A method for determining energy consumption indicators based on big data platform

Abstract

The present invention proposes a method for determining energy consumption indicators based on a big data platform, which includes the following steps: data collection, collecting factors such as temperature, rainfall, PM2.5, permanent population, and meter voltage, and transmitting the information collected by devices such as concentrators and smart meters to the Internet of Things, production libraries, and big data platforms; using KPCA technology to perform feature dimensionality reduction processing to increase the accuracy of prediction results; using a prediction method combining HS-GRU algorithm, gray model, and trend extrapolation method to improve the prediction effect of power grid load; the big data platform analyzes existing data to improve the management efficiency of line loss. The present invention can effectively improve the efficiency of line loss management, predict power grid load, and promote the sustainable development of the power industry.

Description

Technical Field

[0001] The present invention relates to the technical field of human resource management, and in particular to a method for determining energy consumption indicators based on a big data platform. Background Art

[0002] In the era of big data, power energy big data plays a great role in promoting the development of the power industry. With the support of cloud platforms, it promotes the sharing and development of power data. Line loss management, load forecasting, pollution prevention and control often occur in power energy applications. Through the collection and application of big data, the efficiency of line loss management can be effectively improved, the load of the power grid can be predicted, and the efforts to prevent pollution can be strengthened, thus promoting the sustainable development of the power industry. Summary of the invention

[0003] In order to achieve the above object, the present invention proposes a method for determining energy consumption indicators based on a big data platform, comprising the following steps:

[0004] S1. The first step is data collection. To determine the energy consumption of power resources, it is necessary to collect factors such as temperature, rainfall, PM2.5, permanent population, and meter voltage;

[0005] S2. Secondly, KPCA technology is used to reduce the dimension of features to increase the accuracy of prediction results;

[0006] S3, then adopt the prediction method combining HS-GRU algorithm, grey model and trend extrapolation method, each of which accounts for a certain proportion. This comprehensive method can improve the prediction effect of the final power grid load;

[0007] S4. Finally, use the big data platform to analyze the existing data and improve the management efficiency of line losses.

[0008] The HS-GRU algorithm in step S3 shown above uses the harmony algorithm to optimize the GRU. The specific operations are as follows:

[0009] S311, firstly, the harmony search parameters are initialized, including the harmony memory HMS, the harmony memory consideration rate HMCR, and the pitch adjustment probability PAR;

[0010] S312, initializing the harmony memory, randomly generating a set of candidate solutions, that is, a parameter set of the GRU model, and storing them in the harmony memory;

[0011] S313, forward propagation of the GRU unit, processes the input sequence through the standard GRU unit to obtain the hidden state sequence;

[0012] S314, using each candidate solution in the harmony memory library, that is, the GRU model parameter, to calculate the objective function value, where the objective function selects the prediction error;

[0013] S315, generating a new harmony vector, selecting a parameter value from the harmony memory bank with probability HMCR, randomly selecting a parameter value from the parameter space with probability 1-HMCR, and fine-tuning the selected parameter value with probability PAR;

[0014] S316, updating the harmony memory library, if the objective function value of the generated new harmony vector is better than the worst candidate solution in the harmony memory library, replacing the worst candidate solution with the new harmony vector;

[0015] S317. Finally, repeat steps S315 and S316 until the maximum number of iterations is reached or the objective function value converges.

[0016] Preferably, the step S1 transmits the information collected by the concentrator and the smart meter device to the Internet of Things, and transmits the data information to the production library and the big data platform through calculation as the data basis for subsequent work.

[0017] Preferably, the step S2 adopts the KPCA technology to perform feature dimensionality reduction processing as follows:

[0018] S21. Standardize each feature in the data set so that its mean is 0 and its variance is 1.

[0019] S22. Select RBF kernel to construct kernel matrix: where x i ,x j represents two data points, σ ​​is the bandwidth parameter of the RBF kernel function, and for each pair of data points, the RBF kernel function value is calculated to construct the kernel matrix K, whose element K ij = k(x i ,x j );

[0020] S23. Centralize the RBF kernel matrix: K′=K-1 n K-K1 n +1 n K1 n , then perform eigenvalue decomposition on the centralized RBF kernel matrix: K′α i =λ i α i , where λ i is the eigenvalue, α i is the corresponding eigenvector;

[0021] S24, select the eigenvectors corresponding to the first six largest eigenvalues ​​as the new feature space after dimensionality reduction;

[0022] S25. Map the original data to the new feature space to obtain the reduced-dimensional data: Where T represents the eigenvector αi The transpose of .

