ω) represents the prior distribution of parameter ω, and multivariate Gaussian distribution is used in the implementation Where λ = 0.01 is the regularization parameter, and I represents the identity matrix;

[0310] Variational inference is applied to approximate the posterior distribution:

[0311] Using parameterized distribution q φ (ω) approximates the true posterior distribution p(ω|D);

[0312] In the implementation, q φ (ω) uses a diagonal Gaussian distribution where φ={μ φ , σ φ} is the variational parameter, μ φ represents the mean vector of the variational distribution, σ φ represents the standard deviation vector of the variational distribution;

[0313] Minimize the KL divergence KL(q φ (ω)||p(ω|D));

[0314] This is equivalent to maximizing the evidence lower bound:

[0315]

[0316] Among them, ELBO(φ) represents the lower bound of evidence, represents the expected value of the log-likelihood under the variational distribution, KL(q φ (ω)||p(ω)) represents the KL divergence between the variational distribution and the prior distribution;

[0317] The unbiased estimation of gradients is achieved through reparameterization techniques, and the Adam optimizer is used to optimize the variational parameters with a learning rate of 0.001.

[0318] In the line layout scenario, 50 to 100 pairs of expert preference data are usually collected for training, and convergence is achieved after 10,000 iterations;

[0319] Step 5.3, Pareto solution evaluation and ranking;

[0320] Use the learned utility function to evaluate and rank the layout options on the Pareto front:

[0321] Program effectiveness evaluation:

[0322] For each option x on the Pareto frontier, calculate its expected utility:

[0323]

[0324] in, represents the expected utility of option x, Indicates that in the variational distribution q φ (ω), the expected value of the utility function u(x;ω), ω represents the parameter of the utility function, q φ represents an approximation to the posterior distribution of the parameters obtained by variational inference;

[0325] Computing utility uncertainty represents the confidence in the assessment, where σu (x) represents the standard deviation of the utility evaluation of option x, represents the variance of the utility function value under variational distribution;

[0326] Scheme sorting and selection:

[0327] Sort based on upper confidence interval criteria: Among them, UCB(x) represents the upper confidence interval value of solution x, represents the expected utility of option x, σ u (x) represents the uncertainty of the scheme utility, β is the exploration and utilization balance parameter; the K schemes with the highest confidence interval of the utility upper limit are recommended to the decision maker.

[0328] Step 5.4, interactive visual decision interface implementation;

[0329] Implement an interactive visual interface to support decision makers in exploring and selecting the final layout plan:

[0330] Multidimensional target space visualization:

[0331] Use parallel coordinates plots to display the Pareto front in multi-objective space;

[0332] Use dimensionality reduction techniques (such as t-SNE, UMAP) to map the high-dimensional target space into 2D / 3D space for visualization;

[0333] Solution comparison and screening tools:

[0334] Provides a detailed comparison view of the plans, showing the specific performance of different plans in terms of each goal;

[0335] Support dynamic setting of target weights and real-time update of solution rankings;

[0336] Provides scenario filtering function, allowing setting threshold conditions for each target;

[0337] Interactive exploration of layout options:

[0338] The 3D visualization interface shows the specific circuit layout of the selected layout scheme;

[0339] Support interactive operations such as zooming, rotating, and transparency adjustment;

[0340] The electromagnetic field intensity heat map is superimposed and displayed to intuitively show the interference distribution;

[0341] Decision feedback and learning update system:

[0342] Record the decision makers' interactions and final choices;

[0343] Incorporate new decision data into the preference dataset and continuously update the preference model;

[0344] Continuously accumulate experience and improve the quality of recommendations during the use of the system;

[0345] Through the above sub-steps, the preference learning and decision support system is implemented, helping decision makers to efficiently select the layout solution that best meets actual needs from many Pareto optimal solutions.

[0346] The intelligent layout and interference prediction analysis method for strong and weak current lines provided by this embodiment greatly improves the quality and efficiency of line layout design, reduces the risk of electromagnetic interference, and enhances the system's adaptability to environmental changes. It has significant technical value and application prospects.

[0347] Real-world application examples of this implementation:

[0348] This example applies to the design of an electrical control cabinet for a large-scale automated production line. The control cabinet requires numerous power and weak current lines, including three-phase 380V power lines, single-phase 220V control power lines, 24V DC control lines, and various signal transmission lines (analog 4-20mA signals, digital switching signals, and high-speed communication buses). The control cabinet, measuring 2000mm × 800mm × 600mm, is compact and contains numerous structural obstructions, such as mounting brackets and cooling systems.

