Energy consumption data frequency conversion method and system based on electric power data

Abstract

The invention discloses an energy consumption data frequency conversion method based on electric power data. The method comprises the following steps: acquiring annual energy consumption data of a target user; performing data cleaning on the obtained energy consumption data; calculating monthly energy loss data according to the obtained cleaned data; standardized annual energy loss data is calculated according to the monthly energy loss data, and energy consumption data frequency conversion based on the electric power data is completed. The invention also discloses a system for realizing the power data-based energy consumption data frequency conversion method. According to the invention, through collection, cleaning and calculation of the energy consumption data, the frequency conversion of the energy consumption data based on the electric power data is realized, and the reliability is high and the accuracy is good.

Description

Technical Field

[0001] The present invention belongs to the field of electrical automation, and in particular relates to a method and system for frequency conversion of energy consumption data based on power data. Background Art

[0002] With the increasing energy consumption and the diversification of energy structure, energy management has become increasingly complex. In order to manage energy more effectively, it is necessary to monitor and analyze various energy consumption data accurately and in real time. As an important part of energy consumption data, the study of frequency conversion method of power data is of great significance to improving energy management efficiency.

[0003] At present, with the rapid development of smart grid technology, smart grid has realized real-time monitoring and intelligent dispatching of power networks by integrating advanced communication, information technology and control technology, and also obtained massive energy data. However, at this stage, there is indeed a lack of a frequency conversion scheme for energy consumption data based on power data, which makes it impossible to convert massive data of the power system into energy consumption data, thus affecting the accuracy and efficiency of energy management. Summary of the invention

[0004] One of the objectives of the present invention is to provide a method for frequency conversion of energy consumption data based on power data with high reliability and good accuracy.

[0005] A second objective of the present invention is to provide a system for implementing the method for converting the frequency of energy consumption data based on power data.

[0006] The method for converting the frequency of energy consumption data based on power data provided by the present invention comprises the following steps:

[0007] S1. Obtain annual energy consumption data of target users;

[0008] S2. Cleaning the energy consumption data obtained in step S1;

[0009] S3. Calculate monthly energy consumption data based on the cleaned data obtained;

[0010] S4. Based on the monthly energy loss data obtained in step S3, the standardized annual energy loss data is calculated to complete the frequency conversion of energy consumption data based on power data.

[0011] The step S1 of obtaining the annual energy consumption data of the target user specifically includes the following steps:

[0012] From the energy system, obtain energy consumption data;

[0013] The energy system includes an energy consumption management cloud platform, an energy Internet of Things platform and a power system;

[0014] The energy consumption data mentioned above include annual carbon emission data, annual industry energy consumption data and monthly electricity data;

[0015] The monthly power data includes monthly power consumption data, monthly power consumption time data and monthly power consumption equipment data.

[0016] Step S2 of cleaning the energy consumption data obtained in step S1 specifically includes the following steps:

[0017] Performing data cleaning on the energy consumption data obtained in step S1;

[0018] The data cleaning includes outlier cleaning, data conversion and data integration.

[0019] The step S3 of calculating the monthly energy consumption data based on the cleaned data specifically includes the following steps:

[0020] The monthly energy loss data is calculated using the following formula and quadratic optimization algorithm:

[0021]

[0022] In the formula Monthly energy consumption data for users; is the standardized monthly electricity consumption data; A is the coefficient matrix; m represents each month; γ represents the user energy data; Is a mathematical operator used to find the value of the independent variable that minimizes a given function.

[0023] The quadratic energy optimization algorithm specifically includes batch gradient descent algorithm, stochastic gradient descent algorithm, small batch gradient descent algorithm, Newton's method, DFP method, BFGS method, Lagrange multiplier method, penalty function method and interior point method.

[0024] Step S4, according to the monthly energy loss data obtained in step S3, calculates the standardized annual energy loss data, which specifically includes the following steps:

[0025] The standardized annual energy loss is calculated using the following formula:

[0026]

[0027] Where y Y,γ is the standardized annual energy consumption data; t,γ is the standardized energy consumption data for month t; Y represents each year.