[0023] Preferably, the specific steps of using the GM (1,1) grey model for prediction in step S3 are:

[0024] S321. First, assume that the original data sequence is {x(1), x(2), ..., x(n)};

[0025] S322, then generate a cumulative sequence {X(1), X(2), ..., X(n)}, where k is the number of data sequences;

[0026] S323, then establish the differential equation of the cumulative sequence: Where a and b are the parameters that need to be estimated;

[0027] S324. Finally, the prediction model is obtained by solving the differential equation by estimating the parameters a and b using the least squares method.

[0028] Preferably, the specific steps of the trend extrapolation method in step S3 are:

[0029] S331, first collect and organize historical data to form a time series {y1,y2,...,y n}, where y i represents the observation value at the i-th moment;

[0030] S332, then select the polynomial trend model, the data presents a polynomial trend: y t =a0+a1t+a2t 2 +...+a m t m +σ t , where y t is the predicted value at time t, a0,a1,a2..,a m is the model parameter, σ t is the error;

[0031] S333. Estimate the parameters of the polynomial model by least squares method or other appropriate fitting methods;

[0032] S334. Finally, the fitted polynomial model is used to predict the future time. The prediction formula is the polynomial expression of the model: in, Represents the predicted value at time t+k.

[0033] Preferably, step S3 uses a prediction method combining HS-GRU algorithm, grey model and trend extrapolation method: Among them, f1(x) is the prediction result of HS-GRU algorithm, f2(x) is the prediction result of grey model, and f3(x) is the prediction result of trend extrapolation method. is the specific gravity coefficient, satisfying

[0034] Compared with the prior art, the advantages and positive effects of the present invention are:

[0035] 1. Through the big data platform, we comprehensively collect and integrate information on various factors such as temperature, rainfall, PM2.5, permanent population, and meter voltage. We transmit the information collected by concentrators, smart meters and other equipment to the Internet of Things, and then to the production library and big data platform, ensuring the accuracy and comprehensiveness of the data, and providing a reliable foundation for subsequent analysis and decision-making.

[0036] 2. Using KPCA technology for feature dimensionality reduction has the advantages of processing nonlinear data, enhancing model capabilities, improving dimensionality reduction effects and flexibility, and can effectively deal with problems such as increased computational complexity caused by high-dimensional data sets, large training time and resource consumption, and failure of traditional distance metrics.

[0037] 3. The prediction method that combines the HS-GRU algorithm, the grey model and the trend extrapolation method is adopted to effectively reduce the prediction error and uncertainty caused by a single method. The advantages of each method complement each other in different situations, thus improving the stability and reliability of the overall prediction. DETAILED DESCRIPTION

[0038] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described below in conjunction with embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0039] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways than those described herein. Therefore, the present invention is not limited to the specific embodiments of the following disclosure.

[0040] Embodiment: Line loss management and load often occur in electric power energy applications. Traditional methods require a lot of manpower and material resources when facing such situations. In the era of big data, electric power energy big data plays a great role in promoting the development of the power industry. Through big data technology and artificial intelligence, it can accurately and quickly complete the prediction of line loss management and load conditions, which is beneficial to the safe production, power grid planning and provision of high-quality services of power companies. The first is data collection. It is necessary to collect factors such as temperature, rainfall, PM2.5, permanent population, and meter voltage through the big data platform, that is, to transmit the information collected by concentrators, smart meters and other equipment to the Internet of Things, and to transmit the data information to the production library and big data platform through calculation. The comprehensive collection and integration of information of various factors through the big data platform ensures the accuracy and comprehensiveness of the data, providing a reliable basis for subsequent analysis and decision-making.