[0349] Specific line requirements are shown in Table 1:

[0350] Table 1: List of wiring requirements in the control cabinet;

[0351]

[0352] The circuit layout faces the following challenges:

[0353] Small space: The available space in the cabinet is limited and needs to be used efficiently;

[0354] Complex interference: Strong power lines generate strong magnetic fields, which interfere with weak power signal lines, especially analog signal lines and communication buses;

[0355] Load changes: When the production line is running, the load of each device will change, causing the electromagnetic interference characteristics to change dynamically;

[0356] Environmental uncertainty: The temperature distribution inside the cabinet is uneven, and the electromagnetic shielding effect varies at different locations;

[0357] Cost control: The amount of cables and shielding materials used must be reduced while ensuring performance.

[0358] Implementation process example:

[0359] For the above application scenarios, we implemented the intelligent layout and interference prediction analysis of strong and weak power lines according to the steps of the embodiment:

[0360] Step 1 Example: State Characterization and Disturbance Modeling;

[0361] The control cabinet space was discretized into a 100×40×30 three-dimensional grid, with each cell measuring 20mm×20mm×20mm. Based on electromagnetic field theory, an interference propagation model between lines was established. Table 2 shows the electromagnetic interference baseline values for some key locations:

[0362] Table 2: Electromagnetic interference reference values at key locations within the control cabinet;

[0363] Position Description Coordinates (x, y, z) Electric field strength (V / m) Magnetic field strength (μT) Interference sensitivity coefficient α PLC analog input area (30,25,15) 0.5 0.8 10.0 Communication bus junction area (45,15,20) 0.4 0.6 15.0 High-voltage line concentration area (20,10,10) 25.0 35.0 0.1 Power transformer peripherals (10,5,5) 30.0 50.0 0.05 Equipment cooling area (60,30,25) 1.2 1.5 0.5

[0364] Step 2 Example: Uncertainty Quantification and Bayesian Reasoning;

[0365] By installing 12 interference monitoring sensors in the control cabinet, interference data was collected under different load conditions. A particle filter algorithm was applied to estimate the posterior probability of the interference distribution based on historical observation data. Table 3 shows the main uncertainty factors identified in the system and their modeling parameters:

[0366] Table 3: Environmental uncertainty modeling parameters;

[0367]

[0368] Based on the above uncertainty model, 1,000 sets of simulation scenarios were generated to evaluate the performance of each layout scheme under different conditions.

[0369] Step 3 Example: Multi-objective Pareto optimization algorithm execution;

[0370] For the control cabinet layout problem, an improved NSGA-III algorithm was implemented to simultaneously optimize interference suppression, space utilization, and cost control. With an initial population size of 100, 1000 iterations were performed, ultimately obtaining 28 non-dominated solutions on the Pareto front. Table 4 shows five typical Pareto-optimal solutions.

[0371] Table 4: Performance comparison of some Pareto optimal layout solutions;

[0372]

[0373] Step 4 Example: Robust reinforcement learning policy optimization;

[0374] A layout strategy optimization system based on a distributed deep Q-network (D-DQN) was implemented, and a robust strategy was obtained through 1 million training cycles of interaction with the layout environment. In the control cabinet layout, the input observation dimension was set to 16, including information such as the electromagnetic field strength at the current location, the distribution of nearby obstacles, and the status of the routed lines. The action space was defined as 26 discrete actions, including movement in six directions, placement of various lines, and special processing actions. Table 5 shows the performance of the strategy under different load conditions:

[0375] Table 5: Performance of reinforcement learning strategies under different working conditions;

[0376] Working condition description Average reward value Interference control effect Average path length Adaptive adjustment times Light load (30%) 845.6 Excellent 325.7 3 Medium load (60%) 782.3 good 331.4 5 Heavy load (90%) 720.8 medium 338.2 8 Fluctuating load (30%-90%) 755.4 good 335.6 12 Sudden interference conditions 698.2 medium 342.1 15

[0377] Step 5 Example: Preference Learning and Decision Support;

[0378] We collected 85 pairs of layout solution preferences from 10 experienced electrical design engineers and constructed a preference learning model. After model training, we evaluated and ranked 28 non-dominated solutions on the Pareto front. Table 6 shows the model's evaluation results for some of the solutions.

