Power distribution network photovoltaic access optimization positioning method based on improved PSO

Abstract

The invention discloses a power distribution network photovoltaic access optimization positioning method and system based on improved PSO, and relates to the technical field of power distribution network optimization, and the method comprises the steps: obtaining irradiance time sequence data and temperature field distribution data, constructing a photovoltaic output calculation model, and mapping the output quantity of the calculation model to an inertia weight function of a particle swarm optimization algorithm; establishing a dynamic evaluation model based on an inertia weight function in combination with photovoltaic output dynamic characteristics, setting adaptive adjustment rules of cognitive factors and social factors, and inputting the adaptive adjustment rules into an improved speed position updating equation to realize dynamic optimization of particle swarm search characteristics; a multi-target evaluation system is established, a regularization evaluation function is constructed, and an improved particle swarm optimization algorithm is combined to determine the photovoltaic optimal access position meeting the steady-state operation requirement of the power distribution network through iterative optimization. According to the method, the photovoltaic optimal access position meeting the system operation constraint can be effectively determined, and an important reference basis is provided for optimization and adjustment of the system operation mode.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network optimization, and particularly to an optimized positioning method for photovoltaic access in a distribution network based on an improved PSO. Background Art

[0002] With the rapid development of photovoltaic power generation technology and the continuous promotion of renewable energy consumption policies, the access scale of distributed photovoltaic power generation in the distribution network has been continuously expanding. Reasonably selecting the access location of photovoltaic power has important significance for giving full play to its technical advantages and improving the system operation efficiency. The existing optimized methods for photovoltaic access locations mainly evaluate based on single indicators such as voltage deviation and network loss rate, and it is difficult to comprehensively reflect the operation state of the system; or multi-objective optimization methods are adopted, but the dimensions of various indicators are different and the weights are difficult to determine, resulting in the reliability of the optimization results being affected. In terms of optimization algorithms, the traditional particle swarm optimization algorithm is widely adopted due to its simple structure and easy implementation, etc., but the basic particle swarm optimization algorithm has problems such as being prone to falling into local optima and slow convergence speed.

[0003] Although many improvement strategies have been proposed in existing research, such as introducing inertia weight, adaptive parameters, etc., these improvement methods often only focus on the performance of a certain aspect of the algorithm and lack systematic consideration of the dynamic balance between the global search ability and local development ability in the optimization process. In addition, existing methods mostly use the weighted summation method to construct the objective function when dealing with multiple operation constraints, which not only requires artificial setting of weight coefficients but also is difficult to ensure strict satisfaction of the constraints. Therefore, how to establish a scientific and reasonable evaluation system, coordinate and process multiple constraint indicators, and improve the performance of the optimization algorithm at the same time is a key problem to be solved urgently. These technical difficulties seriously restrict the application effect of the photovoltaic access location optimization method in practical engineering and affect the improvement of the photovoltaic consumption capacity of the distribution network. Summary of the Invention

[0004] In view of the problems existing in the existing optimized positioning method for photovoltaic access in a distribution network based on an improved PSO, the present invention proposes an optimized positioning method for photovoltaic access in a distribution network based on an improved PSO.

[0005] Therefore, the problem to be solved by the present invention lies in the technical problem of how to achieve the dynamic balance between the global search and local development capabilities of the adaptive weight particle swarm optimization algorithm and coordinate and process multiple operation constraint indicators.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides an optimized positioning method for photovoltaic access in a distribution network based on an improved PSO, which includes: based on the photovoltaic conversion mechanism of photovoltaic modules, obtaining irradiance time-series data and temperature field distribution data, constructing a photovoltaic output calculation model, and mapping the output of the calculation model to the inertia weight function of the particle swarm optimization algorithm to realize the coupling of physical characteristics and the optimization algorithm; establishing a dynamic evaluation model based on the inertia weight function combined with the dynamic characteristics of photovoltaic output, setting adaptive adjustment rules for the cognitive factor and the social factor, and inputting them into the improved velocity-position update equation to realize the dynamic optimization of the particle swarm search characteristics; establishing a multi-objective evaluation system by integrating network loss indicators, voltage deviation indicators, and photovoltaic benefit indicators, constructing a regularization evaluation function, introducing a penalty term based on power flow constraints to meet the operation constraints of the distribution network, and determining the optimal photovoltaic access position that meets the steady-state operation requirements of the distribution network through iterative optimization in combination with the improved particle swarm optimization algorithm.

[0008] As a preferred solution of the optimized positioning method for photovoltaic access in a distribution network based on the improved PSO of the present invention, wherein: the photovoltaic output calculation model receives the light intensity sequence and the temperature sequence as inputs, and calculates the theoretical photovoltaic output power, expressed as:

[0009] P 0 (t) = η × S × G(t) × [1 - γ(T(t) - T ref )]

[0010] Wherein, P 0 (t) represents the theoretical photovoltaic output power at time t, η represents the photovoltaic conversion efficiency, S represents the effective light-receiving area of the photovoltaic array, γ represents the sensitivity coefficient of the photovoltaic cell output power to temperature change, and T ref represents the reference temperature under standard test conditions; based on the light intensity sequence, temperature sequence, and theoretical photovoltaic output power, a BP neural network prediction model with a double hidden layer structure is constructed to predict the actual photovoltaic output, output a predicted power sequence, and calculate the relative prediction error through the comparison of the predicted value and the actual value; based on the standardized predicted power sequence and error sequence, a double-layer weight function is constructed to calculate the inertia weight value in the particle swarm optimization algorithm, and an inertia weight function is constructed, expressed as:

[0011] w(t) = w b (t) + w c (t)

[0012] w b (t) = w max - (w max - w min ) × P n (t)

[0013] wc w(t) = -β × E n w(t) × [w max - w b (t)]

[0014] Where, w(t) is the inertia weight value of the particle swarm optimization algorithm at time t, w b (t) is the basic weight term, w c (t) is the error compensation term, w max is the maximum weight value, w min is the minimum weight value, p n (t) is the predicted power sequence normalized by the maximum and minimum values, with a value range of [0, 1]; β is the error compensation coefficient, E n (t) is the standardized error after exponential mapping, [w max - w b (t)] represents the adjustable weight space at the current moment.

[0015] As a preferred solution of the method for optimizing the positioning of photovoltaic access to the distribution network based on the improved PSO of the present invention, wherein: the inertia weight function is applied to the speed update process of the particle swarm optimization algorithm. The speed update considers three factors: the current speed of the particle, the individual optimal position, and the global optimal position. By comprehensively adjusting the influence degrees of these three factors through the inertia weight, it is expressed as:

[0016] v i (t + 1) = w(t) × v i (t) + c 1 r 1 [p i - x i (t)] + c 2 r 2 [g - x i (t)]

[0017] Where, v i (t) is the speed of the i-th particle at time t, v i (t + 1) is the speed of the i-th particle at time t + 1, x i (t) is the position of the i-th particle at time t, p i is the historical optimal position of the i-th particle, g is the group historical optimal position, c 1 , c 2 is the learning factor, r 1 , r 2 is a random number in the range of [0, 1].