[0041] Considering that high-dimensional data sets usually contain a large number of features, this will lead to an increase in computational complexity, thereby increasing training time and resource consumption. In addition, in high-dimensional space, the distance between data points will become less obvious, resulting in the failure of traditional distance metrics and affecting the effect of the model. Therefore, the present invention uses KPCA technology for dimensionality reduction. The use of KPCA technology for feature dimensionality reduction has the advantages of processing nonlinear data, enhancing model capabilities, improving dimensionality reduction effects and flexibility, making KPCA perform well in complex data analysis and processing tasks. First, standardize each feature in the data set so that its mean is 0 and its variance is 1; select the RBF kernel to construct the kernel matrix: where x i ,x j represents two data points, σ ​​is the bandwidth parameter of the RBF kernel function, and for each pair of data points, the RBF kernel function value is calculated to construct the kernel matrix K, whose element K ij = k(x i ,x j ) ; Centralize the RBF kernel matrix: K′=K-1 n K-K1 n +1 n K1 n , then perform eigenvalue decomposition on the centralized RBF kernel matrix: K′α i =λ i α i , where λ i is the eigenvalue, α i is the corresponding eigenvector; select the eigenvectors corresponding to the first 6 largest eigenvalues ​​as the new feature space after dimensionality reduction; map the original data to the new feature space to obtain the data after dimensionality reduction: Where T represents the eigenvector α i The transpose of .

[0042] In order to further increase the prediction accuracy of power grid load, the present invention proposes a prediction method combining HS-GRU algorithm, grey model and trend extrapolation method: Among them, f1(x) is the prediction result of HS-GRU algorithm, f2(x) is the prediction result of grey model, and f3(x) is the prediction result of trend extrapolation method. is the specific gravity coefficient, satisfying

[0043] The HS-GRU algorithm uses the harmony algorithm to optimize GRU. First, the harmony search parameters are initialized, including the harmony memory HMS, the harmony memory consideration rate HMCR, the pitch adjustment probability PAR and the adjustment range; the harmony memory is initialized, a set of candidate solutions, that is, the parameter set of the GRU model, is randomly generated, and they are stored in the harmony memory; the forward propagation of the GRU unit processes the input sequence through the standard GRU unit to obtain the hidden state sequence; each candidate solution in the harmony memory, that is, the GRU model parameter, is used to calculate the objective function value, and the objective function selects the prediction error; a new harmony vector is generated, and the parameter value is selected from the harmony memory with probability HMCR, and the parameter value is randomly selected from the parameter space with probability 1-HMCR, and the selected parameter value is fine-tuned with probability PAR; the harmony memory is updated, and if the objective function value of the generated new harmony vector is better than the worst candidate solution in the harmony memory, the worst candidate solution is replaced by the new harmony vector; finally, the generation of new harmony vectors and the updating of the harmony memory are repeated until the maximum number of iterations is reached or the objective function value converges.

[0044] The specific steps of using the GM(1,1) grey model for prediction are as follows: first, assume the original data sequence {x(1), x(2), ..., x(n)}; then generate the cumulative sequence {X(1), X(2), ..., X(n)}, where k is the number of data sequences; then the differential equation of the cumulative sequence is established: Among them, a and b are the parameters that need to be estimated; finally, the parameters a and b are estimated by the least squares method to solve the differential equation to obtain the prediction model.

[0045] The specific steps of the trend extrapolation method are to first collect and organize historical data to form a time series {y1,y2,...,y n}, where y i represents the observed value at time i; then the polynomial trend model is selected, and the data presents a polynomial trend: y t =a0+a1t+a2t 2 +...+a m t m +σ t , where y tis the predicted value at time t, a0,a1,a2..,a m is the model parameter, σ t is the error; the parameters of the polynomial model are estimated by the least squares method or other appropriate fitting methods; finally, the fitted polynomial model is used to predict the future moments, and the prediction formula is the polynomial expression of the model: in, Represents the predicted value at time t+k.

[0046] By combining the three methods, the prediction error and uncertainty caused by a single method can be effectively reduced. In addition, the advantages of each method in different situations complement each other, improving the stability and reliability of the overall prediction.