[0379] Table 6: Evaluation results of the scheme based on preference learning;

[0380]

[0381] Based on the above evaluation results, Scheme P2 was finally selected as the wiring layout scheme for the control cabinet. This scheme achieved the best balance between interference suppression and cost control and best met the implicit preferences of the design experts.

[0382] Technical effect verification:

[0383] Layout Plan P2 of this implementation was compared with traditional manual layout solutions and single-objective optimization solutions, focusing on verifying the two core technical effects of "multi-objective balancing" and "interference robustness." The tests were conducted in an actual control cabinet, using professional electromagnetic interference testing equipment and an industrial control system status monitoring system to collect data.

[0384] Multi-objective balance capability verification:

[0385] This method is compared with traditional single-objective optimization methods. Traditional methods optimize for single objectives: interference minimization, path shortesting, and cost minimization. Table 7 shows the performance comparison of each layout scheme on these three objectives:

[0386] Table 7: Comparison of multi-objective performance of different layout schemes;

[0387]

[0388] As shown in Table 7, this method achieves a better overall balance across multiple objectives than single-objective optimization and traditional manual layout. Although it performs worse than single-objective methods optimized for a single dimension, it achieves the best overall performance, improving the overall score by approximately 35%, while also offering moderate implementation difficulty.

[0389] Anti-interference robustness verification:

[0390] To verify the robustness of different layout schemes in the face of environmental uncertainty and interference changes, we tested the performance of each scheme under five different load conditions and recorded key signal quality indicators. Table 8 shows the error rate of high-sensitivity signals under different conditions:

[0391] Table 8: Comparison of signal quality under different working conditions (signal error rate %);

[0392] Load conditions This method (P2) Traditional manual layout Single-objective optimization method Improvement (%) No load (10%) 0.05 0.08 0.06 37.5 / 16.7 Light load (30%) 0.08 0.15 0.12 46.7 / 33.3 Medium load (60%) 0.12 0.28 0.22 57.1 / 45.5 Heavy load (90%) 0.18 0.45 0.35 60.0 / 48.6 Load fluctuation (20%-80%) 0.15 0.52 0.41 71.2 / 63.4 Sudden interference conditions 0.22 0.68 0.54 67.6 / 59.3 Average improvement - - - 56.7 / 44.5

[0393] In addition, we also tested the system's adaptability to sudden load changes. Figure 1 (Table display) records the real-time changes in communication bus signal quality when the load changes from 30% to 90%:

[0394] Table 9: Signal quality recovery capability under load mutation (SNR, dB);

[0395] Time point (seconds) This method (P2) Traditional manual layout Single-objective optimization method t=0 (before mutation) 65.8 63.2 64.5 t=0.5(after mutation) 48.6 32.5 38.2 t＝1.0 52.3 34.6 39.5 t＝2.0 58.7 38.2 42.1 t＝5.0 61.4 42.8 48.3 t＝10.0 63.5 45.6 52.7 Recovery rate (%) 96.5 72.2 81.7

[0396] The data in Tables 8 and 9 demonstrate that this method demonstrates excellent robustness in the face of environmental uncertainty and interference variations, improving its robustness by approximately 56.7% and 44.5% compared to traditional manual placement and single-objective optimization methods, respectively. The improvement is particularly significant under load fluctuations and sudden interference conditions, fully validating the robustness improvements achieved by this method through Bayesian reasoning and uncertainty modeling.

[0397] This application example proves that the intelligent layout and interference prediction analysis method of strong and weak current lines proposed in this embodiment can be effectively applied in actual industrial control systems, significantly improving the layout quality and enhancing the reliability of the system under complex working conditions, with obvious technical and economic value.

[0398] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Ordinary technicians in this field can also make more forms of equivalent embodiments based on the inspiration of this embodiment, all of which are protected by this embodiment.