[0018] As a preferred embodiment of the method for optimizing the positioning of photovoltaic access in a distribution network based on the improved PSO of the present invention, the following steps are included: Based on the inertia weight function, a dynamic evaluation model is established in combination with the characteristics of photovoltaic output. The dynamic evaluation model uses the basic weight term w b (t) and the error compensation term w c (t) of the inertia weight function as benchmark parameters; By setting the sampling time window for the characteristics of photovoltaic output changes, the change trend coefficient, fluctuation amplitude coefficient, and prediction error coefficient of photovoltaic output are extracted within each time window, and these three coefficients are weighted and combined with the benchmark parameters to construct a multi-dimensional evaluation matrix; The change trend coefficient is obtained by calculating the difference sequence of photovoltaic output at adjacent moments within the time window and normalizing this sequence to obtain a trend coefficient reflecting the change direction and rate of output; The fluctuation amplitude coefficient is based on the maximum value, minimum value, and average value of photovoltaic output within the time window to construct an index reflecting the severity of output fluctuations; The prediction error coefficient is calculated using the deviation between the actual value and the predicted value of photovoltaic output within the time window, and the root mean square error is calculated and normalized to obtain the prediction error coefficient; The row vectors of the multi-dimensional evaluation matrix represent the characteristics of photovoltaic output in different time windows, and the column vectors correspond to different evaluation dimensions of photovoltaic output. The matrix element values reflect the intensity of the corresponding characteristics in the corresponding dimensions.

[0019] As a preferred embodiment of the method for optimizing the positioning of photovoltaic access in a distribution network based on the improved PSO of the present invention, the following steps are included: Based on the dynamic evaluation model, an adaptive mapping relationship between the cognitive factor and the social factor is constructed. The cognitive factor increases linearly with the basic weight term w b (t), and the social factor decreases linearly with the error compensation term w c (t); The adaptive mapping relationship includes three levels, namely: the basic mapping layer, the dynamic adjustment layer, and the adaptive balance layer; The basic mapping layer establishes a positive correlation mapping between the cognitive factor and the basic weight term w b (t), and a negative correlation mapping between the social factor and the error compensation term w c (t); The dynamic adjustment layer corrects the basic mapping by introducing the change trend coefficient and the fluctuation amplitude coefficient in the evaluation matrix; The adaptive balance layer is based on the prediction error coefficient to establish a dynamic balance mechanism between the cognitive factor and the social factor; Through the three-layer mapping relationship, the collaborative adaptive adjustment of the cognitive factor and the social factor is realized, and then the cognitive factor and the social factor are substituted into the improved velocity-position update equation, expressed as:

[0020] v i (t + 1) = w(t)×v i (t) + c 1 (t)r 1 [p i -x i(t)]+c 2 (t)r 2 [g-x i (t)]

[0021] Among them, v i (t + 1) is the particle velocity of the i-th particle at the moment of t + 1, c 1 (t) is the cognitive factor, c 2 (t) is the social factor.

[0022] As a preferred scheme of the method for optimizing the positioning of photovoltaic access to the distribution network based on the improved PSO according to the present invention, wherein: based on the photovoltaic output calculation model, calculate the dynamic change characteristics of the photovoltaic output, introduce the standardized photovoltaic output, the prediction error of the photovoltaic output and its first derivative and second derivative into the dynamic characteristic evaluation model, comprehensively consider the magnitude, change speed and acceleration of the prediction error, and establish a photovoltaic output fluctuation sensitivity index, expressed as:

[0023] S(t) = α 1 e(t) + α 2 e'(t) + α 3 e”(t)

[0024] Among them, α 1 、α 2 、α 3 are sensitivity weighting coefficients, used to balance the influence degree of each order of errors, e(t) is the prediction error of the photovoltaic output, e'(t) is the first derivative of the prediction error of the photovoltaic output, and e”(t) is the second derivative of the prediction error of the photovoltaic output.

[0025] As a preferred scheme of the method for optimizing the positioning of photovoltaic access to the distribution network based on the improved PSO according to the present invention, wherein: based on the photovoltaic output fluctuation sensitivity index, combine the inertia weight function to construct a network loss index function, a voltage deviation index function and a photovoltaic benefit index function; weight each index function by a basic weight term in different time periods, and introduce a fluctuation sensitivity adjustment coefficient to establish a multi-objective evaluation system; based on the multi-objective evaluation system, introduce an error compensation term to construct a regularization evaluation function F(x), expressed as:

[0026]

[0027] Among them, ω 1 is the weight of the network loss index, ω 2 is the weight of the voltage deviation index, ω 3 is the weight of the photovoltaic benefit index, f 1 (x) is the network loss index function, f 2 (x) is the voltage deviation index function, f 3 (x) is the photovoltaic benefit index function, λp is the penalty factor, w d (t) is the inertia weight of constraint violation, T is the evaluation period, P c (x, t) is the degree of constraint violation, indicating the violation of the distribution network operation constraints at time t and access location x.

[0028] As a preferred solution of the distribution network photovoltaic access optimization positioning method based on the improved PSO described in the present invention, wherein: candidate access locations that satisfy all operation constraints are output based on the regularization evaluation function; the regularization evaluation value is calculated for the candidate access locations, the evaluation value is used as the dominant term of the particle fitness, and the degree of constraint violation is used as the penalty term of the fitness; initial particle positions are randomly generated within the system accessible area, the initial particle velocity range is set according to the maximum power output capacity of the system, and the initial individual optimal position and the global optimal position of each particle are recorded; iteration is performed based on the improved velocity-position update equation, expressed as:

[0029] v i (t + 1) = w(t) × v i (t) + c 1 (t)r 1 [p i - x i (t)] + c 2 (t)r 2 [g - x i (t)]

[0030] Calculate the operation constraint status of the new position, check the degree of constraint violation, and update the individual optimal position and the global optimal position; if the constraint violation of the new position decreases, directly update; if the constraint violation of the new position increases, judge whether to update based on the acceptance probability; when the constraints are satisfied, judge whether to update based on the regularization evaluation value; continuously iterate until the maximum number of iterations is reached, or the improvement amplitude of the evaluation value is less than the threshold for consecutive multiple iterations, or the number of iterations when the global optimal position remains stable reaches the set value; output the optimal access position that satisfies all operation constraints.

[0031] As a preferred solution of the method for optimizing the positioning of photovoltaic access to the distribution network based on the improved PSO of the present invention, the following is included: the regularization evaluation function introduces a penalty term to ensure that the optimization result meets the various operating constraints of the distribution network and evaluates the situation of constraint violation, including: if at time t, the node voltage exceeds the allowable upper and lower limit ranges at the access position x, the voltage violation degree is calculated according to the degree of voltage over-limit. The greater the voltage violation degree, the greater the voltage penalty value generated; if at the same time the power of a certain line exceeds its maximum capacity limit, the power violation degree is calculated according to the degree of power over-limit. The greater the power violation degree, the greater the power penalty value generated; if there is a deviation between the actual value and the predicted value of the photovoltaic output during operation, the corresponding error compensation is introduced, and the inertia weight will increase as the prediction error increases; if during the optimization process, the current photovoltaic access position x 1 has a node voltage exceeding the system allowable limit, the load rate of some lines exceeding the rated capacity, and the node power imbalance exceeding the system requirements, the search algorithm proposes a new candidate access position x 2 , and calculates the violation penalty value P 1 of the position x 1 based on the operating constraints, and calculates the violation penalty value P 2 of the position x 2 based on the same constraints; through the violation value difference ΔP = P 1 -P 2 judgment: if the new position reduces the constraint violation, that is, ΔP < 0, then accept the new position x 2 as the current solution, record the updated constraint violation status, increase the search step size to continue the optimization, and update the optimal solution record during the iteration process; if the new position increases the constraint violation, that is, ΔP > 0, then calculate the acceptance probability based on the current iteration temperature, generate a random number for probability judgment; if the probability test is passed, accept the new position, if not, keep the current position, reduce the search step size for fine search, and appropriately shrink the search space range.

