Power distribution network source load joint portraying method and system based on photovoltaic historical data

Abstract

The invention discloses a power distribution network source-load joint portraying method and system based on photovoltaic historical data, and relates to the technical field of power distribution networks, and the method comprises the steps: taking the historical data of a photovoltaic generator set as a fuzzy soft set, and carrying out the prediction through employing the fuzzy soft set after the normalization processing, and obtaining the power generation prediction power of a future set time period; determining the initial model meeting the training ending condition as a load prediction model; inputting the actual operation data into a load prediction model for processing to obtain a predicted load in a future set time period; establishing a sensitivity matrix in the power distribution network by using the power generation prediction power and the prediction load, and analyzing a local optimal matching combination of a power supply and the load in the power distribution network based on the sensitivity matrix; the sensitivity matrix in the power distribution network is established based on the predicted power generation power and the load, so that the local optimal matching combination of the power supply and the load under the operation condition of the active power distribution network can be analyzed, the optimal distribution of the power flow is realized, and the stability and the economical efficiency of a power system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution networks, and more specifically, to a method and system for jointly depicting the power sources and loads of a distribution network based on photovoltaic historical data. Background Art

[0002] At present, due to its advantages such as low cost and zero pollution, photovoltaic power generation has become the most promising renewable energy power generation technology at the present stage. However, photovoltaic power generation is affected by various factors, especially the volatility of weather factors, which results in strong volatility and large intermittency of photovoltaic power generation output. As the main energy source in the future renewable energy field, accurately predicting the photovoltaic power generation is an important basis for realizing the grid connection of photovoltaic power generation and improving the consumption efficiency of photovoltaic power generation. The distribution and behavior of power sources and loads will greatly affect the operation behavior of the distribution network. The analysis of the distribution and behavior of power sources and loads plays an important role in aspects such as grid load forecasting, power consumption management, demand response, operation status analysis, and safety warning. However, in the existing technology, the photovoltaic power generation (source) and the load (load) cannot be effectively combined, which results in instability in power distribution of the distribution network and waste of more electric power resources. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for jointly depicting the power sources and loads of a distribution network based on photovoltaic historical data to solve the problems existing in the above background art.

[0004] The above technical purpose of the present invention is achieved through the following technical solutions: In the first aspect, the present application provides a method for jointly depicting the power sources and loads of a distribution network based on photovoltaic historical data, including the following specific steps: Taking the historical data of photovoltaic generator sets as a fuzzy soft set, performing normalization processing on the fuzzy soft set, and predicting the power generation prediction power for a future set time period by using the normalized fuzzy soft set; Training a preset initial model based on historical data until the initial model reaches a preset training end condition, and determining the initial model that reaches the training end condition as the load prediction model; Obtaining actual operation data, and inputting the actual operation data into the load prediction model for processing to obtain the predicted load for a future set time period; Using the power generation prediction power and the predicted load to establish a sensitivity matrix in the distribution network, and analyzing the locally optimal matching combination of power sources and loads in the distribution network based on the sensitivity matrix.

[0005] On the basis of the above technical solutions, the present invention can be further improved as follows.

[0006] Further, the above initial model is a three-layer BP perceptron neural network, including an input layer, an output layer, and a hidden layer, specifically: ; In the formula, represents the output of the output layer in the initial model, represents the number of nodes in the hidden layer, represents the transfer function of the output layer in the initial model, which is an identity function, K represents the number of historical data input to the initial model, represents the historical data input to the initial model k , represents the hidden layer weight, represents the output layer weight, represents the hidden layer bias, represents the output layer bias, and the superscript h represents the label of the hidden layer, tanh represents the hyperbolic tangent function.

[0007] Furthermore, the above training end condition is that the error function of the initial model does not exceed the threshold, and the error function is specifically: ; In the formula, represents the error function value, T represents the sample size in the historical data used to train the initial model, represents the i th training sample response variable in the historical data Y value, represents the conditional quantile i obtained through the explanatory variable of the th training sample , is an indicator function.

[0008] Furthermore, the model parameters of the above initial model are expressed as: , where: ; In the formula, represents the penalty parameter, represents the number of hidden layer nodes, K represents the number of historical data input to the initial model, is the quadratic norm, represents the hidden layer weight vector, T represents the sample size in the historical data used to train the initial model, represents the error function value of the initial model.