[0047] The above description is only a preferred embodiment of the present invention and does not limit the present invention in other forms. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for determining energy consumption indicators based on a big data platform, characterized in that: The following steps are involved: S1. The first step is data collection. To determine the energy consumption of power resources, it is necessary to collect factors such as temperature, rainfall, PM2.5, permanent population, and meter voltage; S2. Secondly, KPCA technology is used to reduce the dimension of features to increase the accuracy of prediction results; S3, then adopt the prediction method combining HS-GRU algorithm, grey model and trend extrapolation method, each of which accounts for a certain proportion. This comprehensive method can improve the prediction effect of the final power grid load; S4. Finally, use the big data platform to analyze the existing data and improve the management efficiency of line loss; The HS-GRU algorithm in step S3 shown above uses the harmony algorithm to optimize the GRU. The specific operations are as follows: S311, firstly, the harmony search parameters are initialized, including the harmony memory HMS, the harmony memory consideration rate HMCR, and the pitch adjustment probability PAR; S312, initializing the harmony memory, randomly generating a set of candidate solutions, that is, a parameter set of the GRU model, and storing them in the harmony memory; S313, forward propagation of the GRU unit, processes the input sequence through the standard GRU unit to obtain the hidden state sequence; S314, using each candidate solution in the harmony memory library, that is, the GRU model parameter, to calculate the objective function value, where the objective function selects the prediction error; S315, generating a new harmony vector, selecting a parameter value from the harmony memory bank with probability HMCR, randomly selecting a parameter value from the parameter space with probability 1-HMCR, and fine-tuning the selected parameter value with probability PAR; S316, updating the harmony memory library, if the objective function value of the generated new harmony vector is better than the worst candidate solution in the harmony memory library, replacing the worst candidate solution with the new harmony vector; S317. Finally, repeat steps S315 and S316 until the maximum number of iterations is reached or the objective function value converges.

2. The method for determining energy consumption indicators based on a big data platform according to claim 1, characterized in that: The step S1 transmits the information collected by the concentrator and the smart meter equipment to the Internet of Things, and transmits the data information to the production library and the big data platform through calculation as the data basis for subsequent work.

3. The method for determining energy consumption indicators based on a big data platform according to claim 1, characterized in that: The steps of using KPCA technology to perform feature dimensionality reduction processing in step S2 are: S21. Standardize each feature in the data set so that its mean is 0 and its variance is 1. S22. Select RBF kernel to construct kernel matrix: where x i ,x j represents two data points, σ ​​is the bandwidth parameter of the RBF kernel function, and for each pair of data points, the RBF kernel function value is calculated to construct the kernel matrix K, whose element K ij = k(x i ,x j ); S23. Centralize the RBF kernel matrix: K′=K-1 n K-K1 n +1 n K1 n , then perform eigenvalue decomposition on the centralized RBF kernel matrix: K′α i =λ i α i , where λ i is the eigenvalue, α i is the corresponding eigenvector; S24, select the eigenvectors corresponding to the first six largest eigenvalues ​​as the new feature space after dimensionality reduction; S25. Map the original data to the new feature space to obtain the reduced-dimensional data: Where T represents the eigenvector α i The transpose of .

4. The method for determining energy consumption indicators based on a big data platform according to claim 1, characterized in that: The specific steps of using the GM (1,1) grey model to perform prediction in step S3 are: S321. First, assume that the original data sequence is {x(1), x(2), ..., x(n)}; S322, then generate a cumulative sequence {X(1), X(2), ..., X(n)}, where k is the number of data sequences; S323, then establish the differential equation of the cumulative sequence: Where a and b are the parameters that need to be estimated; S324. Finally, the prediction model is obtained by solving the differential equation by estimating the parameters a and b using the least squares method.

5. The method for determining energy consumption indicators based on a big data platform according to claim 1, characterized in that: The specific steps of the trend extrapolation method in step S3 are: S331, first collect and organize historical data to form a time series {y1,y2,...,y n }, where y i represents the observation value at the i-th moment; S332, then select the polynomial trend model, the data presents a polynomial trend: y t =a0+a1t+a2t 2 +...+a m t m +σ t , where y t is the predicted value at time t, a0,a1,a2..,a m is the model parameter, σ t is the error; S333. Estimate the parameters of the polynomial model by least squares method or other appropriate fitting methods; S334. Finally, the fitted polynomial model is used to predict the future time. The prediction formula is the polynomial expression of the model: in, Represents the predicted value at time t+k.

6. The method for determining energy consumption indicators based on a big data platform according to claim 1, characterized in that: The step S3 uses a prediction method combining the HS-GRU algorithm, the grey model and the trend extrapolation method: Among them, f1(x) is the prediction result of HS-GRU algorithm, f2(x) is the prediction result of grey model, and f3(x) is the prediction result of trend extrapolation method. is the specific gravity coefficient, satisfying

Images