[0032] Second aspect, an embodiment of the present invention provides an optimized positioning system for photovoltaic access in a distribution network based on improved PSO, which includes: a coupling optimization module, configured to obtain irradiance time series data and temperature field distribution data, construct a photovoltaic output calculation model, and map the output of the calculation model to the inertia weight function of the particle swarm optimization algorithm to realize the coupling of physical characteristics and the optimization algorithm; a dynamic optimization module, configured to establish a dynamic evaluation model according to the inertia weight function combined with the dynamic characteristics of photovoltaic output, set the adaptive adjustment rules of the cognitive factor and the social factor, and input them into the improved velocity-position update equation to realize the dynamic optimization of the particle swarm search characteristics; an access optimization module, configured to establish a multi-objective evaluation system by integrating network loss indicators, voltage deviation indicators, and photovoltaic benefit indicators, construct a regularization evaluation function, introduce a penalty term based on power flow constraints to meet the operation constraints of the distribution network, and determine the optimal photovoltaic access position that meets the steady-state operation requirements of the distribution network through iterative optimization in combination with the improved particle swarm optimization algorithm.

[0033] The beneficial effects of the present invention are as follows: The present invention uses the inertia weight function to adjust the search behavior of the particle swarm optimization algorithm, realizes the dynamic balance of the global search and local development capabilities during the optimization process, and avoids the defect that the basic particle swarm optimization algorithm is prone to fall into local optimum. The improved velocity-position update equation incorporates time-varying inertia weight, cognitive factor, and social factor, enabling the optimization process to adaptively adjust the search strategy according to the search requirements at different stages, ensuring the convergence efficiency of the algorithm. The present invention can not only effectively determine the optimal photovoltaic access position that meets the system operation constraints, but also provides an important reference basis for the optimization and adjustment of the system operation mode through the quantitative evaluation of the constraint indicators, which has important practical guiding significance for improving the photovoltaic accommodation capacity of the distribution network and ensuring the safe and stable operation of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings according to these drawings without creative efforts. Among them:

[0035] Figure 1 It is a flowchart of an optimized positioning method for photovoltaic access in a distribution network based on improved PSO. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0036] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following provides a detailed description of the specific embodiments of the present invention in conjunction with the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0037] In the following description, many specific details are set forth to facilitate a thorough understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0038] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that mutually excludes other embodiments.

[0039] Embodiment 1

[0040] Referring to Figure 1 , this is the first embodiment of the present invention. This embodiment provides an optimized positioning method for photovoltaic access in a distribution network based on improved PSO, including:

[0041] S1: Based on the photovoltaic conversion mechanism of photovoltaic modules, irradiance time series data and temperature field distribution data are obtained, a photovoltaic output calculation model is constructed, and the output of the calculation model is mapped into the inertia weight function of the particle swarm optimization algorithm to realize the coupling of physical characteristics and the optimization algorithm.

[0042] Historical light intensity data and ambient temperature data with a 24-hour resolution in the target distribution network area are collected, the original data is preprocessed, and outliers are removed through the 3σ criterion to obtain the light intensity sequence G(t) and the temperature sequence T(t).

[0043] The preprocessing includes time alignment of the collected original data to ensure that the data at different sampling points have the same time tag:

[0044] For missing data with less than 3 consecutive time points, linear interpolation is used to supplement it;

[0045] For missing data with more than 3 consecutive time points, historical data of the same period is used for replacement.

[0046] The data after removing outliers are sorted by time. For each time point t, the average value of all sampling points is taken as the value of G(t), forming an illumination intensity sequence with a 24-hour resolution. The illumination intensity sequence records the illumination intensity values at each time point.

[0047] For each time point t, the average value of all sampling points is taken as the value of T(t), forming a temperature sequence with a 24-hour resolution. The temperature sequence records the ambient temperature values at the corresponding time points.

[0048] Taking the two sequences as the basic input data, a photovoltaic output calculation model is established.

[0049] The photovoltaic output calculation model receives the illumination intensity sequence and the temperature sequence as inputs and calculates the theoretical photovoltaic output power. The calculation process takes into account key factors such as the photoelectric conversion efficiency, the area of the photovoltaic panel, and the influence of temperature on the efficiency, and outputs a theoretical photovoltaic power sequence, expressed as:

[0050] P 0 (t) = η × S × G(t) × [1 - γ(T(t) - T ref )]

[0051] Among them, P 0 (t) represents the theoretical photovoltaic output power at time t, η represents the photoelectric conversion efficiency, S represents the effective light-receiving area of the photovoltaic array, γ represents the sensitivity coefficient of the output power of the photovoltaic cell to temperature change, and T ref represents the reference temperature under standard test conditions.

[0052] Based on the environmental parameters, the theoretical output power of the photovoltaic system is calculated through the photovoltaic output calculation model, and the calculation result reflects the basic physical characteristics of photovoltaic power generation.

[0053] A BP neural network prediction model is constructed, adopting a double hidden layer structure. Taking the illumination intensity sequence G(t), the temperature sequence T(t), and the theoretical output power of the photovoltaic system as inputs, the actual photovoltaic output is predicted, and a predicted power sequence is output. By comparing the predicted value with the actual value, the relative prediction error is calculated.

[0054] Specifically, if the change in the illumination intensity sequence G(t) causes a change in the theoretical power P 0 (t), then the predicted power P p (t) output by the neural network is collected, the actually observed photovoltaic output power P r (t) is obtained, and the relative error E(t) of the neural network is calculated, expressed as:

[0055]

[0056] Adjust the weight parameters of the neural network according to the magnitude of the relative error E(t), and the adjustment rule is expressed as:

[0057] Multiply the relative error, the learning rate, and the current input quantity to obtain the magnitude of the adjustment value required for the weight. Subtract the calculated adjustment amount from the current weight to obtain the new weight value at the next moment.

[0058] Starting from the output layer, use the relative error E(t) as the initial error signal and backpropagate it to each hidden layer to calculate the error value corresponding to the hidden layer nodes.

[0059] If the output error of the calculated hidden layer node is at a certain node j in the kth hidden layer of the neural network, first calculate the error gradient of this node, specifically by multiplying the output error by the derivative of the activation function of this node; multiply the error gradient of this node, the learning rate, and the output value of the previous layer node to obtain the weight adjustment amount; add the calculated adjustment amount to the current weight to obtain the new weight value of this connection at the next moment.

[0060] Taking the update of the second hidden layer as an example: The relative error signal starts from the output layer, first passes to the last hidden layer, and then continues to pass to the second hidden layer. For each node in the second hidden layer, calculate the error gradient and update the weight until the weight update of all layers is completed.

[0061] Realize the optimal adjustment of all weights of the entire neural network, and apply the adjusted weight parameters to the power prediction at the next moment.