[0009] Furthermore, the optimal values ​​of the above penalty parameters and the number of hidden layer nodes are determined by the AIC criterion, specifically: ; In the formula, To choose different and hour The minimum error function under quantile, T represents the sample size of historical data used to train the initial model, K represents the amount of historical data input to the initial model, Represents the number of hidden layer nodes.

[0010] Furthermore, the above-mentioned fuzzy soft set prediction after normalization is used to obtain the predicted power generation power for the future set time period, specifically: The parameter values ​​used to predict the power generation prediction are determined by using the normalized fuzzy soft sets, and the power generation prediction is calculated based on the parameter values ​​and the surface temperature of the photovoltaic modules.

[0011] Furthermore, the above parameter values ​​include open circuit voltage, output current, maximum power point current and maximum power point voltage, and the parameter values ​​are determined by fuzzy soft sets, specifically: ; , , ; In the formula, represents the open circuit voltage, Represents the output current, represents the maximum power point voltage, represents the maximum power point current, A and k represents the diode characteristic factor and the Boltzmann constant, and represent the surface temperature, series internal resistance and parallel internal resistance of the photovoltaic module respectively. q is the unit charge, represents the diode saturation current, represents the current generated by light radiation, represents the total composite current, represents the total composite voltage, The fill factor or curve factor of the photovoltaic cell is and The area of ​​the rectangle formed by and The ratio of the areas of the curves they form; The surface temperature of the photovoltaic module is specifically: , where Indicates the surface temperature of the photovoltaic module, represents the ambient temperature, represents the coefficient, , represents the light intensity.

[0012] In a second aspect, the present application provides a distribution network source-load joint portrait system based on photovoltaic historical data, which is applied to a distribution network source-load joint portrait method according to any one of the first aspect, and includes: A first module, configured to use the historical data of the photovoltaic power generation unit as a fuzzy soft set, perform normalization processing on the fuzzy soft set, and predict the power generation prediction power for a future set time period by using the normalized fuzzy soft set; A second module, configured to train a preset initial model based on historical data until the initial model reaches a preset training end condition, and determine the initial model that reaches the training end condition as a load prediction model; A third module, configured to obtain actual operation data, and input the actual operation data into the load prediction model for processing to obtain a predicted load for a future set time period; A fourth module, configured to establish a sensitivity matrix in the distribution network by using the power generation prediction power and the predicted load, and analyze the locally optimal matching combination of the power source and the load in the distribution network based on the sensitivity matrix.

[0013] In a third aspect, the present application provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the method according to any one of the first aspect is implemented.

[0014] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium, and the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the method according to any one of the first aspect.

[0015] Compared with the prior art, the present invention has at least the following beneficial effects: In the present application, by processing the historical data of photovoltaic power generation, outliers and missing values in the data can be removed, improving the accuracy and reliability of the data, and providing a reliable data basis for subsequent power generation power prediction; among them, by predicting the load of the distribution network through the load prediction model, an accurate estimation of the future load can be achieved, thereby facilitating the distribution network to better grasp the load change during the dispatching and planning process, improving the stability and economy of the power system; establishing a sensitivity matrix in the distribution network based on the predicted power generation power and load helps to analyze the locally optimal matching combination of the power source and the load under the operating conditions of the active distribution network, realize the optimal distribution of the power flow, and improve the stability and economy of the power system. Description of the Drawings

[0016] The accompanying drawings described herein are used to provide a further understanding of the embodiments of the present invention, form a part of this application, and do not limit the embodiments of the present invention. In the drawings: Figure 1 is a flowchart of the image method in an embodiment of the present invention; Figure 2 is a schematic connection diagram of the image system in an embodiment of the present invention; Figure 3 is a schematic connection diagram of the electronic device in an embodiment of the present invention. Detailed Embodiments

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations.

[0018] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0019] It should be noted that: similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0020] In the description of the embodiments of the present invention, "a plurality" represents at least two.

[0021] Embodiment 1: In order to effectively combine photovoltaic power generation (source) with load (load) and avoid problems such as instability in power distribution of the distribution network and waste of more power resources, this embodiment provides a method for jointly imaging the source and load of a distribution network based on historical photovoltaic data, as Figure 1 shown, including the following specific steps: S1. Take the historical data of the photovoltaic generator set as a fuzzy soft set, perform normalization processing on the fuzzy soft set, and use the normalized fuzzy soft set to predict the power generation prediction power for a future set time period.

