Household electricity consumption power load prediction and electricity saving method based on classification regression mapping
Through the improved interquartile range method and the constraint formula of power flow design, the objective function of household power load is established, and the problems of inaccuracy and inefficiency of household power load prediction in the prior art are solved, so as to achieve More accurate and efficient power consumption prediction and power saving effect.
Patent Information
- Application Number
- CN202411229381.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-03
- Publication Date
- 2024-12-31
AI Technical Summary
The existing household electricity load prediction methods have poor volatility, poor prediction effect, and low efficiency, and lack effective power saving methods.
Using a method based on classification regression mapping, the household electricity data is processed through the improved interquartile distance method, an improved short-term load prediction model is established, and the objective function of household electricity load is established in combination with the constraint formula designed by the power flow. Minimize electricity costs.
It improves the accuracy and efficiency of household power load prediction, reduces the variance and standard deviation of data, ensures the feasibility of the prediction results and the practicality of the optimized solution.
Smart Images
Figure CN119231488A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of electric power systems, and in particular to a household power load prediction and power saving method based on classification regression mapping. Background Art
[0002] With the vigorous development of smart grids, the high requirements for the safety of power grids, the economy of power transmission and the quality of power are becoming more and more stringent. There are still many problems with the large-scale storage of electric energy in the power system, so the production, transmission and consumption of electric energy are essentially carried out at the same time. Ensuring the continuity of transmission and distribution is the basis for maintaining the stable operation of the entire power system. In order to ensure the economic and stable operation of the power system, it is necessary to carry out detailed planning and analysis of the generation, transmission and distribution of electric energy, reduce unnecessary line losses and production consumption, and effectively improve economic benefits. Power load is an important basis for planning and analysis, but it is affected by many uncertain factors, including weather factors, holidays or household electricity consumption behavior, and the power load curve reflects periodicity and volatility, so power load prediction is still an urgent problem to be solved. According to the nature of the user, the power load can be roughly divided into agricultural power load, industrial power load, transportation power load and household life power load. Among them, according to surveys, household electricity consumption has exceeded one-third of the total social electricity consumption. Not only that, household electricity load is closely related to people's living standards, which also reflects the rapid economic development of society from the side. However, household electricity consumption is not only affected by weather factors and holidays. Users' living habits are also important reasons why household electricity load is highly volatile and difficult to predict. Household electricity load forecasting is an important part of power load forecasting. Accurate household electricity load forecasting can provide a basis for power distribution and transmission, and can also provide effective guarantees for the formulation of electricity prices in the power market. After reasonable prediction and planning, the power transmission process can reduce unnecessary power loss and improve power quality and economic benefits.
[0003] At present, researchers have conducted various studies on power load forecasting methods and proposed various methods to achieve power load forecasting for household electricity consumption.
[0004] For example, Cao Wenzhi et al. of Hunan University of Technology applied for a method and related equipment for predicting household power load in a multi-user scenario, which has been authorized on October 25, 2022. The patent application number is CN202211311699.4. The invention discloses a method and related equipment for predicting household power load in a multi-user scenario, including: obtaining a historical power load data set of household users, performing principal component analysis based on the historical power load data set of each household user to obtain an initial data set, and clustering the initial data set to obtain N types of pool user data, using the N types of pool user data to train the preset deep learning model respectively, and obtain N power load prediction models; determining the membership coefficient of the single household user data to be predicted in the N types of pool user data; using the power load prediction model to predict the power load of the single household user data to be predicted, and obtaining N types of prediction results corresponding to the single household user data to be predicted; using the membership coefficient to perform weighted fusion processing on the N types of prediction results to obtain the power load prediction value of the single household to be predicted. The invention can improve the accuracy of power load prediction.
[0005] There is also a residential user power load forecasting method that has been authorized and applied for by Li Qiqiang and others from Shandong University on May 4, 2017. The patent application number is CN201710309330.2. The invention discloses a residential user power load forecasting method, which aims to solve the problem that it is cumbersome for users to formulate electrical appliance work plans when participating in power demand response. The invention adopts a point-to-point ratio method to divide electrical appliances into three categories according to the usage rules of electrical appliances; the first category is for electrical appliances used throughout the day, and the power consumption of such appliances in the next day is predicted based on the hourly power consumption of the appliances in the previous three days; the second category is for electrical appliances with uncertain usage time and relatively stable power, and the usage time of such appliances is predicted based on the user's usage records on the same day last week and the last three days; the third category is for electrical appliances with large power changes and uncertain usage time, and the usage time and power consumption of such appliances are predicted based on the user's usage records on the same day last week and the last three days. The invention can effectively predict the use behavior and power consumption of electrical appliances in users' homes, and has the characteristics of small amount of historical data required and easy to implement, which effectively reduces the workload of users in formulating electrical appliance work plans.