[0028] The present invention also provides a system for implementing the method for frequency conversion of energy consumption data based on power data, comprising a data acquisition module, a data cleaning module, a loss calculation module and a frequency conversion module; the data acquisition module, the data cleaning module, the loss calculation module and the frequency conversion module are connected in series in sequence; the data acquisition module is used to acquire the annual energy consumption data of the target user, and upload the data information to the data cleaning module; the data cleaning module is used to clean the acquired energy consumption data according to the received data information, and upload the data information to the loss calculation module; the loss calculation module is used to calculate the monthly energy loss data according to the received data information and the obtained cleaned data, and upload the data information to the frequency conversion module; the frequency conversion module is used to calculate the standardized annual energy loss data according to the received data information and the obtained monthly energy loss data, and complete the frequency conversion of energy consumption data based on power data.

[0029] The method and system for converting the frequency of energy consumption data based on power data provided by the present invention not only realizes the frequency conversion of energy consumption data based on power data through the collection, cleaning and calculation of energy consumption data, but also has high reliability and good accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 The figure is a schematic diagram of the method flow of the present invention.

[0031] Figure 2 A schematic diagram comparing actual and frequency-converted energy consumption of large-scale thermal power plants according to an embodiment of the method of the present invention.

[0032] Figure 3 Schematic diagram of the functional modules of the system of the present invention. DETAILED DESCRIPTION

[0033] like Figure 1 The method flow chart of the method of the present invention is shown as follows: the method for converting the frequency of energy consumption data based on power data disclosed in the present invention comprises the following steps:

[0034] S1. Obtain the annual energy consumption data of the target user; the step includes the following steps:

[0035] From the energy system, obtain energy consumption data;

[0036] The energy system includes an energy consumption management cloud platform, an energy Internet of Things platform and a power system;

[0037] The energy consumption data mentioned above include annual carbon emission data, annual industry energy consumption data and monthly electricity data;

[0038] The monthly power data includes monthly power consumption data, monthly power consumption time data and monthly power consumption equipment data;

[0039] S2. Clean the energy consumption data obtained in step S1; specifically, the following steps are included:

[0040] Performing data cleaning on the energy consumption data obtained in step S1;

[0041] The data cleaning includes outlier cleaning, data conversion and data integration;

[0042] S3. Calculate monthly energy consumption data based on the cleaned data; specifically include the following steps:

[0043] The monthly energy loss data is calculated using the following formula and quadratic optimization algorithm:

[0044]

[0045] In the formula Monthly energy consumption data for users; is the standardized monthly electricity consumption data; A is the coefficient matrix; m represents each month; γ represents the user energy data; Is a mathematical operator used to find the value of the independent variable that minimizes a given function.

[0046] Among them, the quadratic energy optimization algorithm specifically includes batch gradient descent algorithm, stochastic gradient descent algorithm, mini-batch gradient descent algorithm, Newton method, DFP method, BFGS method, Lagrange multiplier method, penalty function method and interior point method;

[0047] In practice, a quadratic form is a quadratic homogeneous polynomial about a variable, that is, it contains terms of the variable and its square, but does not contain higher powers of the variable. It can be expressed as Q(x)=x T Ax, where A is a symmetric matrix. Through orthogonal transformation, any quadratic form can be transformed into a standard form, that is, a diagonal matrix form in which the coefficients of the square terms are 1 or -1. The standard form reflects the essential characteristics of the quadratic form, such as positive definiteness and rank. If for any non-zero vector x, Q(x)>0, the quadratic form is called positive definite. The necessary and sufficient condition for a quadratic form to be positive definite is that all coefficients of its standard form are positive.

[0048] Quadratic optimization problems can be divided into unconstrained optimization problems and constrained optimization problems.

[0049] Unconstrained optimization problem: The objective function is a quadratic form and there are no constraints. The solutions to this type of problem include analytical methods and numerical methods. The analytical method finds the optimal solution by solving the extreme points of the quadratic form, while the numerical method finds the optimal solution through iterative approximation.

[0050] Constrained optimization problem: The objective function is a quadratic form and is constrained by a series of linear or nonlinear equations or inequalities. The solution to this type of problem is more complicated and requires the use of special algorithms such as the Lagrange multiplier method, penalty function method, and interior point method.