[0022] Among them, a significant feature of photovoltaic power generation is its intermittency and volatility. Since photovoltaic resources are affected by natural factors (parameters) such as weather and seasons, the historical data of photovoltaic power generation has the characteristics of time series, randomness, diversity, and large quantity. If these historical data are not processed, it is difficult to provide certain data support for photovoltaic power generation.

[0023] Specifically, the historical data includes data such as solar radiation, environmental temperature measurement range, battery charging current, and battery discharging current. These historical data are regarded as fuzzy soft sets, and the fuzzy soft sets are processed to normalize the fuzzy soft sets, and then the photovoltaic power generation is predicted according to the normalized fuzzy soft sets.

[0024] Specifically, a fuzzy soft set can be regarded as a parameterized family of fuzzy sets on a given universe of discourse. It is a fuzzy extension of a soft set and an object expression model with parameters as the framework and fuzzy information as the representation. It is no longer limited to the simple description of 0 / 1 object parameters by an exact soft set, but expresses the membership degree of parameters in a more flexible fuzzy form, so it can be applied to the uncertain situation of photovoltaic power generation in the power system. Among them, the forms of fuzzy information contained in the fuzzy soft sets are different, and the constructed decision models are also independent of each other. Obviously, in order to fuse mixed fuzzy information to support the solution of the power generation prediction problem and meet the real-time and efficient decision-making requirements, the following definition of a one-dimensional mixed fuzzy soft set is given in this embodiment: For the fuzzy set on the universe of discourse where , if there exist two elements and A such that the membership degrees of to A are expressed by different types of fuzzy numbers, then the fuzzy set A can be called a mixed fuzzy set; otherwise, is called a non-mixed fuzzy set; for example: ; in the formula, the membership degrees of the three elements and U are expressed by a general fuzzy number, an interval value, and a trapezoidal fuzzy number respectively. For the fuzzy soft set on the universe of discourse , if the types of fuzzy numbers describing the membership degrees of object parameters are not unique, then the fuzzy soft set

[0025] can be called a mixed fuzzy soft set. Furthermore, let U be the set of all non-mixed fuzzy subsets on the given universe of discourse E , be the parameter set, and U . For the mixed fuzzy soft set on the universe of discourseS Approximate function is a mapping with values in a non - hybrid fuzzy set, then is called a one - dimensional hybrid fuzzy soft set. Taking any parameter , then ; Specifically, a one - dimensional hybrid fuzzy soft set is regarded as a specific type of hybrid fuzzy soft set, characterized by different fuzzy membership degree representations between different parameters, while maintaining consistency within the same parameter. This characteristic is realistically and feasibly reflected in the power generation prediction operation system, that is, the representation forms between the attribute indicators of each subject are diverse, but the representations within the indicators are consistent.

[0026] Among them, after establishing the one - dimensional hybrid fuzzy soft set model, this embodiment further proposes a normalization method based on distance measure. When normalizing a one - dimensional hybrid fuzzy soft set, in order to maximize the retention of the degree information of the original fuzzy numbers, not only the magnitude relationship between the original parameter values needs to be maintained, but also the relative distance relationship between them needs to be transmitted. Since the essence of the parameter values is fuzzy membership degrees, in order to avoid mutual interference, direct comparison or mixed operations are not performed on different parameters, thus ensuring their independence; based on the above considerations, the basic strategy for normalizing the hybrid fuzzy soft set adopted in this embodiment is as follows: First, calculate the relative distances of various fuzzy numbers from 0 / 1, then use these distances to locate the fuzzy numbers within the interval, and finally convert them into general fuzzy numbers; as follows: Suppose is a one - dimensional hybrid fuzzy soft set on the given universe of discourse , the parameter set is is the expression matrix in tabular form, where is the parameter of the object value. Obviously is a hybrid fuzzy matrix. If is normalized to the expression matrix corresponding to the general fuzzy soft set is , then according to the normalization strategy, the corresponding relationship between the matrix and can be described as: ; In the formula, is the distance between the fuzzy number and 0, is the distance between the fuzzy number and 1.

[0027] Furthermore, since there are various types of fuzzy numbers and different distance measure methods, this embodiment takes interval - valued fuzzy sets (IVFS) as an example to demonstrate and generation process: Interval number is defined as a subset of the set of real numbers : . Therefore, can also be denoted as ; for two interval numbers , the distance between them is: .