[0006] Although many experts and scholars have conducted extensive research on household electricity load forecasting methods, such as the methods proposed in two related patents, few corresponding power-saving methods have been proposed for household electricity load forecasting models. In addition, most of the current household electricity load forecasting methods have defects, such as poor volatility of forecast data, poor power load forecasting effect, low power load forecasting efficiency, etc. Summary of the invention
[0007] In view of the above-mentioned defects of the prior art, the present invention provides a household electricity load forecasting and power saving method based on classification regression mapping, adopts an improved interquartile range method to process household electricity data, establishes an improved short-term load forecasting model in combination with household electricity data, establishes a constraint formula based on power flow design, and establishes an objective function of household electricity load according to the constraint formula, thereby minimizing the household electricity cost.
[0008] To achieve the above-mentioned purpose, the present invention adopts the following technical scheme: a household power load prediction and power saving method based on classification regression mapping, comprising:
[0009] The improved interquartile range method is used to process the data of household electricity consumption;
[0010] An improved short-term load forecasting model is established based on the processed household electricity consumption data;
[0011] Establish constraint formulas based on power flow design according to the short-term load forecasting model;
[0012] The objective function of household electricity load is established according to the constraint formula.
[0013] Furthermore, the improved interquartile range method is used to process household electricity data, specifically using an improved interquartile range method based on outlier detection, and the specific formula is expressed as follows:
[0014] IQR=Quart3-Quart1, LB=Quart1-1.5×IQR, UB=Quart3+1.5×IQR
[0015] Among them, Quart1 and Quart3 represent the first and third quartiles in the data respectively, IQR in the data represents the interquartile range, which is the data value of the first quartile subtracted from the third quartile, UB and LB are the upper and lower limits of the quartile data respectively;
[0016] Through the above method, outlier data is defined as data above the upper limit of the interquartile range and below the lower limit of the interquartile range, and is deleted from the data set; then the data processed by outliers is divided into five bands, and the data is predicted based on classification regression mapping combined with multivariate regression technology, and the final result is the data processing result of household electricity consumption.
[0017] Furthermore, the improved short-term load forecasting model is established based on the processed household electricity consumption data, specifically:
[0018] The feedforward neural network predicts the 24-hour load pattern of the d+1 day. Its input vector is in the form of 26 elements. The specific formula is x=[p(d,1),...,p(d,24),season,day], where p(d,i) represents the input data of the ith hour of the dth day, season represents the input data of the load pattern of different seasons in a year, and day represents the input data of the load pattern of weekdays and weekends. The 24-hour power consumption pattern of the d+1th day is expressed by the formula d=[p(d+1,1),...,p(d+1,24)];
[0019] The input attributes of these predictors are created using an autoencoder, which is a deep neural network using several hidden layers. The specific formula can be expressed as h=f(x), where the input data x is encoded as a reduced dimensionality vector; wherein the parameters of the network are adjusted to minimize the actual input vector and its reconstructed representation from the encoded vector;
[0020] The learning is repeated for subsequent hidden layers individually and the reconstruction part is removed from the structure. The dimension-reduced signal from the last hidden layer is considered as a diagnostic feature and serves as the input attribute of the feed-forward neural predictor.
[0021] Furthermore, a constraint formula based on power flow design is established according to the short-term load forecasting model, and the constraint formula based on power flow design takes microgrid generation, distributed generation, distributed energy storage, compensation system and feeder configuration into consideration, wherein:
[0022] The microgrid generation constraint is based on the interconnection line power capacity, and the distributed generation constraint is based on the output capacity constraint. The specific formula is expressed as Among them, P MG and P DG They are the power generation of microgrid and the power generation of distributed generation;
[0023] Distributed energy storage needs to meet the requirements of charging and discharging cycles. The specific constraint formula is expressed as Among them, P DS+ Indicates charging power, P DS- Indicates the discharge power, Q DS represents the power compensation resulting from energy loss, Indicates the maximum distributed energy storage supply;
[0024] The compensation system constraint is composed of reactive power compensation. The reactive power compensation of the system is expressed as Among them, Q STC Indicates reactive power compensation of the system;
[0025] The feeder configuration constraint formula is expressed as in, represents the feeder configuration power at time t, N tot represents the total feeder configuration power at time t;
[0026] The linear distribution flow equation is used to calculate the constraints based on power flow design, which is expressed as That is, the final household power load is obtained based on the constraint formula of power flow design.