[0062] Standardize the predicted power sequence of the BP neural network prediction model. Adopt the maximum-minimum normalization method to find the maximum and minimum values in the predicted power sequence, map each predicted power value proportionally to the interval from 0 to 1, and maintain the relative relationship of the output change through normalization.

[0063] Adopt the exponential mapping method for the prediction error sequence, map each moment's relative error value through the exponential function, and introduce the error sensitivity coefficient α to adjust the influence degree of the error.

[0064] When α > 1, the exponential mapping will amplify the original error value;

[0065] When α < 1, the exponential mapping will compress the original error value;

[0066] When α = 1, maintain the linear relationship of the error.

[0067] Set the error threshold according to the requirements of power system dispatching and wind farm operation, and perform non-linear mapping on the relative error through the exponential function:

[0068] When the relative error is large, that is, when the relative error exceeds the set error threshold, a larger weight value is generated after exponential mapping;

[0069] When the error is small, that is, when the relative error approaches 0, the change in the weight value after exponential mapping is relatively gentle.

[0070] Through normalization, data with different dimensions are unified, and through non-linear mapping, the influence degree of errors of different magnitudes is strengthened, realizing the adjustability of error response and enhancing the sensitivity of the model to prediction quality.

[0071] Based on the normalized predicted power sequence and error sequence, a two-layer weight function is constructed to calculate the inertia weight value w(t) in the particle swarm optimization algorithm, which is expressed as:

[0072] w(t) = w b (t) + w c (t)

[0073] w b (t) = w max - (w max - w min ) × P n (t)

[0074] w c (t) = -β × E n (t) × [w max - w b (t)]

[0075] Among them, w(t) is the inertia weight value of the particle swarm optimization algorithm at time t, w b (t) is the basic weight term, w c (t) is the error compensation term, w max is the maximum weight value, w min is the minimum weight value, p n (t) is the predicted power sequence normalized by the maximum and minimum values, and the value range is [0, 1]. When p n (t) = 0, w b (t) reaches the maximum value w max ; when p n (t) = 1, w b (t) reaches the minimum value w min ; β is the error compensation coefficient, controlling the intensity of the error compensation term, E n (t) is the normalized error after exponential mapping, [w max - w b (t)] represents the adjustable weight space at the current moment.

[0076] Construct a double-layer inertia weight function with error adaptability, where the basic weight term changes linearly with the normalized output value to ensure the basic dynamic characteristic tracking ability; the error compensation term is driven by the normalized error. When the prediction error is large, the weight value is increased through a negative feedback mechanism to enhance the global search ability of the algorithm.

[0077] Apply the inertia weight function to the velocity update process of the particle swarm optimization algorithm. The velocity update considers three factors: the current velocity of the particle, the individual best position, and the global best position. The influence degrees of these three factors are dynamically adjusted through the comprehensive inertia weight, which is expressed as:

[0078] v i (t + 1) = w(t) × v i (t) + c 1 r 1 [p i -x i (t)] + c 2 r 2 [g - x i (t)]

[0079] Among them, v i (t) is the velocity of the i-th particle at time t, v i (t + 1) is the velocity of the i-th particle at time t + 1, x i (t) is the position of the i-th particle at time t, p i is the historical best position of the i-th particle, g is the historical best position of the group, c 1 , c 2 is the learning factor, r 1 , r 2 is a random number in the interval [0, 1].

[0080] By introducing an error-driven compensation mechanism, the PSO algorithm can dynamically adjust the search strategy according to the prediction accuracy. When the prediction error is large, the exploration ability of the particle swarm is increased through the error compensation term, and the adaptability of the algorithm to the uncertainty of photovoltaic output is improved. At the same time, the basic weight term maintains the tracking ability of the time-series characteristics of photovoltaic output, and the two work together to improve the optimization performance.

[0081] S2: Based on the inertia weight function, establish a dynamic evaluation model in combination with the dynamic characteristics of photovoltaic output, set the adaptive adjustment rules for the cognitive factor and the social factor, and input them into the improved velocity-position update equation to realize the dynamic optimization of the particle swarm search characteristics.

[0082] According to the double-layer adaptive inertia weight function, establish a dynamic evaluation model in combination with the characteristics of photovoltaic output. The dynamic evaluation model combines the basic weight term w b (t) and the error compensation term wc (t) is used as a reference parameter.

[0083] Among them, the basic weight term w b (t) mainly reflects the fluctuation law and daily variation characteristics of photovoltaic output, and the error compensation term w c (t) compensates for the impact brought by the photovoltaic output prediction error e(t).

[0084] By setting the sampling time window of the photovoltaic output change characteristics, the change trend coefficient, fluctuation amplitude coefficient and prediction error coefficient of the photovoltaic output are extracted within each time window, and these three coefficients are weighted and combined with the reference parameter to construct a multi-dimensional evaluation matrix.

[0085] Among them, the change trend coefficient is obtained by calculating the difference sequence of the photovoltaic output at adjacent moments within the time window and normalizing this sequence to obtain a trend coefficient that reflects the change direction and rate of the output. A positive change trend coefficient indicates an upward output trend, a negative value indicates a downward trend, and the absolute value reflects the change rate;

[0086] The fluctuation amplitude coefficient is based on the maximum value, minimum value and average value of the photovoltaic output within the time window to construct an index that reflects the severity of the output fluctuation. By calculating the difference between the maximum value and the minimum value divided by the average value and normalizing it, the fluctuation amplitude coefficient is obtained. The larger the value of the fluctuation amplitude coefficient, the more severe the output fluctuation;

[0087] The prediction error coefficient uses the deviation between the actual value and the predicted value of the photovoltaic output within the time window to calculate the root mean square error and normalizes it to obtain the prediction error coefficient. The prediction error coefficient reflects the accuracy of the prediction model in the current time window.

[0088] The row vector of the multi-dimensional evaluation matrix represents the photovoltaic output characteristics of different time windows, and the column vector corresponds to different evaluation dimensions of the photovoltaic output. The matrix element value reflects the intensity of the corresponding characteristic in the corresponding dimension.

[0089] Through the multi-dimensional quantitative characterization of the dynamic characteristics of the photovoltaic output, the multi-dimensional evaluation matrix enables the optimization search process to adjust corresponding strategies as the photovoltaic output changes.

[0090] Based on the dynamic evaluation model, an adaptive mapping relationship between the cognitive factor and the social factor is constructed. The cognitive factor increases linearly with the basic weight term w b (t), and the social factor decreases linearly with the error compensation term w c (t). Through this mapping relationship, the group search ability is enhanced when the photovoltaic output is large, and the individual exploration ability is improved when the output is small, realizing the dynamic balance of the search strategy.

[0091] Specifically, the adaptive mapping relationship includes three levels, namely: the basic mapping layer, the dynamic adjustment layer, and the adaptive balance layer;

[0092] The basic mapping layer establishes a positive correlation mapping between cognitive factors and the basic weight term w b (t), and a negative correlation mapping between social factors and the error compensation term w c (t). Three operating intervals of high, medium, and low photovoltaic power output are set. In the low power output interval, the cognitive factor reaches the maximum value to enhance the individual search ability; in the high power output interval, the social factor reaches the maximum value to strengthen the group information interaction; in the middle interval, a smooth transition is achieved through a piecewise function.