[0028] Since , thus is an interval number. Let According to the above formula, the distance between and 1 and the distance between and 0 are respectively: .

[0029] Optionally, through the above fuzzy number conversion method, a general fuzzy matrix is obtained from the mixed fuzzy matrix , and thus the tabular form of the general fuzzy soft set is obtained, and the normalization of the mixed fuzzy soft set is completed. Subsequently, the power generation power of the photovoltaic can be predicted according to the normalized fuzzy soft set.

[0030] Furthermore, the principle of photovoltaic power generation is through the photovoltaic effect. A photovoltaic cell has a structure similar to a diode PN junction. When light irradiates on the cell, a voltage will be generated at both ends of the PN junction. The output power of a single photovoltaic panel is small. Therefore, usually, a large number of photovoltaic cells are connected in series and parallel to form a photovoltaic array for grid-connected power generation. The modeling method of the output characteristics of photovoltaic modules is usually to convert the photovoltaic cell into an equivalent circuit model; specifically, for given external conditions such as stability, light intensity, etc., when the load changes from 0 to , the open-circuit voltage will change from to 0; at the same time, the output current will change from to 0. Under the given external conditions, the output voltage and current of the photovoltaic cell are distributed on a curve, and the magnitude of the output power also changes with the value of the voltage. The optimal operating point of the photovoltaic cell, the corresponding voltage and current are the maximum power point voltage and the maximum power point current respectively; the rectangular area formed by and is also called the maximum output power of the photovoltaic cell, that is: , where is the fill factor or curve factor of the photovoltaic cell, and its magnitude is and the rectangular area formed and and the curve area formed the ratio of.

[0031] Among them, due to the dynamic characteristics of the photovoltaic cell being affected by various external conditions, such as light intensity, ambient temperature, particle radiation, etc., the influences of these factors on the performance of the photovoltaic cell often exist simultaneously, such as temperature and light intensity. During the application process, the parameter values determined by the above-mentioned normalized fuzzy soft set In actual applications, considering the ambient temperature and light intensity on the surface temperature of the photovoltaic module the influence of, and compensating for the light intensity and photovoltaic cell problems to obtain the parameter values under different light intensities and cell temperatures.

[0032] Optionally, the above parameter values include open-circuit voltage, output current, maximum power point current, and maximum power point voltage. The parameter values are determined by the fuzzy soft set, specifically: ; , , ; In the formula, represents the open-circuit voltage, represents the output current, represents the maximum power point voltage, represents the maximum power point current, A and k represent the diode characteristic factor and the Boltzmann constant, and respectively represent the surface temperature, series internal resistance, and parallel internal resistance of the photovoltaic module, q is the unit charge, represents the diode saturation current, represents the current generated by light radiation, represents the total recombination current, represents the total recombination voltage, represents the fill factor or curve factor of the photovoltaic cell, which is and the ratio of the rectangular area formed and and the curve area formed; The surface temperature of the photovoltaic module is specifically: , in the formula, represents the surface temperature of the photovoltaic module, represents the ambient temperature, represents the coefficient, , represents the light intensity.

[0033] Specifically, assume that the attribute standards generated by the performance requirements of the photovoltaic cell are hybrid fuzzy numbers: ((0.6, 0.7, 0.8), (0.6, 0.8), (0.6, 0.4), (0.5, 0.6, 0.7, 0.8)).

[0034] Assume that the dynamic characteristic requirement of the photovoltaic cell is the comprehensive utility of 0.75. Then, the application of the normalization strategy is transformed into a general fuzzy soft set as shown in Table 1, and the respective comprehensive performances are generated.

[0035] Table 1 hesitant fuzzy number interval number trapezoidal fuzzy number comprehensive performance open-circuit voltage 0.63 0.59 0.45 0.56 output current 0.77 0.70 0.79 0.75 maximum power point voltage 0.70 0.85 0.62 0.72 maximum power point current 0.73 0.74 0.89 0.77 At this time, the parameters , The corresponding determined comprehensive performance is greater than 0.75. Thus, this parameter can be determined as the parameter for predicting the power generation. The other two parameters can be determined in the same way.

[0036] S2. Train the preset initial model based on historical data until the initial model reaches the preset training end condition, and determine the initial model that reaches the training end condition as the load prediction model.