[0027] Furthermore, the objective function of household electricity load is established according to the constraint formula, and the objective function is expressed as follows:
[0028]
[0029] in, represents the power generation cost of the microgrid at time t, represents the distributed generation cost at time t, represents the distributed energy storage cost at time t, represents the compensation system cost at time t, represents the feeder configuration cost at time t.
[0030] The present invention has the following technical effects:
[0031] 1. The present invention uses an improved interquartile range method based on outlier detection to process household electricity data, which can reduce the variance and standard deviation of the data, thereby more accurately establishing a household electricity load prediction model.
[0032] 2. The input data of the autoencoder neural network used in the present invention is created from the hourly data of the entire household electricity consumption, so it consists of a multi-element vector representing the data of the previous day. The autoencoder generates a reduced-dimensional diagnostic feature from the multi-element vector as the signal of the last hidden layer. Based on the input features passed to the neural predictor, the network creates a power forecast for household electricity consumption. This solution can predict the 24-hour household electricity load on all days. In contrast to ordinary feedforward neural networks, the feedforward neural network of the present invention is generated by an automatic self-organizing process with data representation formed by the autoencoder without human intervention, which can greatly improve the efficiency of short-term load forecasting.
[0033] 3. The present invention incorporates multiple power constraints into the constraint formula based on power flow design, which can directly ensure that the solution process meets all constraints, thereby improving the quality and accuracy of the solution. At the same time, it can also ensure that the solution obtained is feasible in multiple dimensions, thereby improving the practicality of the optimization solution. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] Figure 1 The present invention is a flow chart of a household electricity load prediction and power saving method based on classification regression mapping. DETAILED DESCRIPTION
[0035] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings to make the technical content clearer and easier to understand. The present invention can be embodied in many different forms of embodiments, and the protection scope of the present invention is not limited to the embodiments mentioned in the text.
[0036] The present invention proposes a household electricity load forecasting and power saving method based on classification regression mapping. Firstly, an improved interquartile range method is used to process household electricity data. Then, an improved short-term load forecasting model is established in combination with household electricity data. Then, a constraint formula based on power flow design is established. Finally, an objective function of household electricity load is established according to the constraint formula, thereby minimizing household electricity costs.
[0037] In view of the above technical problems in the prior art, the present invention provides a household power load prediction and power saving method based on classification regression mapping. The specific steps of the method can be as follows:
[0038] Step 1: Use the improved interquartile range method to process household electricity data
[0039] The present invention collects household electricity consumption data through a household electricity consumption management terminal and processes the data. In terms of data processing, an improved interquartile range method based on outlier detection is adopted. The specific formulas can be expressed as IQR=Quart3-Quart1, LB=Quart1-1.5×IQR, UB=Quart3+1.5×IQR, wherein Quart1 and Quart3 represent the first and third quartiles in the data respectively, IQR in the data represents the interquartile range, which is the data value of the first quartile subtracted from the third quartile, UB and LB are the upper and lower limits of the quartile data respectively, and outlier data is defined as data above the upper limit of the interquartile range and below the lower limit of the interquartile range in this paper, and is deleted from the data set; then the data processed by the outliers are divided into five bands, and the data are predicted according to the classification regression mapping combined with the multivariate regression technology.
[0040] The final result is the data processing result of household electricity consumption. Beneficial effect: This step uses the improved interquartile range method based on outlier detection to process household electricity consumption data, which can reduce the variance and standard deviation of the data, thereby more accurately establishing a household electricity load prediction model.
