Method and system for evaluating energy-saving and carbon-reducing effects of transformer substation

Abstract

The invention discloses a method and a system for evaluating energy-saving and carbon-reducing effects of a transformer substation, and relates to the field of power systems. Constructing an energy consumption prediction model, and training and verifying by using a transformer substation historical data set; constructing and training a variable electric quantity prediction model, predicting future variable electric quantity data of the to-be-evaluated transformer substation, determining future equipment data and future environment data of the to-be-evaluated transformer substation, performing arrangement to obtain a prediction sample, and predicting the predicted total energy consumption of the transformed transformer substation by using the trained energy consumption prediction model; and according to the predicted total energy consumption and the future variable electric quantity data, calculating the energy consumption ratio after transformation, and according to the ratio of the energy consumption ratio before and after transformation, evaluating the energy-saving and carbon-reducing effects. The method comprehensively considers the influence of the power consumption data and the external environment on the overall energy-saving and carbon-reducing effect of the transformer substation, improves the prediction and evaluation accuracy of the overall energy-saving and carbon-reducing effect of the transformer substation, and can provide an effective reference for a specific transformation scheme of the transformer substation.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and more particularly to a method and system for evaluating energy-saving and carbon-reduction effects of a substation. Background Art

[0002] Substations are an important part of the power system, mainly used for the conversion, distribution, dispatching and monitoring of electric energy. During the operation of substations, there are losses in transformers and lines, monitoring and control systems, cooling systems and energy consumption of various auxiliary systems. Energy-saving and carbon-reduction transformation of various systems can effectively reduce power losses, thereby reducing operating costs and carbon emissions, and play a positive role in addressing climate change.

[0003] However, in the existing substation energy-saving and carbon-reduction transformation process, the energy-saving and carbon-reduction effect evaluation is often limited to the energy-saving effect evaluation of a single system, ignoring the impact of electricity consumption data and external environment on the overall energy-saving and carbon-reduction effect of the substation. In addition, the evaluation process mostly relies on expert review and other methods, and the energy-saving effect of the substation after energy-saving transformation is analyzed and predicted based on the experience of experts. The overall energy-saving and carbon-reduction effect prediction and evaluation after the transformation is not accurate enough, and can only roughly estimate the carbon-reduction effect after the transformation, which is not of sufficient guiding significance for specific transformation plans.

[0004] Therefore, how to improve the accuracy of prediction and evaluation of the overall energy-saving and carbon-reduction effects of substations is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0005] In view of this, the present invention provides a method and system for evaluating the energy-saving and carbon-reduction effects of a substation, so as to accurately predict the energy-saving and carbon-reduction effects after the substation transformation.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] The present invention discloses a method for evaluating the energy-saving and carbon-reduction effect of a substation, and the specific steps are as follows:

[0008] Obtain historical operation data, historical equipment data, and historical environmental data of several substations, and sort them out after preprocessing to obtain several samples to obtain a substation historical data set;

[0009] Building an energy consumption prediction model and using the substation historical data set for training and verification;

[0010] Construct a substation quantity prediction model, train it based on the historical operation data, and predict the future substation quantity data of the substation to be evaluated; determine the future equipment data of the substation to be evaluated according to the substation transformation plan; obtain the future environmental data of the substation to be evaluated;

[0011] Based on the future variable power data, the future equipment data, and the future environmental data, prediction samples are sorted out, and the trained energy consumption prediction model is used to predict the total predicted energy consumption of the transformed substation; according to the total predicted energy consumption and the future variable power data, the energy consumption ratio after transformation is calculated, and the energy-saving and carbon-reduction effect is evaluated according to the ratio of the energy consumption ratios before and after transformation.

[0012] Further, the historical operation data includes: the input power, output power, energy consumption of monitoring and control equipment, energy consumption of cooling equipment, and energy consumption of auxiliary equipment at each moment of the substation;

[0013] The historical equipment data includes: the energy consumption coefficients of substation equipment, transmission lines, monitoring equipment, cooling equipment, and auxiliary equipment, and the rated power of substation equipment;

[0014] The historical environmental data includes: the monthly average sunshine duration, average temperature, average humidity, and average wind speed at the location of the substation.