[0037] Among them, the above initial model is a three-layer BP perceptron neural network, including an input layer, an output layer, and a hidden layer. The model is expressed as: ; where: is the hidden layer node, satisfying: .

[0038] Specifically, the hidden layer transfer function of the load prediction model established in this embodiment selects the hyperbolic tangent tanh function, and the output layer transfer function selects the equal value function. Therefore, the following non-linear relationship is shown from the input layer to the output layer: In the formula, represents the output of the output layer in the initial model, represents the number of nodes in the hidden layer, represents the transfer function of the output layer in the initial model, which is the equal value function, K represents the number of historical data input to the initial model, represents the historical data input to the initial model k , Denote the hidden layer weights, Denote the output layer weights, Denote the hidden layer bias, Denote the output layer bias, superscript h Denote the label of the hidden layer, tanh Denote the hyperbolic tangent function.

[0039] Specifically, the model parameters in the initial model can be uniformly represented by the parameter vector as: ; Similar to the parameter estimation formula of the quantile regression model, the estimation of the parameter vector in the load forecasting model also aims to minimize the error function.

[0040] Optionally, the above training end condition is that the error function of the initial model does not exceed a threshold, and the error function is specifically: ; In the formula, Denote the error function value, T Denote the sample size used for training the initial model in the historical data, Denote the i th training sample response variable Y value in the historical data, Denote the conditional quantile i obtained through the explanatory variable of the th training sample , is the indicator function.

[0041] Specifically, to prevent the initial model from falling into overfitting, a penalty term is added to the objective function. At this time, the parameter estimation of the initial model is transformed into the following optimization problem: The model parameters of the above initial model are represented as: , where: ; In the formula, Denote the penalty parameter, Denote the number of hidden layer nodes, K Denote the quantity of historical data input to the initial model, is the quadratic norm, Denote the hidden layer weight vector, T Denote the sample size used for training the initial model in the historical data, Denote the value of the error function of the initial model.

[0042] Further, the optimal values of the above penalty parameter and the number of hidden layer nodes are determined by the AIC criterion, specifically as follows: ; In the formula, is the minimum error function at different and quantiles, represents the sample size used to train the initial model in historical data, T represents the number of historical data input into the initial model, K represents the number of hidden layer nodes.

[0043] S3. Obtain the actual operation data, and input the actual operation data into the load prediction model for processing to obtain the predicted load for a future set time period.

[0044] Specifically, according to the AIC criterion, the optimal values of the penalty parameter and the number of hidden layer nodes are screened out, and the optimal parameter that minimizes the error function is calculated; since the indicator function and the activation function in the error function are not differentiable at the origin, the traditional gradient-based nonlinear optimization algorithm is not suitable for directly optimizing the formula. It is proposed to first use the everywhere-differentiable Huber norm smoothing approximation function to approximately replace and , and then use the nonlinear optimization method to solve.

[0045] S4. Use the predicted power generation and the predicted load to establish a sensitivity matrix in the distribution network, and analyze the locally optimal matching combination of power sources and loads in the distribution network based on the sensitivity matrix.

[0046] Among them, the dynamic sensitivity is selected as the operation analysis index. Based on the inherent admittance matrix of the power grid, multiple influence factors are formed and constructed in the way of regional density. Its overall complexity and cost are superior to the separate reconstruction of the distribution network; the construction of tightness mainly starts from sensitivity analysis. Sensitivity calculation can describe the sensitivity degree between grid nodes based on various dependent variable relationships and is the most intuitive index reflecting tightness. Specifically, sensitivity calculation is generally used to study the differential relationship of various corresponding physical variables under the operation state of the power grid, to obtain the sensitivity degree between the dependent variable and the independent variable, so as to analyze a specific variable of the power grid operation. It has a wide range of applications in the power grid, such as power grid reliability analysis, line loss analysis, transmission capacity, etc., and can improve system security, increase system stability margin and economy, etc.

[0047] ​Furthermore, the power flow sensitivity matrix constructed in the embodiment involves three types of variables: independent parameters , including invariant parameters such as line admittance, denoted by ; state variables , including load node voltage, phase angle, generator node phase angle, etc., denoted by ; control variables , generator node active power and voltage, balance node voltage and phase angle, etc., denoted by .