[0051] Unconstrained Optimization Algorithm:

[0052] (1) Gradient descent method: Gradient descent method is a commonly used unconstrained optimization algorithm that approximates the minimum value of a function by iteratively updating parameters. In each iteration, the gradient of the function is calculated and the parameters are updated according to the gradient. This process continues until the value of the function reaches an acceptable threshold or the number of iterations reaches a preset upper limit. The core idea of ​​the gradient descent method is to update the parameters in the opposite direction of the objective function's gradient (that is, the direction in which the function value decreases fastest) to gradually approach the minimum value of the objective function in an iterative manner. The "gradient" here is the direction in which the multivariate function changes fastest at a certain point, and the opposite direction of the gradient is the direction in which the function value decreases fastest.

[0053] The algorithm steps are as follows:

[0054] a. Choose an initial value for the parameter, usually a small random number or zero.

[0055] b. Calculate the gradient vector based on the objective function and the current parameter value. Each element in the gradient vector corresponds to a partial derivative of the parameter.

[0056] c. Update the parameter value in the opposite direction of the gradient, usually using the formula θ = θ - α * ▽f(θ), where θ represents the parameter, α is the learning rate (step size), and ▽f(θ) is the gradient of the objective function f(θ) at θ.

[0057] d. Check whether the termination conditions are met, such as reaching the preset number of iterations, the change in the loss function value is less than a certain threshold, etc. If the termination conditions are not met, return to step b to continue iterating; otherwise, output the final parameter value.

[0058] It should be noted that there are many variants of the gradient descent method to adapt to different application scenarios and data set sizes:

[0059] (1) Batch Gradient Descent: In each iteration, the entire training dataset is used to calculate the gradient and update the parameters. This method can guarantee convergence to the global optimal solution (when the loss function is a convex function), but the computational complexity is large and the computational cost is high for large-scale datasets.

[0060] (2) Stochastic Gradient Descent (SGD): In each iteration, a sample is randomly selected to calculate the gradient and update the parameters. This method is fast, but the gradient estimation may be inaccurate, resulting in unstable parameter updates, which may oscillate around the optimal solution.

[0061] (3) Mini-batch Gradient Descent: It combines the advantages of batch gradient descent and stochastic gradient descent, using a small batch of samples (usually dozens to hundreds) to calculate the gradient and update the parameters each time. This method achieves a good balance between computational efficiency and parameter update stability.

[0062] It should be noted that the learning rate has a great impact on the performance of the gradient descent method. A learning rate that is too large may cause unstable parameter updates or even divergence; a learning rate that is too small may lead to slow convergence. Therefore, it is necessary to choose a suitable learning rate. Before performing gradient descent, it is usually necessary to scale the features to ensure that different features are on the same scale. This helps to speed up convergence and improve the stability of the algorithm. In practical applications, you can judge whether the algorithm has converged by observing the changes in the loss function value. If the loss function value no longer decreases significantly after multiple iterations, it can be considered that the algorithm has converged.

[0063] (2) Newton's method: Newton's method is an optimization algorithm based on the second-order derivative of a function. It finds the extreme points of a function by solving the second-order derivative of the function (i.e., the Hessian matrix). Newton's method has the advantage of fast convergence speed, but calculating the Hessian matrix and solving the linear equations may be complicated. The core idea of ​​Newton's method is to perform Taylor series expansion on the function and find the optimal solution of the function by continuously approaching the root or minimum point of the function.

[0064] For a univariate function f(x), perform a Taylor series expansion at point x0, ignore quadratic terms and above, and approximate the nonlinear equation to a linear equation. For a multivariate function, a second-order Taylor series expansion can be used to approximate the function. By solving the linear equations, we get the update direction dx (or Δθ, in the case of multiple variables). Update the current point x (or θ), that is where H is the Hessian matrix (in the multivariate case) and ▽f is the gradient vector.

[0065] (3) Quasi-Newton method: The quasi-Newton method is an unconstrained optimization method that uses the gradient information of a function to simulate the iteration direction of the Newton method. This method has the advantages of fast convergence speed and no need to calculate second-order derivatives, so it is suitable for large-scale problems. The basic idea of ​​the quasi-Newton method is to approximate the inverse matrix of the Hessian matrix in the Newton method (that is, the second-order derivative matrix of the objective function) with a matrix that does not contain second-order derivatives, thereby avoiding the calculation of complex second-order derivatives. Through continuous iteration, the quasi-Newton method can gradually approximate the solution of nonlinear equations.