[0093] The dynamic adjustment layer corrects the basic mapping by introducing the change trend coefficient and the fluctuation amplitude coefficient in the evaluation matrix. A hierarchical response mechanism is established based on the photovoltaic power output change rate threshold. When the change trend coefficient indicates a rapid increase in power output, the social factor is dynamically adjusted according to the change rate to accelerate convergence; when the fluctuation amplitude coefficient exceeds the set threshold, the cognitive factor is correspondingly increased according to the fluctuation intensity to maintain search diversity;

[0094] The adaptive balance layer is based on the prediction error coefficient to establish a dynamic balance mechanism between the cognitive factor and the social factor. By setting multiple thresholds of the prediction error, a piecewise adjustment function of the factor ratio is constructed. In the high error interval, the cognitive factor dominates to enhance the anti-interference ability of the algorithm; in the low error interval, the social factor dominates to accelerate the convergence speed.

[0095] Through the three-layer mapping relationship, the collaborative adaptive adjustment of the cognitive factor and the social factor is realized, ensuring the search efficiency of the algorithm under different photovoltaic power output conditions and ensuring the stability and reliability of the search process.

[0096] Furthermore, the cognitive factor and the social factor are substituted into the improved velocity-position update equation, expressed as:

[0097] v i (t + 1) = w(t) × v i (t) + c 1 (t)r 1 [p i -x i (t)] + c 2 (t)r 2 [g - x i (t)]

[0098] Among them, v i (t + 1) is the particle velocity of the i-th particle at time t + 1, c 1 (t) is the cognitive factor, and c 2 (t) is the social factor.

[0099] The inertia of the particle is controlled by the inertia weight function w(t), and at the same time, the cognitive factor c 1 (t) and the social factor c 2 (t) are used to adjust the learning intensities of the particle for individual experience and group information respectively, realizing the collaborative adaptation of the three search dimensions of inertia, cognition, and society.

[0100] Furthermore, the inertia weight function w(t) is optimized to establish a non-linear mapping relationship between the weight function and the normalized photovoltaic output P n (t). It includes:

[0101] The basic weight term w b (t) is constructed by a piecewise exponential function to map with the normalized photovoltaic output, so that the weight remains a large value in the low output interval to enhance the global exploration ability and rapidly decays in the high output interval to improve the local precise search ability;

[0102] The error compensation term w c (t) constructs a dynamic response mechanism by introducing the first derivative e′(t) and the second derivative e″(t) of the photovoltaic output prediction error. When the prediction error shows an accelerating increasing trend, the compensation weight is rapidly increased through the second derivative term. When the prediction error shows a decreasing trend, the compensation weight is steadily reduced through the first derivative term, enabling the weight function to quickly adjust to the change trend of the output prediction error and improving the adaptability of the algorithm to the photovoltaic output fluctuation.

[0103] S3: Establish a multi-objective evaluation system by integrating the network loss index, voltage deviation index, and photovoltaic benefit index, and construct a regularization evaluation function. Introduce a penalty term based on the power flow constraint to meet the operation constraints of the distribution network. Combine the improved particle swarm optimization algorithm to determine the optimal photovoltaic access position that meets the steady-state operation requirements of the distribution network through iterative optimization.

[0104] By comprehensively considering the network operation efficiency, voltage quality, and photovoltaic power generation benefit, establish a multi-objective evaluation system, and perform dynamic adjustment in combination with the photovoltaic output fluctuation characteristics to finally determine the optimal photovoltaic access position.

[0105] Before establishing the multi-objective evaluation system, based on the constructed normalized photovoltaic output calculation model, further understand the dynamic change characteristics of the photovoltaic output. Introduce the normalized photovoltaic output, photovoltaic output prediction error and its first derivative, and second derivative into the dynamic characteristic evaluation model. By comprehensively considering the magnitude, change speed, and acceleration of the prediction error, establish a photovoltaic output fluctuation sensitivity index, expressed as:

[0106] S(t) = α 1 e(t) + α 2 e'(t) + α 3 e”(t)

[0107] Among them, α 1 , α 2 , α 3 are sensitivity weighting coefficients, used to balance the influence degree of each order error. e(t) is the photovoltaic output prediction error, e'(t) is the first derivative of the photovoltaic output prediction error, and e”(t) is the second derivative of the photovoltaic output prediction error.

[0108] The photovoltaic output fluctuation sensitivity index comprehensively reflects the dynamic characteristics of photovoltaic output through the combination of prediction error and its change rate.

[0109] Based on the photovoltaic output fluctuation sensitivity index, three evaluation index functions are constructed by combining with the inertia weight function w(t), which are respectively: the network loss index function f 1 (x), the voltage deviation index function f 2 (x), and the photovoltaic benefit index function f 3 (x); each index function is weighted by time periods using the basic weight term, and a fluctuation sensitivity adjustment coefficient is introduced to enable the evaluation index to be adaptively adjusted with the dynamic change of photovoltaic output.

[0110] The construction process of the network loss index function includes:

[0111] Introduce the basic weight term w b (t) for time-period weighting to reflect the importance of different time periods;

[0112] Introduce the fluctuation sensitivity adjustment term [1, β 1 S(t)] to enable the network loss index to be dynamically adjusted with the fluctuation of photovoltaic output;

[0113] Take the loss Pl(t) of the network at time t as the evaluation object;

[0114] Sum up the weighted network losses within the entire evaluation period T to obtain the network loss index function.

[0115] The construction process of the voltage deviation index function includes:

[0116] Use the basic weight term w b (t) for time-period weighting;

[0117] Introduce the fluctuation sensitivity adjustment term [1, β 2 S(t)] to dynamically adjust the weighted result with the fluctuation of photovoltaic output;

[0118] Calculate the square difference between the voltage V i (t) of each node and the rated voltage V n to reflect the degree of voltage deviation;

[0119] Accumulate the voltage deviations of all nodes;

[0120] Perform weighted summation within the evaluation period T to obtain the voltage deviation index function.

[0121] The construction process of the photovoltaic benefit index function includes:

[0122] Use the basic weight term w b (t) for weighted time segmentation;

[0123] Introduce the fluctuation sensitivity adjustment term [1, β 3 S(t)], and dynamically adjust the weighted result with the fluctuation of photovoltaic output;

[0124] Take the normalized photovoltaic output P n (t) as the evaluation object;

[0125] Perform weighted summation within the evaluation period T to obtain the photovoltaic benefit index function.

[0126] Among them, β 1 、β 2 、β 3 are the fluctuation sensitivity adjustment coefficients, reflecting the sensitivity of different indicators to photovoltaic fluctuations.

[0127] The network loss index function considers the network loss conditions in each period, the voltage deviation index function measures the deviation degree of the node voltage from the rated value, and the photovoltaic benefit index function evaluates the economy of photovoltaic power generation.

[0128] Based on the network loss index function, the voltage deviation index function and the photovoltaic benefit index function, construct a multi-objective evaluation system including network loss, voltage deviation and photovoltaic benefit.

[0129] To ensure that the optimization result meets the operation constraints of the distribution network, on the basis of the multi-objective evaluation system, introduce an error compensation term to construct a regularization evaluation function F(x), expressed as:

[0130]

[0131] Among them, ω 1 is the weight of the network loss index, ω 2 is the weight of the voltage deviation index, ω 3 is the weight of the photovoltaic benefit index, f 1 (x) is the network loss index function, f 2 (x) is the voltage deviation index function, f 3 (x) is the photovoltaic benefit index function, λ p is the penalty factor, w d (t) is the inertia weight of constraint violation, T is the evaluation period, P c(x, t) represents the degree of constraint violation, indicating the violation of the distribution network operation constraints at time t and connection location x.