[0015] Further, the energy consumption prediction model is a generalized regression neural network model, which consists of an input layer, a hidden layer, a summation layer, and an output layer;

[0016] The number of nodes in the input layer is the same as the dimension of the input sample, and the sample information is transmitted to the hidden layer; the hidden layer uses a kernel function to perform feature extraction and non-linear transformation on the input information and outputs it to the summation layer; the summation layer uses an arithmetic summation function and a weighted summation function to perform summation respectively and outputs the results to the output layer; the output layer outputs the final prediction result.

[0017] Further, the kernel function is:

[0018]

[0019] where, P i is the output of the kernel function, α∈(0,1) is the proportionality coefficient, σ is the width coefficient of the Gaussian function, β is the coefficient of the sigmoid function, X=(X1,X2…X i …X n ) T represents the input sample, X i represents the i-th dimension of the input sample, n represents the dimension of the input sample, and i represents a natural number;

[0020] The arithmetic summation function is:

[0021]

[0022] The weighted summation function is:

[0023]

[0024] Among them, S D and S Nj are the arithmetic sum and the weighted sum of the j-th output node respectively, and W ij is the output weight from the i-th neuron in the hidden layer to the j-th output node in the output layer;

[0025] The output Y of the output layer is:

[0026] Y = (y1, y2,... y j ... y m ) T ;

[0027] Among them, is the output of the j-th output node in the output layer, and m is the dimension of the output sample.

[0028] Furthermore, the generalized regression neural network model uses the tuna school algorithm to find the optimal hyperparameters, specifically:

[0029] Step 1: Initialize the maximum number of iterations, the search lower limit, the search upper limit, and the random position vector of each individual;

[0030] Step 2: Update the velocity coefficient and the time coefficient;

[0031] Step 3: Update the position of each individual according to the position update function;

[0032] Step 4: Calculate the fitness value of each position and determine the optimal position;

[0033] Step 5: When the fitness value of the optimal position is greater than the set threshold, or the number of iterations reaches the maximum number of iterations, determine the optimal hyperparameters according to the current optimal position; otherwise, return to Step 2.

[0034] Furthermore, the velocity coefficient is: V = k * z * r3 + 1, and the time coefficient is T = 2 * z * r2 - z; where V is the velocity coefficient, T is the time coefficient, k is a constant, r1, r2, r3 are random numbers within [0, 1], and z is the attenuation coefficient, specifically:

[0035]

[0036] Among them, e is the natural constant, l is the current number of iterations, and l max is the maximum number of iterations;

[0037] The position update function is:

[0038]

[0039] Among them, Z l+1 is the updated position, Z l 、V l 、T l are respectively the optimal position, the current search position, the velocity coefficient, and the time coefficient of the l-th iteration.

[0040] Furthermore, the operation data prediction model is an autoregressive moving average-LSTM model, and the specific training steps include:

[0041] Sample the output power of the substation to be evaluated in the historical operation data to obtain the original variable electricity data sequence, and perform sliding window segmentation to obtain several original sub-variable electricity data sequences;

[0042] Calculate the variance of each of the original sub-variable electricity data sequences. Those with variances less than the set threshold are identified as stationary sequences, and those greater than the threshold are subjected to d-order difference processing to obtain all stationary variable electricity data sequences;

[0043] Use the least squares method for parameter estimation to determine the best parameter estimation of the autoregressive moving average model, and obtain the autoregressive moving average model part;

[0044] Input each of the stationary variable electricity data sequences into the autoregressive moving average model part for prediction, and calculate the residual sequence of each predicted variable electricity sequence and the corresponding time period of the original variable electricity data sequence;

[0045] Perform EMD processing on each residual sequence to separate the residual into g intrinsic mode components;

[0046] Use the intrinsic mode components to train the LSTM model to obtain the trained LSTM model part;

[0047] The sum of the prediction sequence output by the autoregressive moving average model part and the prediction residual sequence output by the LSTM model part is the prediction result of the autoregressive moving average-LSTM model.

[0048] Furthermore, the formula of the autoregressive moving average model part is:

[0049]

[0050] where p is the order of the autoregressive model, D is the number of differences, q is the order of the moving average model, L is the lag operator, X t represents the input data sequence, represents the autoregressive part coefficient, θ r represents the moving average part coefficient, ε t represents the error term, and r = 1, 2,..., p are natural numbers.