[0066] Specifically, the quasi-Newton method constructs a quadratic model of the objective function in each iteration, and takes the optimal solution of the quadratic model as the search direction to obtain a new iteration point. The key to this process is how to update the approximate Hessian matrix (or its inverse matrix) in each iteration. The quasi-Newton method updates this approximate matrix by using the information of the previous iteration, as well as the gradient and function value of the current iteration point.

[0067] The quasi-Newton method formula is the core expression of the algorithm, which is used to update the numerical value of the iterative solution. Its general form is:

[0068] x(k+1)=x(k)-[H(k)] -1 *g(k)

[0069] Among them, x(k) represents the solution vector of the kth iteration; x(k+1) represents the solution vector of the k+1th iteration; g(k) represents the gradient vector of the kth iteration; H(k) represents the quasi-Newton matrix of the kth iteration, that is, the approximation of the inverse matrix of the Hessian matrix. The key to the quasi-Newton method is how to choose the quasi-Newton matrix H. Commonly used quasi-Newton methods include the DFP method and the BFGS method. Both methods are based on the recursive formula of the inverse Hessian matrix that satisfies the quasi-Newton condition.

[0070] The recursive formula of the DFP method is:

[0071] H(k+1)=H(k)+(s(k)*s(k)') / (s(k)'*y(k))

[0072] -(H(k)*y(k)*y(k)'*H(k)) / (y(k)'*H(k)*y(k))

[0073] Among them, s(k)=x(k+1)-x(k) represents the change of the iterative solution, and y(k)=g(k+1)-g(k) represents the change of the gradient.

[0074] The recursive formula of the BFGS method is:

[0075] H(k+1)=H(k)+(1+(y(k)'*H(k)*y(k)) / (y(k)'*s(k)))*(s(k)*s(k)') / (y(k)'*s(k))

[0076] =-(H(k)*y(k)*s(k)'-s(k)*y(k)'*H(k)) / (y(k)'*s(k))

[0077] Similarly, s(k) and y(k) represent the change in the iterative solution and the change in the gradient, respectively.

[0078] Constrained optimization algorithm:

[0079] (1) Lagrange multiplier method: Lagrange multiplier method is an algorithm for solving constrained optimization problems. It integrates the objective function and constraints into a new function (i.e., Lagrange function), and then solves the extreme points of the Lagrange function. In the solution process, Lagrange multipliers are needed to transform the constraints into unconstrained conditions.

[0080] The basic principle of the Lagrange multiplier method is to introduce a new parameter (Lagrange multiplier) to link the constraint condition function with the original function to form a new function (Lagrange function). By solving the extreme value problem of the Lagrange function, the extreme value solution of the original function under the constraint conditions can be obtained.

[0081] Let the objective function be f(x) and the constraint be g(x) = 0 (multiple constraints can be handled similarly). The Lagrangian function L(x,λ) is defined as L(x,λ) = f(x) + λg(x). Finding the gradient of the Lagrangian function L(x,λ) yields Let the gradient be zero, that is At the same time, consider the constraint condition g(x) = 0, forming a system of equations. Solve this system of equations to obtain the values ​​of x and λ. Among them, the value of x is the extreme value solution of the original function under the constraint conditions.

[0082] It is important to note that when constructing the Lagrangian function, make sure that the constraints are properly incorporated into the function. When solving for the gradient, take partial derivatives with respect to all variables. When setting up the system of equations, make sure that all conditions are taken into account. When solving the system of equations, you may need to use numerical methods or specific mathematical software to solve it.

[0083] (2) Penalty function method: The penalty function method is a method for transforming a constrained optimization problem into an unconstrained optimization problem. It transforms a constrained optimization problem into an unconstrained optimization problem by introducing a penalty function to transform the constraint into a part of the objective function. The value of the penalty function is related to the degree of violation of the constraint, so the optimization process can be controlled by adjusting the parameters of the penalty function.

[0084] The traditional penalty function method is generally divided into the external penalty function method and the internal penalty function method (also called the barrier penalty function method). The external penalty function method refers to constructing a penalty function based on the characteristics of the constraint, and then adding it to the objective function to transform the constraint problem into a series of unconstrained problems. This "penalty strategy" penalizes those target points that attempt to violate the constraints in the process of solving the unconstrained problem. The internal penalty function method searches inside the feasible domain, and the constraint boundary plays a role similar to a fence. If the current solution is far away from the constraint boundary, the penalty function value is very small; otherwise, the penalty function value is close to infinity.