[0132] The distribution network operation constraints include node voltage constraints, line capacity constraints, power balance constraints, and power flow constraints.

[0133] By introducing penalty terms, it is ensured that the optimization results meet the various operation constraints of the distribution network, and the situation of constraint violation is evaluated, including:

[0134] If at time t, the node voltage exceeds the allowable upper and lower limits at connection location x, the voltage violation degree is calculated according to the degree of voltage over-limit. The greater the voltage violation degree, the greater the voltage penalty value generated;

[0135] If at the same time the power of a certain line exceeds its maximum capacity limit, the power violation degree is calculated according to the degree of power over-limit. The greater the power violation degree, the greater the power penalty value generated;

[0136] If there is a deviation between the actual value and the predicted value of the PV output during operation, the corresponding error compensation is introduced, and the inertia weight will increase with the increase of the prediction error;

[0137] If during the optimization process, the current PV connection location x 1 has a node voltage exceeding the system allowable limit, the load rate of some lines exceeding the rated capacity, and the node power imbalance exceeding the system requirements, the search algorithm proposes a new candidate connection location x 2 , and calculates the violation penalty value P 1 of location x 1 based on the operation constraints, and calculates the violation penalty value P 2 of location x 2 ;

[0138] Through the violation value difference ΔP = P 1 - P 2 judgment:

[0139] If the new location reduces the constraint violation, that is, ΔP < 0, then accept the new location x 2 as the current solution, record the updated constraint violation status, increase the search step size to continue the optimization, and update the optimal solution record during the iteration process;

[0140] If the new location increases the constraint violation, that is, ΔP > 0, then calculate the acceptance probability based on the current iteration temperature, generate a random number for probability judgment; if the probability test is passed, accept the new location, if not, keep the current location, reduce the search step size for fine search, and appropriately shrink the search space range.

[0141] Further, the calculation of the acceptance probability takes into account the influence of the constraint violation value difference, the influence of the temperature parameter, and the decreasing law of the temperature parameter. During the calculation process, by obtaining the difference between the constraint violation values, dividing it by the current temperature value, taking the negative sign, and then calculating the exponential function, a probability value between 0 and 1 is obtained; the greater the difference in violation values, the smaller the acceptance probability, and the deeper the iteration, the smaller the acceptance probability.

[0142] Further, the decreasing law of the temperature parameter is expressed as: setting the initial temperature to a relatively large value, and multiplying it by a decay coefficient less than 1 after each iteration, so that the temperature value continuously decreases with the iteration process, resulting in a generally lower acceptance probability in the later stage.

[0143] Continuously perform iterative optimization calculations until the maximum number of iterations is reached, or there is no improvement after consecutive multiple searches, or the constraint violation value is lower than the termination threshold, or the search step size is less than the accuracy requirement.

[0144] Finally, output the access position that satisfies all operating constraints, the corresponding constraint violation status, and the iteration records of the convergence process.

[0145] Thus, the search process continuously escapes from the high constraint violation area, avoids local optima through the probability acceptance mechanism, dynamically adjusts the search step size to balance globality and locality, and ensures that it finally converges to a position that satisfies all constraints.

[0146] Calculate the regularization evaluation value for the candidate access positions that satisfy all operating constraints. The evaluation value serves as the dominant term of the particle fitness, and the degree of constraint violation serves as the penalty term for the fitness.

[0147] Randomly generate the initial particle positions within the system accessible area, set the initial particle velocity range according to the maximum power output capacity of the system, and record the initial individual optimal positions and the global optimal positions of each particle;

[0148] Perform iterations according to the improved velocity-position update equation, which is expressed as:

[0149] v i (t + 1) = w(t) × v i (t) + c 1 (t)r 1 [p i -x i (t)] + c 2 (t)r 2 [g - x i (t)]

[0150] Calculate the operating constraint status of the new position, check the degree of constraint violation, update the individual optimal position and the global optimal position. If the constraint violation of the new position decreases, directly update; if the constraint violation of the new position increases, judge whether to update based on the acceptance probability.

[0151] When the constraint is satisfied, it is judged whether to update based on the regularized evaluation value.

[0152] Iterate continuously until the maximum number of iterations is reached, or the improvement amplitude of the evaluation value in consecutive multiple iterations is less than the threshold, or the number of iterations when the global optimal position of the population remains stable reaches the set value.

[0153] Output the optimal access position that satisfies all operation constraints, the corresponding regularized evaluation function value, the specific values of each constraint index, and the iteration record of the optimization convergence process.

[0154] In summary, the present invention uses an inertia weight function to adjust the search behavior of the particle swarm optimization algorithm, realizes the dynamic balance between global search and local development capabilities during the optimization process, and avoids the defect that the basic particle swarm optimization algorithm is prone to falling into local optima. The regularized evaluation function is used as the fitness function of the improved particle swarm optimization algorithm. Through the dual judgment mechanism of the constraint violation degree and the acceptance probability, it is ensured that the search process continuously escapes from the high constraint violation area while not completely rejecting the sub-optimal solution, enhancing the ability of the algorithm to find the global optimal solution. The improved velocity-position update equation incorporates time-varying inertia weight, cognitive factor, and social factor, enabling the optimization process to adaptively adjust the search strategy according to the search requirements at different stages, ensuring the convergence efficiency of the algorithm. The present invention can not only effectively determine the optimal PV access position that satisfies the system operation constraints, but also provide an important reference basis for the optimization and adjustment of the system operation mode through the quantitative evaluation of the constraint indicators, and has important practical guiding significance for improving the PV accommodation capacity of the distribution network and ensuring the safe and stable operation of the system.

[0155] This embodiment further provides a distribution network PV access optimization and positioning system based on the improved PSO, including:

[0156] A coupling optimization module, configured to obtain irradiance time-series data and temperature field distribution data, construct a PV output calculation model, and map the output quantity of the calculation model into the inertia weight function of the particle swarm optimization algorithm to realize the coupling of physical characteristics and the optimization algorithm;

[0157] A dynamic optimization module, configured to establish a dynamic evaluation model according to the inertia weight function in combination with the dynamic characteristics of PV output, set the adaptive adjustment rules of the cognitive factor and the social factor, and input them into the improved velocity-position update equation to realize the dynamic optimization of the particle swarm search characteristics;

[0158] An access optimization module, configured to establish a multi-objective evaluation system by integrating network loss index, voltage deviation index, and PV benefit index, construct a regularized evaluation function, introduce a penalty term based on power flow constraints to meet the distribution network operation constraints, and determine the optimal PV access position that meets the steady-state operation requirements of the distribution network through iterative optimization in combination with the improved particle swarm optimization algorithm.

[0159] This embodiment also provides a computer device, which is applicable to the situation of the optimized positioning method for photovoltaic access in a distribution network based on the improved PSO. It includes a memory and a processor. The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the optimized positioning method for photovoltaic access in a distribution network based on the improved PSO as proposed in the above embodiment.