[0041] Step 2: Establish an improved short-term load forecasting model based on the household electricity consumption data obtained in step 1
[0042] The present invention uses an improved autoencoder feedforward neural network to perform short-term load forecasting on the household electricity consumption data obtained in step 1, and all these autoencoder networks share the same input attribute set, which is formed by information related to the past hourly power demand of the system. Since the typical prediction mathematical model needs to consider the input data of the 24-hour load pattern of the previous day, the load pattern of weekdays and weekends, and the load pattern of different seasons in a year. Therefore, the input vector of the feedforward neural network responsible for predicting the 24-hour load pattern of the d+1th day is in the form of 26 elements, and the specific formula can be expressed as x=[p(d,1),...,p(d,24),season,day], where p(d,i) represents the input data of the i-th hour of the dth day, season represents the input data of the load pattern of different seasons in a year, day represents the input data of the load pattern of weekdays and weekends, and the 24-hour power consumption pattern of the d+1th day can be expressed by the formula d=[p(d+1,1),...,p(d+1,24)]. Next, the input attributes of these predictors are created using an autoencoder, which is a deep neural network using several hidden layers. The role of subsequent layers is to gradually reduce the dimensionality of the input data. During the learning process, the input data x is encoded into a reduced-dimensional vector, which can be expressed as h = f (x). The parameters of the network are adjusted in a way that minimizes the actual input vector and its reconstructed representation from the encoded vector (auto-association mode). Then, the learning process is repeated for subsequent hidden layers separately, and the reconstruction part is removed from the structure. The reduced-dimensional signal from the last hidden layer is considered as a diagnostic feature and used as the input attribute of the feedforward neural predictor.
[0043] What is finally obtained is a short-term load forecasting model for household electricity consumption. Beneficial effect: The input data of the autoencoder neural network used in this step is created from the hourly data of the entire household electricity consumption, so it consists of a multi-element vector representing the data of the previous day. The autoencoder generates a reduced-dimensional diagnostic feature from the multi-element vector as the signal of the last hidden layer. Based on the input features passed to the neural predictor, the network creates a power forecast for household electricity consumption. This solution can predict 24-hour household electricity load on all days. In contrast to ordinary feedforward neural networks, the feedforward neural network of the present invention is generated by an automatic self-organizing process in which the data representation formed by the autoencoder is generated without human intervention, which can greatly improve the efficiency of short-term load forecasting.
[0044] Step 3: Establish the constraint formula based on power flow design based on the short-term load forecasting model obtained in step 2
[0045] The constraint formula designed by the present invention based on power flow design needs to take into account microgrid generation, distributed generation, distributed energy storage, compensation system and feeder configuration. The microgrid generation constraint is based on the interconnection line power capacity, while the distributed generation constraint is based on the output capacity constraint. The specific formula can be expressed as Among them, P MG and P DG They are the power generation of microgrid and the power generation of distributed power generation; distributed energy storage needs to meet the requirements of charging and discharging cycles. The specific constraint formula can be expressed as Among them, P DS+ Indicates charging power, P DS- Indicates the discharge power, Q DS represents the power compensation resulting from energy loss, represents the maximum distributed energy storage supply; the compensation system constraint is mainly composed of reactive power compensation, and the reactive power compensation of the system can be expressed as Among them, Q STC Represents the reactive power compensation of the system; in the constraint formula of the present invention, the main goal of feeder configuration is to manage power congestion, and radiality is the main criterion for ensuring feeder configuration. Whenever the household power management terminal controls the resident electrical appliances to work in a power-saving state or staggers the power grid power peak to delay power consumption, the radiality condition must be evaluated. If the new state change of the system is essentially radial, the feeder configuration constraint must be included in the constraint formula of the power flow design. The specific formula can be expressed as in, represents the feeder configuration power at time t, N tot represents the total feeder configuration power at time t. The linear distributed flow equation is used to calculate the constraints based on power flow design, which can be expressed as
[0046] The final result is the constraint formula of household power load based on power flow design. Beneficial effect: This step incorporates multiple power constraints into the constraint formula based on power flow design, which can directly consider ensuring that the solution process meets all constraints, thereby improving the quality and accuracy of the solution, and also ensuring that the solution obtained is feasible in multiple dimensions, thereby improving the practicality of the optimization solution.
[0047] Step 4: Establish the objective function of household electricity load for the constraint formula obtained in step 3, so as to minimize the household electricity cost.
[0048] The main goal of the present invention is to reduce the operating cost within the allowable constraints of the predicted load, thereby minimizing the household electricity cost. The specific objective function can be expressed as in, represents the power generation cost of the microgrid at time t, represents the distributed generation cost at time t, represents the distributed energy storage cost at time t, represents the compensation system cost at time t, represents the feeder configuration cost at time t. The final result is the objective function of household power load.
[0050] In summary, the present invention uses an improved interquartile range method based on outlier detection to process household electricity data, which can reduce the variance and standard deviation of the data, thereby more accurately establishing a household electricity load forecasting model. The data representation formed by the feedforward neural network of the present invention is generated by an automatic self-organizing process without human intervention, which can greatly improve the efficiency of short-term load forecasting. Incorporating multiple power constraints into the constraint formula based on power flow design can directly consider ensuring that the solution process meets all constraints, thereby improving the quality and accuracy of the solution, and at the same time ensuring that the solution obtained is feasible in multiple dimensions, thereby improving the practicality of the optimized solution.