[0085] For constrained optimization problems, it can be expressed as follows:

[0086] minimize f(x)subject to g(x)≤0,h(x)=0

[0087] Among them, f(x) is the objective function, g(x) and h(x) are constraint functions.

[0088] The general form of the penalty function can be expressed as:

[0089] F(x,ρ)=f(x)+ρ*p(x)

[0090] Among them, F(x,ρ) is the penalty function, ρ is the penalty coefficient of the penalty function (also called the penalty factor), and p(x) is the penalty term used to penalize solutions that violate the constraints.

[0091] It is worth noting that the external penalty function method is a method that starts from an infeasible solution and gradually moves to the feasible region. Its general form can be expressed as:

[0092] B(x)=f(x)+Σr i *G i +Σc j *H j

[0093] Among them, B(x) is the new objective function in the optimization process, G i and H j The constraints g are i (x) and h j (x), r i and c j is a constant, called the penalty factor. i and H j The most common forms are:

[0094] G i =max[0,g i (x)] a

[0095] H j =|h j (x)| b

[0096] Among them, a and b are generally 1 or 2.

[0097] The internal penalty function method searches inside the feasible domain, and the constraint boundary acts like a fence. If the current solution is far away from the constraint boundary, the penalty function value is very small; otherwise, the penalty function value is close to infinity. The specific formula of the internal penalty function method may vary depending on the constraint conditions.

[0098] (3) Interior point method: The interior point method is an algorithm for solving constrained optimization problems. It finds the optimal solution by iterating within the feasible domain. The interior point method has the advantages of fast convergence speed and insensitivity to the initial point, but it requires solving a complex set of nonlinear equations.

[0099] For the optimization problem with inequality constraints, the general expression of the penalty function constructed by the interior point method is:

[0100] minf(x),x∈R n

[0101] stg i (x)≤0,i=1,2,...,m

[0102] Among them, f(x) is the objective function, g i (x) is an inequality constraint.

[0103] S4. Based on the monthly energy loss data obtained in step S3, the standardized annual energy loss data is calculated to complete the frequency conversion of energy consumption data based on power data; specifically, the steps include:

[0104] The standardized annual energy loss is calculated using the following formula:

[0105]

[0106] Where y Y,γ is the standardized annual energy loss data, which is represented by y Y,γ =[y 1,γ ,y 2,γ ,...,y 12,γ ],y 1,γ ~y 12,γ are the standardized energy consumption data from January to December; y t,γ is the standardized energy consumption data for month t; Y represents each year.

[0107] The method of the present invention can convert the annual energy consumption data directly obtained in step S1 into the final standardized monthly energy consumption data based on the rules of the monthly electricity consumption data of the power system, thereby completing the frequency conversion of the data dimension from year to month.

[0108] The method of the present invention is further described below in conjunction with an embodiment:

[0109] First, select the user's energy consumption data, including the user's annual industry energy consumption data and annual carbon emission data. In this example, the annual energy consumption data of a large thermal power plant in a certain province from 2021 to 2022 is used, with one data point per month, that is, 12 data points per year.

[0110] When conducting experimental comparisons, this example uses three model evaluation indicators, namely mean square error (MSE), root mean square error (RMSE), and mean relative error (MRE), specifically:

[0111] Mean Square Error:

[0112] Root Mean Square Error:

[0113] Average relative error: In the formula, n is the number of predictions, Y i represents the predicted value, Indicates actual value.

[0114] The evaluation indicators of the total energy consumption data frequency conversion method are shown in Table 1:

[0115] Table 1 Evaluation index table

[0116] index Numeric MSE <![CDATA[1.5167×10 3 ]]> RMSE 38.945 MRE 0.15467

[0117] The original user energy consumption data is input into the frequency conversion algorithm to obtain the user's energy activity. Through the quadratic optimization algorithm, the user's annual industry energy consumption data and annual carbon emission data are converted into monthly industry energy consumption data and monthly carbon emission data. Figure 2 This is a comparison of the actual energy consumption of large-scale thermal power plants in the province and the energy consumption converted by frequency.