[0048] The current main analysis and calculations using the sensitivity matrix include: ① bus sensitivity, which can determine the reactive power compensation position and judge the weak state of the bus; ② branch sensitivity, which can judge the influence degree of a certain branch on the system voltage stability; ③ photovoltaic power generation sensitivity, which judges the most important photovoltaic power generation for voltage stability near the critical point. The methods for calculating power flow sensitivity mainly include static sensitivity and trajectory sensitivity, which analyze the system from the static section and dynamic response perspectives respectively. The general expression of steady-state power flow, according to the above-mentioned parameters, is: .

[0049] Among them, it can be assumed that the system stable point is , and if the stable point after system disturbance is , according to Taylor expansion, we have: ; substituting the above two equations and solving, we get: ; The above equation is the sensitivity matrix of the change of causing the change of .

[0050] Specifically, for example, when voltage is used as the main reference variable for system security, the sensitivity matrix is based on the voltage fluctuation sensitivity relationship between the power supply node and the load node. The weak part in the voltage changes of different nodes in the system is used as the constraint of the safety boundary during system operation, which can guide the optimal operation of the system. For the differential relationship matrix of generators and load nodes with respect to node voltage changes, the reactive power balance equation injected into the node is: ; In the formula, is the output of the power grid to the load, is the imaginary part, is the total amount of nodes, is the node j voltage amplitude at.

[0051] Furthermore, the power generation and load nodes are separately rewritten in matrix form: ; In the formula, 、 、 、 are invariant parameters corresponding to the load and the power generation node, is the voltage amplitude at the load, is the voltage amplitude at the power generation, are the change amounts of the load and the power generation node. Assume that after adjusting it remains unchanged, that is which is namely: , , .

[0052] Among them, is the sensitivity matrix of the voltage between the power generation and the load nodes. Through this calculation, the power generation node with the greatest impact on the change of the load node can be analyzed. If there is a phenomenon of voltage rise at the power supply node, it will affect the qualified rate of the load voltage. Therefore, according to this sensitivity matrix, the system security boundary and the influence ranking of reactive power changes can be further determined.

[0053] Specifically, establishing the sensitivity matrix in the distribution network based on the predicted power generation and load helps to analyze the local optimal matching combination of the power supply and the load under the operation of the active distribution network, realize the optimal distribution of the power flow, and improve the stability and economy of the power system.

[0054] Embodiment 2: The embodiment of the present application provides a distribution network source-load joint portrait system based on photovoltaic historical data, which is applied to a distribution network source-load joint portrait method according to any one of Embodiments 1, as Figure 2 shown, including: The first module is used to use the historical data of the photovoltaic power generation unit as a fuzzy soft set, perform normalization processing on the fuzzy soft set, and predict the future set time period of the power generation prediction power by using the normalized fuzzy soft set.

[0055] The second module is used to train the preset initial model based on the historical data until the initial model reaches the preset training end condition, and determine the initial model that reaches the training end condition as the load prediction model.

[0056] The third module is used to obtain the actual operation data and input the actual operation data into the load prediction model for processing to obtain the predicted load for the future set time period.

[0057] The fourth module is configured to establish a sensitivity matrix in the distribution network by using the predicted power generation and the predicted load, and analyze the locally optimal matching combination of power sources and loads in the distribution network based on the sensitivity matrix.

[0058] Embodiment 3: An embodiment of the present application provides an electronic device. As Figure 3 shown, it includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the method according to any one of Embodiment 1 is implemented.

[0059] Embodiment 4: An embodiment of the present application provides a non-transitory computer-readable storage medium. The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the method according to any one of Embodiment 1.

[0060] The above specific embodiments further elaborate on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A distribution network source-load joint profiling method based on photovoltaic historical data, characterized in that: The specific steps include: The historical data of the photovoltaic power generation group is used as a fuzzy soft set, the fuzzy soft set is normalized, and the normalized fuzzy soft set is used to predict the power generation prediction of a future set time period; Training a preset initial model based on the historical data until the initial model reaches a preset training end condition, and determining the initial model that reaches the training end condition as a load forecasting model; Acquire actual operation data, and input the actual operation data into the load forecasting model for processing to obtain the forecast load for a future set time period; A sensitivity matrix in the distribution network is established by using the predicted power generation and the predicted load, and based on the sensitivity matrix, a local optimal matching combination of power sources and loads in the distribution network is analyzed.