[0051] Furthermore, the LSTM model includes an input layer, an LSTM layer, and a fully connected layer;

[0052] The LSTM layer includes g sub-LSTM models. The input layer inputs g types of connotative modal components into the corresponding sub-LSTM models, and the fully connected layer superimposes the outputs of the sub-LSTM models to obtain a predicted residual sequence.

[0053] The present invention also discloses a substation energy-saving and carbon-emission reduction effect evaluation system, including:

[0054] Data acquisition module: Acquire the historical operation data, historical equipment data, and historical environmental data of several substations, preprocess and organize them to obtain several samples, and obtain the substation historical data set;

[0055] Energy consumption model module: Construct an energy consumption prediction model, and train and verify it using the substation historical data set;

[0056] Predicted data module: Construct a variable electricity quantity prediction model, train it based on the historical operation data, and predict the future variable electricity quantity data of the substation to be evaluated; Determine the future equipment data of the substation to be evaluated according to the substation transformation plan; Obtain the future environmental data of the substation to be evaluated;

[0057] Effect evaluation module: According to the future variable electricity quantity data, the future equipment data, and the future environmental data, organize and obtain prediction samples, and use the trained energy consumption prediction model to predict the predicted total energy consumption of the substation after transformation; According to the predicted total energy consumption and the future variable electricity quantity data, calculate the energy consumption ratio after transformation, and evaluate the energy-saving and carbon-emission reduction effect according to the ratio of the energy consumption ratios before and after transformation.

[0058] It can be seen from the above technical solutions that compared with the prior art, the present invention discloses a substation energy-saving and carbon-emission reduction effect evaluation method and system, which integrates historical operation data, equipment data, and environmental data to form a systematic data set, making the evaluation process more comprehensive and reliable; By constructing an energy consumption prediction model and a variable electricity quantity prediction model, accurate energy consumption prediction can be achieved, providing a scientific basis for the energy-saving transformation of substations; By calculating the energy consumption ratios before and after transformation, the energy-saving and carbon-emission reduction effect can be clearly evaluated, providing a quantitative basis for the transformation decision of substations and promoting sustainable development; The present invention comprehensively considers the influence of electricity consumption data and the external environment on the overall energy-saving and carbon-emission reduction effect of substations, improves the accuracy of predicting and evaluating the overall energy-saving and carbon-emission reduction effect of substations, and can provide an effective reference for the specific transformation plan of substations. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on the provided drawings.

[0060] Figure 1 It is a schematic diagram of the overall process of the embodiment of the present invention. Specific implementation manners

[0061] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0062] The embodiment of the present invention discloses a method for evaluating the energy-saving and carbon-reduction effects of a substation, as Figure 1 shown, the specific steps are as follows:

[0063] Obtain the historical operation data, historical equipment data, and historical environmental data of several substations, preprocess and organize them to obtain several samples, and obtain the substation historical data set;

[0064] Construct an energy consumption prediction model, and train and verify it using the substation historical data set;

[0065] Construct a variable electricity prediction model, train it based on the historical operation data, and predict the future variable electricity data of the substation to be evaluated; determine the future equipment data of the substation to be evaluated according to the substation transformation plan; obtain the future environmental data of the substation to be evaluated;

[0066] According to the future variable electricity data, future equipment data, and future environmental data, organize and obtain prediction samples, and use the trained energy consumption prediction model to predict the predicted total energy consumption of the substation after transformation; calculate the energy consumption ratio after transformation according to the predicted total energy consumption and the future variable electricity data, and evaluate the energy-saving and carbon-reduction effects according to the ratio of the energy consumption ratios before and after transformation.

[0067] In a specific embodiment, the historical operation data includes: the input power, output power, energy consumption of monitoring and control equipment, energy consumption of cooling equipment, and energy consumption of auxiliary equipment at each moment of the substation;

[0068] The historical equipment data includes: the energy consumption coefficients of substation equipment, transmission lines, monitoring equipment, cooling equipment, and auxiliary equipment, and the rated power of the substation equipment;

[0069] Historical environmental data includes: monthly average sunshine duration, average temperature, average humidity, and average wind speed at the location where the substation is located.