Claims

1. A method for intelligent layout and interference prediction analysis of strong and weak current lines, characterized in that: The following steps are involved: Construct state representation and interference propagation model of electronic system layout, including: establishing electromagnetic interference propagation model; forming interference impact matrix; Using the constructed interference propagation model and interference impact matrix, uncertainty quantification is performed based on Bayesian reasoning and particle filtering methods, including: collecting interference data under different working conditions; obtaining the probability distribution of major uncertainty factors; Based on state representation, interference propagation model and probability distribution of uncertainty factors, an improved multi-objective optimization algorithm is applied to solve the Pareto optimal layout solution set of power and weak current lines, including: setting multiple optimization objectives; generating a set of non-dominated solutions; Using the set of non-dominated solutions as training data, robust reinforcement learning strategy optimization is implemented based on a distributed deep Q-network, including: defining the environment state space, action space, and reward function; optimizing the layout decision strategy; and performing robustness evaluation and optimization adjustments on the trained strategy. Combining the set of non-dominated solutions and the optimal layout decision strategy, decision support is provided through a preference learning model, including: constructing a preference learning model; evaluating and ranking all non-dominated solutions; and recommending the optimal layout solution.

2. A method for intelligent layout and interference prediction and analysis of strong and weak current lines according to claim 1, characterized in that: The construction of the state representation and interference propagation model of the electronic system layout specifically includes: Discretize the layout space into G = {g i，j，k }, where G represents the grid set of the entire layout space, i, j, and k represent the X-axis, Y-axis, and Z-axis coordinates of the space, respectively. i，j，k A grid cell representing a specific location in three-dimensional space; For any two circuit elements e a and e b Potential interference between I(e a , e b ), and establish a calculation model based on electromagnetic field theory: Among them, I(e a , e b ) represents the element e a to element e b The electromagnetic interference intensity, P a Indicates the interference source power, S b Indicates the sensitivity of the interfered component, d ab Represents the distance between two elements, f(θ ah ) represents the direction factor, g(ω a ,ω b ) represents the frequency correlation function, α is the correction coefficient, and α is the comprehensive correction coefficient; The calculation results are organized into the interference influence matrix M n×n , where n is the total number of circuit elements in the system.

3. A method for intelligent layout and interference prediction and analysis of strong and weak current lines according to claim 1, characterized in that: The uncertainty quantification based on Bayesian reasoning and particle filtering method specifically includes: Install multiple electromagnetic interference monitoring sensors {s1, s2, ..., s m }, collect interference data under different load conditions, where s1, s2, s m They represent the 1st, 2nd, and mth electromagnetic interference monitoring sensors deployed in the system, respectively, and m represents the total number of sensors; Establish state transition model p(x t |x t-1 ) and the measurement model p(z t |x t ), where x t represents the interference state at time t, x t-1 represents the interference state at time t-1, z t Indicates the corresponding measurement value; The particle filter algorithm is used to estimate the posterior probability of the interference distribution: Among them, p(x t |z 1：t ) indicates that in the observation sequence z 1：t Under the condition, the interference state x at time t t The posterior probability distribution of represents the particle set and the corresponding weight, δ is the Diracdelta function, represents the i-th particle, N represents the total number of particles, represents the sum of all N particles; The probability distribution of the main uncertainty factors is obtained through sampling and importance resampling steps.

4. The method for intelligent layout and interference prediction analysis of strong and weak current lines according to claim 1 is characterized in that: The improved multi-objective optimization algorithm specifically includes: Set the optimization objective function set: F(x)={f1(x),f2(x),...,f k (x)}; Among them, F(x) represents the set of objective functions in the multi-objective optimization problem, X represents the decision variable vector of the layout solution, f1(x) represents the interference suppression metric, f2(x) represents the space utilization, and f k (x) represents cost control, and k represents the total number of optimization objectives; The improved NSGA-III algorithm is used for multi-objective optimization, including: Using adaptive crossover probability p c and mutation probability p m : Among them, p c represents the crossover probability in the current iteration, p c0 represents the initial value of the crossover probability, p m Represents the mutation probability in the current iteration, p m0 represents the initial value of the mutation probability, G is the current generation, β1 and β2 represent the amplitude coefficients of the crossover probability and mutation probability, γ1 and γ2 represent the attenuation rates of the crossover probability and mutation probability respectively; A reference point-based selection mechanism is used to ensure the diversity of the solution set in the target space; Generate non-dominated solution set P * , represents the optimal compromise between multiple objectives.