2. According to claim 1, a distribution network source-load joint portrait method based on photovoltaic historical data is characterized in that: The initial model is a three-layer BP perceptron neural network, including an input layer, an output layer and a hidden layer, specifically: ； In the formula, represents the output of the output layer in the initial model, represents the number of nodes in the hidden layer, represents the conversion function of the output layer in the initial model, which is an equivalent function. K represents the amount of historical data input to the initial model, Represents the historical data input to the initial model k , represents the hidden layer weight, represents the output layer weight, represents the hidden layer offset, Indicates the output layer offset, the superscript h represents the label of the hidden layer, tanh represents the hyperbolic tangent function.

3. A method for joint source-load profiling of a distribution network based on photovoltaic historical data according to claim 2, characterized in that: The training end condition is that the error function of the initial model does not exceed a threshold value, and the error function is specifically: ； In the formula, represents the error function value of the initial model, T represents the sample size of historical data used to train the initial model, Indicates the historical data i training sample response variable Y The value of Indicates that through i Explanatory variables for training samples What you get The conditional quantile of , is the indicator function.

4. A method for joint source-load profiling of a distribution network based on photovoltaic historical data according to claim 3, characterized in that: The model parameters of the initial model are expressed as: ,in: ； In the formula, represents the penalty parameter, represents the number of hidden layer nodes, K represents the amount of historical data input to the initial model, is the quadratic norm, represents the hidden layer weight vector, T represents the sample size of historical data used to train the initial model, Represents the error function value of the initial model.

5. A method for joint source-load profiling of a distribution network based on photovoltaic historical data according to claim 4, characterized in that: The optimal values ​​of the penalty parameter and the number of hidden layer nodes are determined by the AIC criterion, specifically: ； In the formula, To choose different and hour The minimum error function under quantile, T represents the sample size of historical data used to train the initial model, K represents the amount of historical data input to the initial model, Represents the number of hidden layer nodes.

6. A method for joint source-load profiling of a distribution network based on photovoltaic historical data according to claim 1, characterized in that: The predicted power generation power in the future set time period is obtained by using the fuzzy soft set prediction after normalization, specifically: The parameter value for predicting the power generation prediction is determined by using the normalized fuzzy soft set, and the power generation prediction is calculated based on the parameter value and the surface temperature of the photovoltaic module.

7. A method for joint source-load profiling of a distribution network based on photovoltaic historical data according to claim 6, characterized in that: The parameter values ​​include open circuit voltage, output current, maximum power point current and maximum power point voltage, and the parameter values ​​are determined by fuzzy soft sets, specifically: ； ， ， ； In the formula, represents the open circuit voltage, Represents the output current, represents the maximum power point voltage, represents the maximum power point current, A and k represents the diode characteristic factor and the Boltzmann constant, and represent the surface temperature, series internal resistance and parallel internal resistance of the photovoltaic module respectively. q is the unit charge, represents the diode saturation current, represents the current generated by light radiation, represents the total composite current, represents the total composite voltage, The fill factor or curve factor of the photovoltaic cell is and The area of ​​the rectangle formed by and The ratio of the areas of the curves they form; The surface temperature of the photovoltaic module is specifically: , where Indicates the surface temperature of the photovoltaic module, Indicates the ambient temperature, represents the coefficient, , Indicates light intensity.

8. A distribution network source-load joint portrait system based on photovoltaic historical data, applied to a distribution network source-load joint portrait method based on photovoltaic historical data according to any one of claims 1 to 7, characterized in that: include: The first module is used to use the historical data of the photovoltaic power generation group as a fuzzy soft set, normalize the fuzzy soft set, and use the normalized fuzzy soft set to predict the power generation forecast for a future set time period; The second module is used to train a preset initial model based on the historical data until the initial model reaches a preset training end condition, and determine the initial model that reaches the training end condition as a load forecasting model; The third module is used to obtain actual operation data and input the actual operation data into the load forecasting model for processing to obtain the predicted load for a future set time period; The fourth module is used to establish a sensitivity matrix in the distribution network using the predicted power generation and the predicted load, and analyze the local optimal matching combination of power sources and loads in the distribution network based on the sensitivity matrix.

9. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the method according to any one of claims 1 to 7 is implemented when the processor executes the computer program.

10. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium stores computer instructions, which enable a computer to execute the method of any one of claims 1-7.

Images