[0070] Specifically, for the calculation of the energy consumption ratio of the substation, it is usually calculated on an annual basis. The operation data of the substation is sampled at a set sampling interval, and data cleaning and missing value supplementation are performed to obtain historical operation data. The conversion efficiency of the power conversion equipment is used as its energy consumption coefficient; the loss of transmitting a set amount of electricity by the transmission line is used to determine its energy consumption coefficient; the energy consumption level of the monitoring equipment is used as its energy consumption coefficient; the energy consumption level of the cooling equipment is used as its energy consumption coefficient; the auxiliary equipment includes lighting equipment and standby power equipment, and the energy consumption coefficient of the auxiliary equipment is obtained by weighted calculation based on the energy consumption levels of the two, and the weight is determined according to the ratio of the energy consumption of the two. The historical environmental data is extracted from the meteorological monitoring station data at the location where the substation is located.

[0071] Among them, the output power is the output power of the power conversion equipment, that is, the power conversion amount of the substation. According to the output power and the rated power of the power conversion equipment, the no-load rate at each moment can be determined. For the prediction samples of the energy consumption prediction model, the output power sequence is used as the first dimension of the prediction sample; the no-load rate sequence is used as the second dimension of the prediction sample; according to the historical equipment data, the equipment vector is sorted out as the third dimension of the prediction sample; according to the historical environmental data, the environmental sequence is sorted out as the fourth dimension of the prediction sample; the annualized energy consumption of the power conversion equipment (the energy consumption of the power conversion equipment is calculated based on the input power and the output power), the monitoring and control equipment, the cooling equipment, and the auxiliary equipment is used as the sample label to obtain the final prediction sample.

[0072] In a specific embodiment, the energy consumption prediction model is a generalized regression neural network model, which consists of an input layer, a hidden layer, a summation layer, and an output layer;

[0073] The number of nodes in the input layer is the same as the dimension of the input sample, and the sample information is transmitted to the hidden layer; the hidden layer uses a kernel function to perform feature extraction and non-linear transformation on the input information and outputs it to the summation layer; the summation layer uses an arithmetic summation function and a weighted summation function to perform summation respectively and outputs the result to the output layer; the output layer outputs the final prediction result.

[0074] In a specific embodiment, the kernel function is:

[0075]

[0076] Where P i is the output of the kernel function, α∈(0,1) is the proportionality coefficient, σ is the Gaussian function width coefficient, β is the sigmoid function coefficient, X=(X1,X2…X i …X n ) TDenote the input sample as X i Denote the i-th dimension of the input sample, n represents the dimension of the input sample, that is, the four dimensions of the prediction sample in the foregoing embodiment, and i represents a natural number;

[0077] The arithmetic summation function is:

[0078]

[0079] The weighted summation function is:

[0080]

[0081] Among them, S D and S Nj are the arithmetic sum and the weighted sum of the j-th output node respectively, and W ij is the output weight from the i-th neuron in the hidden layer to the j-th output node in the output layer;

[0082] The output Y of the output layer is:

[0083] Y = (y1, y2,... y j ... y m ) T ;

[0084] Among them, is the output of the j-th output node in the output layer, and m is the dimension of the output sample, corresponding to the annualized energy consumption of the power transformation equipment, monitoring and control equipment, cooling equipment, and auxiliary equipment in the sample label of the prediction sample respectively.

[0085] In a specific embodiment, the generalized regression neural network model uses the tuna school algorithm to find the optimal hyperparameters, specifically:

[0086] Step 1: Initialize the maximum number of iterations, search lower limit, search upper limit, and the random position vector of each individual;

[0087] Step 2: Update the velocity coefficient and time coefficient;

[0088] Step 3: Update the position of each individual according to the position update function;

[0089] Step 4: Calculate the fitness value of each position. The fitness value is the mean square error of the prediction value of the generalized regression neural network model, and determine the optimal position;

[0090] Step 5: When the fitness value of the optimal position is greater than the set threshold, or the number of iterations reaches the maximum number of iterations, determine the optimal hyperparameters (i.e., α, σ, β, and W ij ) according to the current optimal position; otherwise, return to Step 2.