[0051] The preferred specific embodiments of the present invention are described in detail above. It should be understood that ordinary technicians in the field can make many modifications and changes based on the concept of the present invention without creative work. Therefore, all technical solutions that can be obtained by technicians in the technical field based on the concept of the present invention through logical analysis, reasoning or limited experiments on the basis of the prior art should be within the scope of protection determined by the claims.
Claims
1. A household power load prediction and power saving method based on classification regression mapping, characterized in that: include: The improved interquartile range method is used to process the data of household electricity consumption; An improved short-term load forecasting model is established based on the processed household electricity consumption data; Establish constraint formulas based on power flow design according to the short-term load forecasting model; The objective function of household electricity load is established according to the constraint formula.
2. The household power load prediction and power saving method based on classification regression mapping according to claim 1 is characterized in that: The improved interquartile range method is used to process household electricity data. Specifically, the improved interquartile range method based on outlier detection is used. The specific formula is: IQR=Quart3-Quart1, LB=Quart1-1.5×IQR, UB=Quart3+1.5×IQR Among them, Quart1 and Quart3 represent the first and third quartiles in the data respectively, IQR in the data represents the interquartile range, which is the data value of the first quartile subtracted from the third quartile, UB and LB are the upper and lower limits of the quartile data respectively; Through the above method, outlier data is defined as data above the upper limit of the interquartile range and below the lower limit of the interquartile range, and is deleted from the data set; then the data processed by outliers is divided into five bands, and the data is predicted based on classification regression mapping combined with multivariate regression technology, and the final result is the data processing result of household electricity consumption.
3. The household power load prediction and power saving method based on classification regression mapping according to claim 1 is characterized in that: The improved short-term load forecasting model is established based on the processed household electricity consumption data, specifically: The feedforward neural network predicts the 24-hour load pattern of the d+1th day, and its input vector is in the form of 26 elements. The specific formula is x=[p(d, 1), ..., p(d, 24), season, day], where p(d, i) represents the input data of the i-th hour of the dth day, season represents the input data of the load pattern of different seasons in a year, and day represents the input data of the load pattern of weekdays and weekends. The 24-hour power consumption pattern of the d+1th day is expressed by the formula d=[p(d+1, 1), ..., p(d+1, 24)]; The input attributes of these predictors are created using an autoencoder, which is a deep neural network using several hidden layers, specifically represented by the formula h=f(x), where the input data x is encoded into a reduced dimensionality vector; wherein the parameters of the network are adjusted in such a way as to minimize the actual input vector and its reconstructed representation from the encoded vector; The learning is repeated for subsequent hidden layers individually and the reconstruction part is removed from the structure. The dimension-reduced signal from the last hidden layer is considered as a diagnostic feature and serves as the input attribute of the feed-forward neural predictor.
4. The household power load prediction and power saving method based on classification regression mapping according to claim 1 is characterized in that: The constraint formula based on power flow design is established according to the short-term load forecasting model. The constraint formula based on power flow design takes microgrid generation, distributed generation, distributed energy storage, compensation system and feeder configuration into consideration, wherein: The microgrid generation constraint is based on the interconnection line power capacity, and the distributed generation constraint is based on the output capacity constraint. The specific formula is expressed as Among them, P MG and P DG They are the power generation of microgrid and the power generation of distributed generation; Distributed energy storage needs to meet the requirements of charging and discharging cycles. The specific constraint formula is expressed as Among them, P DS+ Indicates charging power, P DS- Indicates the discharge power, Q DS represents the power compensation resulting from energy loss, Indicates the maximum distributed energy storage supply; The compensation system constraint is composed of reactive power compensation. The reactive power compensation of the system is expressed as Among them, Q STC Indicates reactive power compensation of the system; The feeder configuration constraint formula is expressed as in, represents the feeder configuration power at time t, N tot represents the total feeder configuration power at time t; The linear distribution flow equation is used to calculate the constraints based on power flow design, which is expressed as That is, the final household power load is obtained based on the constraint formula of power flow design.
5. The household power load prediction and power saving method based on classification regression mapping according to claim 4 is characterized in that: The objective function of household electricity load is established according to the constraint formula, and the objective function is expressed by the formula: in, represents the power generation cost of the microgrid at time t, represents the distributed generation cost at time t, represents the distributed energy storage cost at time t, represents the compensation system cost at time t, represents the feeder configuration cost at time t.