5. The method for intelligent layout and interference prediction analysis of strong and weak current lines according to claim 1 is characterized in that: The robust reinforcement learning strategy optimization based on the distributed deep Q network specifically includes: Define the state space S, action space A and reward function R: The state space contains the current layout state and uncertainty estimates; The action space is defined as the possible placement and direction of the lines; The reward function comprehensively considers interference suppression, space utilization, and cost control; Training a distributed deep Q-network: Use double Q learning strategy to reduce Q value estimation bias; Apply the experience replay mechanism D to store the transformed samples (s t , a t , r t , s t-1 ), where s t represents the state at time t, a t represents the action executed at time t, r t Indicates execution of action a t The immediate reward after t+1 Indicates execution of action a t The new state to which it is transferred; Minimize the loss function: L(θ)=E (s,ar,s')～D [(r+γmax a′ Q(s',a';θ - )-Q(s,a;θ)) 2 ]; Among them, L(θ) represents the loss function of network training, E (s，a，r，s′)～D represents the expectation sampled from the experience replay buffer D, r represents the immediate reward, γ represents the discount factor, and max a′ Q(s′, a′; θ - ) indicates the use of target network parameters θ - The maximum Q value of the next state s′ is calculated. Q(s, a; θ) represents the Q value of the state-action pair (s, a) calculated using the current network parameters θ, θ and θ - Represent the current network and target network parameters respectively; Through Monte Carlo simulation and scenario analysis, the robustness of the strategy in an uncertain environment is evaluated and adjustments and optimizations are made.

6. The method for intelligent layout and interference prediction analysis of strong and weak current lines according to claim 1 is characterized in that: The decision support through the preference learning model specifically includes: Collect expert evaluation dataset D p ={(x i , x j ,y ij )}, where D p represents the expert evaluation dataset, x i 、x j Represents the i-th and j-th layout schemes, y ij Indicates the expert's opinion on solution x i and x j degree of preference; An enhanced twin neural network is constructed for the preference learning model, and its loss function is defined as: Among them, L p (θ) represents the loss function of the preference learning model, Denotes the expert evaluation dataset D p Sum all data samples in, f θ represents the preference learning model, y ij Indicates the expert's opinion on solution x i and x j The preference degree, λ is the regularization parameter, represents the regularization term; Use the trained preference model to analyze the non-dominated solution set P * Score all layout options in: Score(x)=f θ (x))+α·diversity(x) Among them, score(x) represents the comprehensive score of layout plan x, f θ (x) represents the utility score of the preference model for layout option x, diversity(x) is the diversity measure, and α is the balance parameter; Recommend the optimal layout solution based on the comprehensive score.

7. The method for intelligent layout and interference prediction analysis of strong and weak current lines according to claim 1 is characterized in that: It also includes a comprehensive consideration of the importance of different objectives through an adaptive weighting mechanism, where the objective weights are dynamically adjusted according to environmental conditions and user needs: Among them, W i (t) represents the weight of the i-th target at time t, β i (t) is a parameter that changes with environmental conditions, Indicates the exponential transformation of the importance parameter, represents the sum of the exponential transformations of all k objectives, where k represents the total number of optimization objectives.

8. The method for intelligent layout and interference prediction analysis of strong and weak current lines according to claim 1 is characterized in that: The uncertainty quantification based on Bayesian reasoning and particle filtering also includes an uncertainty propagation analysis method for evaluating the propagation characteristics of uncertainty in the system: Build an uncertainty propagation graph: G u =(V,E,W); Among them, G u represents the uncertainty propagation diagram, V represents the system components, E represents the association between components, and W represents the uncertainty propagation intensity; The Fokker-Planck equation is used to describe the evolution of uncertainty in the system: in, represents the rate of change of the probability density function over time, p(x, t) represents the probability density of the system state x at time t, A(x) and B(x) represent the drift and diffusion coefficients respectively, represents the deterministic drift term, represents the random diffusion term.

9. A system for intelligent layout and interference prediction analysis of strong and weak current lines, for implementing the intelligent layout and interference prediction analysis of strong and weak current lines as claimed in claims 1-8, comprising: State representation and interference modeling module, used to build state representation and interference propagation models of electronic system layout; Uncertainty quantification module, used for uncertainty quantification based on Bayesian reasoning and particle filtering methods; Multi-objective optimization module, used to apply an improved multi-objective optimization algorithm to solve the Pareto optimal layout solution set for power and weak current lines; Reinforcement learning strategy module, used to implement robust reinforcement learning strategy optimization based on distributed deep Q network; Preference learning and decision support module, used to provide decision support through preference learning models; Human-computer interaction interface for visualizing layout solutions and receiving user feedback.

10. A computer-readable medium having a computer program recorded thereon, wherein when the computer program is executed by a processor, the method for intelligent layout and interference prediction and analysis of strong and weak current lines according to any one of claims 1 to 8 is implemented.

Images