[0091] Generalized regression neural network model, and use the tuna school algorithm to optimize the hyperparameters of the model, which can effectively improve the performance of the model, reduce the prediction error, thus improving the modeling ability for complex non-linear relationships and enhancing the accuracy of energy consumption prediction.

[0092] In a specific embodiment, the speed coefficient is: V = k * z * r3 + 1, and the time coefficient is T = 2 * z * r2 - z; where, V is the speed coefficient, T is the time coefficient, k is a constant, r1, r2, r3 are random numbers within [0, 1], and z is the attenuation coefficient, specifically:

[0093]

[0094] where, e is the natural constant, l is the current iteration number, l max is the maximum iteration number;

[0095] The position update function is:

[0096]

[0097] where, Z l+1 is the updated position, Z l 、V l 、T l are respectively the optimal position, the current search position, the speed coefficient, and the time coefficient at the l-th iteration.

[0098] In a specific embodiment, the operation data prediction model is an autoregressive moving average-LSTM model, and the specific training steps include:

[0099] Sample the output power of the substation to be evaluated in the historical operation data to obtain the original variable electricity data sequence, and perform sliding window (the sliding window width is usually 1 year) segmentation to obtain several original sub-variable electricity data sequences;

[0100] Calculate the variance of each original sub-variable electricity data sequence, and those with variance less than the set threshold are identified as stationary sequences, and those greater than the threshold are subjected to d-order difference processing to obtain all stationary variable electricity data sequences;

[0101] Use the least squares method for parameter estimation to determine the best parameter estimation of the autoregressive moving average model, and obtain the autoregressive moving average model part; specifically, randomly generate p, q, D, θ rMultiple combinations of parameters are obtained, and the stationary variable power data sequence is used as the input to obtain the output results of the autoregressive moving average models corresponding to each parameter combination. According to the output results, the Bayesian information criterion (BIC) is used to calculate the BIC values of the autoregressive moving average models respectively, and the parameter combinations with BIC values less than the preset threshold are selected. Finally, according to the selected parameter combinations, the least squares method is used to fit the selected optimal parameter combination, which is the best parameter estimate.

[0102] Each stationary variable power data sequence is input into the autoregressive moving average model part for prediction, and the residual sequence of each predicted variable power sequence and the original variable power data sequence in the corresponding time period is calculated.

[0103] EMD processing is performed on each residual sequence, and the residual is separated into g intrinsic mode components.

[0104] The LSTM model is trained using the intrinsic mode components to obtain the trained LSTM model part.

[0105] The sum of the prediction sequence output by the autoregressive moving average model part and the prediction residual sequence output by the LSTM model part is the prediction result of the autoregressive moving average-LSTM model.

[0106] In a specific embodiment, the formula of the autoregressive moving average model part is:

[0107]

[0108] where p is the order of the autoregressive model, D is the number of differences, q is the order of the moving average model, L is the lag operator, X t represents the input data sequence, t represents time, represents the autoregressive part coefficient, θ r represents the moving average part coefficient, ε t represents the error term, and r = 1, 2,..., p are natural numbers.

[0109] In a specific embodiment, since the autoregressive moving average model is fitted according to the least squares method, there must be residuals in its predicted values. By learning the residuals through the LSTM model, the prediction error of the autoregressive moving average model can be further corrected, and the accuracy of the overall power prediction can be improved. The LSTM model includes an input layer, an LSTM layer, and a fully connected layer;

[0110] The LSTM layer includes g sub-LSTM models. The input layer takes g intrinsic mode components (i.e., IMF1, IMF1... IMF g)Input the corresponding sub-LSTM model, and the sub-LSTM model is a classic LSTM model; the fully connected layer superimposes the outputs of each sub-LSTM model to obtain the predicted residual sequence. For the training of the LSTM model, the intrinsic mode components are used as the sample features of the prediction samples, and the corresponding intrinsic mode components are obtained by performing EMD processing on the residual sequence of the next prediction period as the sample labels of the prediction samples.