[0160] This computer device can be a terminal. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of this computer device is used to communicate with an external terminal in a wired or wireless manner. The wireless manner can be achieved through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of this computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of this computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the shell of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0161] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the optimized positioning method for photovoltaic access in a distribution network based on the improved PSO as proposed in the above embodiment.

[0162] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. For technical details not described in detail in this embodiment, reference can be made to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0163] Embodiment 2

[0164] This embodiment provides an optimized positioning method for photovoltaic access in a distribution network based on the improved PSO. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0165] To verify the superiority of the improved adaptive weight PSO algorithm of the present invention, the IEEE 33-node standard distribution system is selected for testing. The rated voltage of the system is 12.66 kV, and the total load is 3.715 MW + 2.3 MVar.

[0166] To ensure the fairness of comparison, when comparing the algorithm of the present invention with traditional PSO and linearly decreasing weight PSO, the population size is uniformly set to 50, the maximum number of iterations is 200, the learning factors c1 = c2 = 2, the initial inertia weight is 0.9, and the termination inertia weight is 0.4.

[0167] During the optimization process, four key indicators are mainly investigated: node voltage deviation, network power loss, algorithm convergence performance, and solution stability.

[0168] The experimental results show that the node voltage deviation obtained by the algorithm of the present invention is 0.0156 p.u., which is reduced by 36.6% and 21.2% respectively compared with 0.0246 p.u. of traditional PSO and 0.0198 p.u. of linearly decreasing PSO; the network loss rate is reduced to 3.86%, which is reduced by 19.9% and 9.2% respectively compared with 4.82% of traditional PSO and 4.25% of linearly decreasing PSO.

[0169] From the perspective of the optimization process, the algorithm of the present invention can converge to the optimal solution in about 35 generations, while traditional PSO needs more than 80 generations to converge, and the convergence speed is increased by 56.3%. The standard deviation analysis of 30 independent runs shows that the solution of the algorithm of the present invention has better stability, and the standard deviation is only 0.0052, far better than 0.0145 of traditional PSO and 0.0098 of linearly decreasing PSO, indicating that the algorithm has strong robustness.

[0170] The above experimental results fully verify the superiority of the algorithm of the present invention in the optimization of photovoltaic access in the distribution network, which can not only obtain a better optimization scheme, but also has a faster convergence speed and better optimization stability.

[0171] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not restrictive. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for optimizing the photovoltaic access positioning in a distribution network based on improved PSO, characterized in that: include: Based on the photoelectric conversion mechanism of photovoltaic modules, the irradiance time series data and temperature field distribution data are obtained, and a photovoltaic output calculation model is constructed. The output of the calculation model is mapped to the inertia weight function of the particle swarm optimization algorithm to achieve the coupling of physical characteristics and optimization algorithm. Based on the inertia weight function and the dynamic characteristics of photovoltaic output, a dynamic evaluation model is established, and adaptive adjustment rules of cognitive factors and social factors are set, which are input into the improved speed and position update equation to achieve dynamic optimization of particle swarm search characteristics; A multi-objective evaluation system is established by integrating network loss indicators, voltage deviation indicators and photovoltaic benefit indicators, and a regularized evaluation function is constructed. A penalty term based on power flow constraints is introduced to meet the operation constraints of the distribution network. The improved particle swarm optimization algorithm is combined to determine the optimal photovoltaic access location that meets the steady-state operation requirements of the distribution network through iterative optimization.

2. The method for optimizing the photovoltaic access to the distribution network based on improved PSO as claimed in claim 1, characterized in that: The photovoltaic output calculation model receives the light intensity sequence and the temperature sequence as input and calculates the theoretical photovoltaic output power, which is expressed as: P0(t)=η×S×G(t)×[1-γ(T(t)-T ref )] Where P0(t) represents the theoretical photovoltaic output power at time t, η represents the photoelectric conversion efficiency, S represents the effective light-receiving area of ​​the photovoltaic array, γ represents the sensitivity coefficient of the photovoltaic cell output power to temperature changes, and T ref represents the reference temperature under standard test conditions, G(t) represents the light intensity sequence, and T(t) represents the temperature sequence; Based on the light intensity sequence, temperature sequence and theoretical photovoltaic output power, a BP neural network prediction model with a double hidden layer structure is constructed to predict the actual photovoltaic output, output a predicted power sequence, and calculate the relative prediction error by comparing the predicted value with the actual value; Based on the standardized predicted power sequence and error sequence, a double-layer weight function is constructed, the inertia weight value in the particle swarm optimization algorithm is calculated, and the inertia weight function is constructed, which is expressed as: w(t)=w b (t)+w c (t) w b (t)=w max -(w max -w min )×P n (t) w c (t)=-β×E n (t)×[w max -w b (t)] Among them, w(t) is the inertia weight value of the particle swarm optimization algorithm at time t, w b (t) is the basic weight term, w c (t) is the error compensation term, w max is the maximum weight value, w min is the minimum weight value, p n (t) is the predicted power sequence normalized by the maximum and minimum values, and its value range is [0,1]; β is the error compensation coefficient, E n (t) is the standardized error after exponential mapping, [w max -w b (t)] represents the adjustable weight space at the current moment.

3. The method for optimizing the photovoltaic access to the distribution network based on improved PSO as claimed in claim 2, characterized in that: The inertia weight function is applied to the speed update process of the particle swarm optimization algorithm. The speed update takes into account three factors: the current speed of the particle, the individual optimal position, and the global optimal position. The influence of these three factors is dynamically adjusted by the comprehensive inertia weight, which is expressed as: v i (t+1)=w(t)×v i (t)+c1r1[p i -x i (t)]+c2r2[g-x i (t)] Among them, v i (t) is the velocity of the ith particle at time t, v i (t+1) is the velocity of the ith particle at time t+1, x i (t) is the position of the ith particle at time t, p i is the historical optimal position of the ith particle, g is the historical optimal position of the group, c1 and c2 are learning factors, and r1 and r2 are random numbers in the interval [0,1].

4. The method for optimizing the photovoltaic access to the distribution network based on improved PSO as claimed in claim 3 is characterized by: Based on the inertia weight function, a dynamic evaluation model is established in combination with photovoltaic output characteristics. The dynamic evaluation model converts the basic weight term w of the inertia weight function into b (t) and the error compensation term w c (t) as a benchmark parameter; By setting the sampling time window of the photovoltaic output change characteristics, the change trend coefficient, fluctuation amplitude coefficient and prediction error coefficient of the photovoltaic output are extracted in each time window, and these three coefficients are weightedly combined with the benchmark parameters to construct a multi-dimensional evaluation matrix. The change trend coefficient is obtained by calculating the difference sequence of photovoltaic power output at adjacent moments in the time window and normalizing the sequence to obtain the trend coefficient reflecting the direction and rate of power output change; The fluctuation amplitude coefficient is an indicator reflecting the severity of the power output fluctuation based on the maximum, minimum and average values ​​of the photovoltaic power output within the time window; The prediction error coefficient is obtained by calculating the root mean square error using the deviation between the actual value and the predicted value of the photovoltaic output in the time window, and performing normalization processing; The row vectors of the multidimensional evaluation matrix represent the photovoltaic output characteristics in different time windows, the column vectors correspond to different evaluation dimensions of the photovoltaic output, and the matrix element values ​​reflect the intensity of the corresponding characteristics in the corresponding dimensions.