[0118] Through Table 1 and Figure 2 It can be seen that the solution of the present invention has good conversion accuracy, reliability and precision.

[0119] like Figure 3The figure shows a schematic diagram of the functional modules of the system of the present invention: the system disclosed in the present invention for realizing the method for frequency conversion of energy consumption data based on power data comprises a data acquisition module, a data cleaning module, a loss calculation module and a frequency conversion module; the data acquisition module, the data cleaning module, the loss calculation module and the frequency conversion module are connected in series in sequence; the data acquisition module is used to acquire the annual energy consumption data of the target user, and upload the data information to the data cleaning module; the data cleaning module is used to clean the acquired energy consumption data according to the received data information, and upload the data information to the loss calculation module; the loss calculation module is used to calculate the monthly energy loss data according to the received data information and the obtained cleaned data, and upload the data information to the frequency conversion module; the frequency conversion module is used to calculate the standardized annual energy loss data according to the received data information and the obtained monthly energy loss data, and complete the frequency conversion of energy consumption data based on power data.

Claims

1. A method for converting energy consumption data frequency based on power data, comprising the following steps: S1. Obtain annual energy consumption data of target users; S2. Cleaning the energy consumption data obtained in step S1; S3. Calculate monthly energy consumption data based on the cleaned data obtained; S4. Based on the monthly energy loss data obtained in step S3, the standardized annual energy loss data is calculated to complete the frequency conversion of energy consumption data based on power data.

2. The method for converting energy consumption data frequency based on power data according to claim 1 is characterized in that The step S1 of obtaining the annual energy consumption data of the target user specifically includes the following steps: From the energy system, obtain energy consumption data; The energy system includes an energy consumption management cloud platform, an energy Internet of Things platform and a power system; The energy consumption data mentioned above include annual carbon emission data, annual industry energy consumption data and monthly electricity data; The monthly power data includes monthly power consumption data, monthly power consumption time data and monthly power consumption equipment data.

3. The method for converting energy consumption data frequency based on power data according to claim 2 is characterized in that Step S2 of cleaning the energy consumption data obtained in step S1 specifically includes the following steps: Performing data cleaning on the energy consumption data obtained in step S1; The data cleaning includes outlier cleaning, data conversion and data integration.

4. The method for converting energy consumption data frequency based on power data according to claim 3 is characterized in that The step S3 of calculating the monthly energy consumption data based on the cleaned data specifically includes the following steps: The monthly energy loss data is calculated using the following formula and quadratic optimization algorithm: In the formula Monthly energy consumption data for users; is the standardized monthly electricity consumption data; A is the coefficient matrix; m represents each month; γ represents the user energy data; Is a mathematical operator used to find the value of the independent variable that minimizes a given function.

5. The method for converting energy consumption data frequency based on power data according to claim 4 is characterized in that The quadratic energy optimization algorithm specifically includes batch gradient descent algorithm, stochastic gradient descent algorithm, small batch gradient descent algorithm, Newton method, DFP method, BFGS method, Lagrange multiplier method, penalty function method and interior point method.

6. The method for converting energy consumption data frequency based on power data according to claim 5, characterized in that Step S4, according to the monthly energy loss data obtained in step S3, calculates the standardized annual energy loss data, which specifically includes the following steps: The standardized annual energy loss is calculated using the following formula: Where y Y,γ is the standardized annual energy consumption data; t,γ is the standardized energy consumption data for month t; Y represents each year.

7. A system for implementing the method for converting energy consumption data frequency based on power data as claimed in any one of claims 1 to 6, characterized in that It includes a data acquisition module, a data cleaning module, a loss calculation module and a frequency conversion module; the data acquisition module, the data cleaning module, the loss calculation module and the frequency conversion module are connected in series in sequence; the data acquisition module is used to obtain the annual energy consumption data of the target user and upload the data information to the data cleaning module; The data cleaning module is used to clean the acquired energy consumption data according to the received data information, and upload the data information to the loss calculation module; The loss calculation module is used to calculate the monthly energy loss data based on the received data information and the obtained cleaned data, and upload the data information to the frequency conversion module; the frequency conversion module is used to calculate the standardized annual energy loss data based on the received data information and the obtained monthly energy loss data, and complete the frequency conversion of energy consumption data based on electricity data.

Images