[0111] An embodiment of the present invention also discloses a substation energy-saving and carbon-reduction effect evaluation system, including:

[0112] Data acquisition module: Obtain the historical operation data, historical equipment data, and historical environmental data of several substations, preprocess and organize them to obtain several samples, and obtain the substation historical data set;

[0113] Energy consumption model module: Construct an energy consumption prediction model and train and verify it using the substation historical data set;

[0114] Predicted data module: Construct a variable power prediction model, train it based on historical operation data, and predict the future variable power data of the substation to be evaluated; Determine the future equipment data of the substation to be evaluated according to the substation transformation plan; Obtain the future environmental data of the substation to be evaluated;

[0115] Effect evaluation module: According to the future variable power data, future equipment data, and future environmental data, organize and obtain prediction samples, and use the trained energy consumption prediction model to predict the total predicted energy consumption of the transformed substation; Calculate the energy consumption ratio after transformation according to the predicted total energy consumption and future variable power data, and evaluate the energy-saving and carbon-reduction effect according to the ratio of the energy consumption ratios before and after transformation.

[0116] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0117] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for evaluating energy saving and carbon reduction effects of a substation, characterized in that: The specific steps are as follows: Obtain historical operation data, historical equipment data, and historical environmental data of several substations, and sort them out after preprocessing to obtain several samples to obtain a substation historical data set; Building an energy consumption prediction model and using the substation historical data set for training and verification; Constructing a substation power prediction model, training it based on the historical operation data, and predicting future substation power data of the substation to be evaluated; Determine the future equipment data of the substation to be evaluated according to the substation renovation plan; Obtain future environmental data of the substation to be evaluated; According to the future substation quantity data, the future equipment data, and the future environment data, a prediction sample is obtained, and the total energy consumption of the transformed substation is predicted by using the trained energy consumption prediction model; The energy consumption ratio after the transformation is calculated based on the predicted total energy consumption and the future variable power data, and the energy saving and carbon reduction effect is evaluated based on the ratio of the energy consumption ratios before and after the transformation.

2. A method for evaluating energy conservation and carbon reduction effects of a substation according to claim 1, characterized in that: The historical operation data includes: input power, output power, energy consumption of monitoring and control equipment, energy consumption of cooling equipment, and energy consumption of auxiliary equipment of the substation at each moment; The historical equipment data includes: energy consumption coefficients of substation equipment, transmission lines, monitoring equipment, cooling equipment and auxiliary equipment, and rated power of substation equipment; The historical environmental data include: the monthly average sunshine duration, average temperature, average humidity, and average wind speed at the location of the substation.

3. A method for evaluating energy saving and carbon reduction effects of a substation according to claim 1, characterized in that: The energy consumption prediction model is a generalized regression neural network model, which consists of an input layer, a hidden layer, a summation layer and an output layer; The number of nodes in the input layer is the same as the dimension of the input sample, and the sample information is transmitted to the hidden layer; The hidden layer uses a kernel function to perform feature extraction and nonlinear transformation on the input information, and outputs it to the summation layer; the summation layer uses an arithmetic summation function and a weighted summation function to perform summation respectively and output the result to the output layer; The output layer outputs the final prediction result.

4. A method for evaluating energy saving and carbon reduction effects of a substation according to claim 3, characterized in that: The kernel function is: Among them, P i is the output of the kernel function, α∈(0,1) is the proportional coefficient, σ is the width coefficient of the Gaussian function, β is the coefficient of the sigmoid function, X=(X1,X2…X i …X n ) T represents the input sample, X i represents the i-th dimension of the input sample, n represents the input sample dimension, and i represents a natural number; The arithmetic sum function is: The weighted sum function is: Among them, S D , S Nj are the arithmetic sum and the weighted sum of the j-th output node, W ij is the output weight of the i-th neural unit in the hidden layer to the j-th output node in the output layer; The output Y of the output layer is: Y=(y1,y2,…y j …y m ) T ; in, is the output of the jth output node in the output layer, and m is the dimension of the output sample.