5. The method for optimizing the photovoltaic access to the distribution network based on improved PSO as claimed in claim 4, characterized in that: Based on the dynamic evaluation model, an adaptive mapping relationship between cognitive factors and social factors is constructed, and the cognitive factor is proportional to the basic weight term w. b (t) increases linearly, and the social factor increases with the error compensation term w c (t) linearly decreases; The adaptive mapping relationship includes three levels: basic mapping layer, dynamic adjustment layer and adaptive balancing layer; The basic mapping layer combines the cognitive factor with the basic weight term w b (t) Establish a positive correlation mapping between social factors and error compensation term w c (t) establishing a negative correlation mapping; The dynamic adjustment layer modifies the basic mapping by introducing the change trend coefficient and the fluctuation amplitude coefficient in the evaluation matrix; The adaptive balancing layer establishes a dynamic balancing mechanism between cognitive factors and social factors based on the prediction error coefficient; Through the three-layer mapping relationship, the coordinated adaptive adjustment of cognitive factors and social factors is realized, and then the cognitive factors and social factors are substituted into the improved speed and position update equation, which is expressed as: v i (t+1)=w(t)×v i (t)+c1(t)r1[p i -x i (t)]+c2(t)r2[g-x i (t)] Among them, v i (t+1) is the particle velocity of the i-th particle at time t+1, c1(t) is the cognitive factor, and c2(t) is the social factor.

6. The method for optimizing the photovoltaic access to the distribution network based on improved PSO as claimed in claim 2, characterized in that: Based on the photovoltaic output calculation model, the dynamic change characteristics of photovoltaic output are calculated, and the standardized photovoltaic output, photovoltaic output prediction error and its first-order derivative and second-order derivative are introduced into the dynamic characteristic evaluation model. The size, change speed and acceleration of the prediction error are comprehensively considered to establish the photovoltaic output fluctuation sensitivity index, which is expressed as: S(t)=α1e(t)+α2e'(t)+α3e”(t) Among them, α1, α2, and α3 are sensitivity weighting coefficients, which are used to balance the influence of each order error, e(t) is the photovoltaic output prediction error, e'(t) is the first-order derivative of the photovoltaic output prediction error, and e”(t) is the second-order derivative of the photovoltaic output prediction error.

7. The method for optimizing the photovoltaic access to the distribution network based on improved PSO as claimed in claim 6, characterized in that: Based on the photovoltaic output fluctuation sensitivity index, a network loss index function, a voltage deviation index function and a photovoltaic benefit index function are constructed in combination with an inertia weight function; Each indicator function is weighted by time period using basic weight items, and the fluctuation sensitivity adjustment coefficient is introduced to establish a multi-objective evaluation system. Based on the multi-objective evaluation system, an error compensation term is introduced to construct a regularized evaluation function F(x), which is expressed as: Among them, ω1 is the weight of the network loss index, ω2 is the weight of the voltage deviation index, ω3 is the weight of the photovoltaic benefit index, f1(x) is the network loss index function, f2(x) is the voltage deviation index function, f3(x) is the photovoltaic benefit index function, λ p is the penalty factor, w d (t) is the inertia weight of constraint violation, T is the evaluation period, P c (x, t) is the degree of constraint violation, which indicates the violation of the distribution network operation constraints at time t and access location x.

8. The method for optimizing the photovoltaic access to the distribution network based on improved PSO as claimed in claim 7, characterized in that: Outputting a candidate access position that satisfies all operation constraints based on the regularized evaluation function; Calculating a regularized evaluation value for the candidate access position, the evaluation value is used as a dominant term of the particle fitness, and the degree of constraint violation is used as a penalty term of the fitness; The initial particle position is randomly generated in the area accessible by the system, the initial particle speed range is set according to the maximum delivery capacity of the system, and the initial individual optimal position and group optimal position of each particle are recorded; Iteration is performed based on the improved velocity position update equation, which is expressed as: v i (t+1)=w(t)×v i (t)+c1(t)r1[p i -x i (t)]+c2(t)r2[g-x i (t)] Calculate the operating constraint status of the new position, check the degree of constraint violation, and update the individual optimal position and the group optimal position; If the constraint violation of the new position is reduced, update directly; If the new position constraint violation increases, determine whether to update based on the acceptance probability; When the constraints are satisfied, determine whether to update based on the regularized evaluation value; Continue to iterate until the maximum number of iterations is reached, or the improvement of the evaluation value for multiple consecutive iterations is less than the threshold, or the number of iterations in which the optimal position of the group remains stable reaches the set value; Output the optimal access location that satisfies all operational constraints.

9. The method for optimizing the photovoltaic access to the distribution network based on improved PSO as claimed in claim 7, characterized in that: The regularized evaluation function introduces a penalty term to ensure that the optimization result meets the various operation constraints of the distribution network and evaluates the violation of the constraints, including: If at time t, the node voltage at the access position x exceeds the allowable upper and lower limits, the voltage violation degree is calculated according to the voltage limit violation degree. The greater the voltage violation degree, the greater the voltage penalty value. If the power of a certain line exceeds its maximum capacity limit at the same time, the power violation is calculated according to the degree of power violation. The greater the power violation, the greater the power penalty value. If there is a deviation between the actual value of PV output and the predicted value during operation, the corresponding error compensation is introduced, and the inertia weight increases as the prediction error increases; If, during the optimization process, the node voltage at the current photovoltaic access location x1 exceeds the system allowable limit, the load rate of some lines exceeds the rated capacity, and the node power imbalance exceeds the system requirement, the search algorithm proposes a new candidate access location x2, and calculates the violation penalty value P1 of location x1 based on the operating constraints, and calculates the violation penalty value P2 of location x2 based on the same constraints; By violating the value difference ΔP = P1-P2, we can judge: If the new position reduces the constraint violation, that is, ΔP<0, then accept the new position x2 as the current solution, record the updated constraint violation status, increase the search step size to continue the optimization, and update the optimal solution record during the iteration process; If the new position increases the constraint violation, that is, ΔP>0, the acceptance probability is calculated based on the current iteration temperature, and a random number is generated for probability judgment; if the probability test passes, the new position is accepted; if it fails, the current position is maintained, the search step size is reduced for a refined search, and the search space range is appropriately narrowed.

10. A distribution network photovoltaic access optimization positioning system based on improved PSO, based on the distribution network photovoltaic access optimization positioning method based on improved PSO according to any one of claims 1 to 9, characterized in that: include: A coupling optimization module is used to obtain irradiance time series data and temperature field distribution data, build a photovoltaic output calculation model, map the output of the calculation model to the inertia weight function of the particle swarm optimization algorithm, and realize the coupling of physical characteristics and optimization algorithm; A dynamic optimization module is used to establish a dynamic evaluation model based on the inertia weight function combined with the dynamic characteristics of photovoltaic output, set adaptive adjustment rules for cognitive factors and social factors, input them into the improved speed and position update equation, and realize dynamic optimization of particle swarm search characteristics; The access optimization module is used to establish a multi-objective evaluation system based on comprehensive network loss indicators, voltage deviation indicators and photovoltaic benefit indicators, and to construct a regularized evaluation function. It introduces a penalty term based on power flow constraints to meet the operation constraints of the distribution network, and combines the improved particle swarm optimization algorithm to determine the optimal photovoltaic access position that meets the steady-state operation requirements of the distribution network through iterative optimization.

Images