5. A method for evaluating energy conservation and carbon reduction effects of a substation according to claim 4, characterized in that: The generalized regression neural network model uses the tuna school algorithm to find the optimal hyperparameters, specifically: Step 1: Initialize the maximum number of iterations, search lower limit, search upper limit, and random position vector of each individual; Step 2: Update the speed coefficient and time coefficient; Step 3: Update the position of each individual according to the position update function; Step 4: Calculate the fitness value of each position and determine the optimal position; Step 5: When the fitness value of the optimal position is greater than the set threshold, or the number of iterations reaches the maximum number of iterations, the optimal hyperparameter is determined according to the current optimal position; Otherwise return to step 2.

6. A method for evaluating energy conservation and carbon reduction effects of a substation according to claim 5, characterized in that: The speed coefficient is: V = k*z*r3+1, and the time coefficient is T = 2*z*r2-z; wherein V is the speed coefficient, T is the time coefficient, k is a constant, r1, r2, r3 are random numbers in [0,1], and z is the attenuation coefficient, specifically: Among them, e is a natural constant, l is the current number of iterations, and l max is the maximum number of iterations; The position update function is: Among them, Z l+1 To update the location, Z l 、V l 、T l They are respectively the optimal position of the lth iteration, the current search position, the speed coefficient, and the time coefficient.

7. A method for evaluating energy conservation and carbon reduction effects of a substation according to claim 1, characterized in that: The operation data prediction model is an autoregressive moving average-LSTM model, and the specific training steps include: Sampling the output power of the substation to be evaluated in the historical operation data to obtain an original substation quantity data sequence, and performing sliding window segmentation to obtain a plurality of original substation quantity quantum data sequences; Calculate the variance of each of the original variable quantity quantum data sequences, and identify the sequence with a variance less than a set threshold as a stable sequence, and perform d-order difference processing on the sequence with a variance greater than the threshold to obtain all stable variable quantity data sequences; The least square method is used to perform parameter estimation to determine the best parameter estimation of the autoregressive moving average model, and the autoregressive moving average model part is obtained; Input each of the steady variable quantity data sequences into the autoregressive moving average model part for prediction, and calculate the residual sequence of each predicted variable quantity sequence and the original variable quantity data sequence for the corresponding time period; Perform EMD processing on each residual sequence and separate the residual into g types of intrinsic modal components; Using the connotation modal component to train the LSTM model to obtain a trained LSTM model part; The sum of the prediction sequence output by the autoregressive moving average model part and the prediction residual sequence output by the LSTM model part is the prediction result of the autoregressive moving average-LSTM model.

8. A method for evaluating energy conservation and carbon reduction effects of a substation according to claim 7, characterized in that: The formula for the autoregressive moving average model is: Among them, p is the order of the autoregressive model, D is the number of differences, q is the order of the moving average model, L is the lag operator, and X t Represents the input data sequence, represents the autoregressive coefficient, θ r represents the sliding average coefficient, ε t represents the error term, r=1,2,...,p is a natural number.

9. A method for evaluating energy conservation and carbon reduction effects of a substation according to claim 7, characterized in that: The LSTM model includes an input layer, an LSTM layer, and a fully connected layer; The LSTM layer includes g sub-LSTM models, the input layer inputs g types of intrinsic modal components into the corresponding sub-LSTM models, and the fully connected layer superimposes the outputs of each sub-LSTM model to obtain a prediction residual sequence.

10. A substation energy-saving and carbon-reduction effect evaluation system, using a substation energy-saving and carbon-reduction effect evaluation method according to any one of claims 1 to 9, characterized in that: include: Data acquisition module: obtains historical operation data, historical equipment data, and historical environmental data of several substations, and obtains several samples after preprocessing to obtain a historical data set of the substation; Energy consumption model module: construct an energy consumption prediction model and use the substation historical data set for training and verification; Prediction data module: construct a substation quantity prediction model, train it based on the historical operation data, and predict the future substation quantity data of the substation to be evaluated; determine the future equipment data of the substation to be evaluated according to the substation transformation plan; Obtain future environmental data of the substation to be evaluated; Effect evaluation module: according to the future substation power data, the future equipment data, and the future environment data, a prediction sample is obtained, and the total energy consumption of the transformed substation is predicted by using the trained energy consumption prediction model; The energy consumption ratio after the transformation is calculated based on the predicted total energy consumption and the future variable power data, and the energy saving and carbon reduction effect is evaluated based on the ratio of the energy consumption ratios before and after the transformation.

Images