Electricity selling management system based on data analysis

Abstract

The invention belongs to the field of power sales management, and discloses a data analysis-based power sales management system, which comprises a power consumption acquisition module and a prediction module, the electricity consumption acquisition module is used for acquiring electricity consumption of a user; the prediction module comprises a data preprocessing unit and a prediction unit; the data preprocessing unit is used for performing data cleaning processing on the electricity consumption to obtain the electricity consumption subjected to data cleaning; and the prediction unit is used for analyzing the power consumption data subjected to data cleaning to obtain predicted power consumption of the user. According to the method, the number of the contrast sequences is calculated based on the L, a plurality of contrast sequences are set, different influence weights are adaptively calculated for the calculation results of different contrast sequences, and more accurate abnormal value removal processing is performed on the electricity consumption by integrating the calculation results of the contrast sequences with different lengths, so that the accuracy of the abnormal value removal processing is improved. The data change characteristics in the electricity consumption obtained after abnormal value removal processing can be more effectively reserved.

Description

Technical Field

[0001] The present invention relates to the field of power sales management, and in particular to a power sales management system based on data analysis. Background Art

[0002] The power sales management system obtains the user's power consumption data, analyzes the user's power consumption data, and realizes power consumption forecast based on the analysis results, so that the power sales company can provide power providers with power consumption requirements based on the power consumption forecast, so that the power provider can supply power more accurately and reduce power waste.

[0003] Before making a prediction, it is usually necessary to clean the electricity consumption data and remove outliers. The existing removal method is generally based on a fixed range of electricity consumption data to obtain the average value. This method has certain disadvantages. That is, when the electricity consumption data changes greatly in a short period of time, it is easy to lose too much of the change characteristics of electricity consumption after removing outliers, thereby affecting the accuracy of the predicted electricity consumption results. Summary of the invention

[0004] The purpose of the present invention is to disclose a power sales management system based on data analysis to solve the technical problems raised in the background technology.

[0005] In order to achieve the above object, the present invention adopts the following technical solution:

[0006] The present invention provides a power sales management system based on data analysis, including a power consumption collection module and a prediction module;

[0007] The power consumption collection module is used to obtain the user's power consumption;

[0008] The prediction module includes a data preprocessing unit and a prediction unit;

[0009] The data preprocessing unit is used to perform data cleaning on the power consumption to obtain the power consumption after data cleaning;

[0010] The prediction unit is used to analyze the cleaned power consumption data to obtain the user's predicted power consumption;

[0011] Among them, the power consumption is cleaned and the cleaned power consumption is obtained, including:

[0012] L represents the sequence of power consumption within a set time interval; the power consumption in L is sorted in order from early to late according to the acquisition time, and N represents the total power consumption in L;

[0013] Then for the nth power consumption l in L n, the formula for data cleaning is:

[0014]

[0015] n∈[1,N],wl n For l n The power consumption after data cleaning, wu represents the total number of control sequences, wgt i represents the influence weight of the ith control sequence, c i represents the total amount of electricity consumption in the ith control sequence, wt i,j represents the cleaning weight of the jth power consumption in the ith control sequence, l i,j Represents the jth electricity consumption in the ith comparison sequence.

[0016] Preferably, it also includes a customer management module, which is used to manage user information;

[0017] User information includes electricity purchase records, payment records and account balances.

[0018] Preferably, the power consumption collection module includes an IoT electric meter and a power consumption database;

[0019] The IoT meter is used to upload the user’s electricity consumption to the electricity consumption database in real time;

[0020] The electricity consumption database is used to store the electricity consumption of users.

[0021] Preferably, the cleaned power consumption data is analyzed to predict the power consumption of the user in the next cycle, including:

[0022] The cleaned electricity consumption data is input into the trained neural network model for prediction to obtain the user's predicted electricity consumption.

[0023] Preferably, the trained neural network model includes an LSTM model or a GRU model.

[0024] Preferably, the set time interval is [T s ,T e ], T s is the starting time of the time interval, T e is the time to start cleaning the electricity consumption data, T s and T e The time length between them is the set time length.

[0025] Preferably, the set length of time comprises one year.

[0026] Preferably, the control sequence is determined as follows:

[0027] Calculate the total number of control sequences wu based on L;

[0028] The total amount of electricity consumption in the i-th control sequence is

[0029] The first The Nth power consumption is taken as the power consumption in the ith comparison sequence, and the power consumption in the ith comparison sequence is sorted in order from early to late according to the acquisition time.

[0030] Preferably, wgt i The calculation is done using the following formula:

[0031]

[0032] num i is the total amount of electricity consumption contained in the ith control sequence, num max Indicates the maximum value of the total amount of electricity consumption included in the comparison sequence.

[0033] Preferably, wt i,j The calculation is done using the following formula:

[0034]

[0035] σb represents the maximum value of power consumption in the ith comparison sequence and l n The difference between; ord = nm, m represents l i,j The ranking value corresponding to the power consumption in L.

[0036] In the process of data cleaning of electricity consumption, the electricity sales management system of the present invention does not directly use the average value of electricity consumption data in a fixed range to clean the electricity consumption as in the prior art, but calculates the number of control sequences based on L, sets multiple control sequences, and adaptively calculates different influence weights for the calculation results of different control sequences, and realizes more accurate outlier removal of electricity consumption by integrating the calculation results of multiple control sequences of different lengths. In this way, when the electricity consumption changes greatly, the data change characteristics in the electricity consumption obtained after the outlier removal processing can be more effectively retained, thereby effectively improving the accuracy of the results predicted based on the electricity consumption obtained after the outlier processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0038] Figure 1 A schematic diagram of a power sales management system based on data analysis according to the present invention.

[0039] Figure 2 This is the training process of the neural network model of the present invention. DETAILED DESCRIPTION

[0040] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the 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 the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0041] like Figure 1 In one embodiment shown, the present invention provides a power sales management system based on data analysis, including a power consumption collection module and a prediction module;

[0042] The power consumption collection module is used to obtain the user's power consumption;

[0043] The prediction module includes a data preprocessing unit and a prediction unit;

[0044] The data preprocessing unit is used to perform data cleaning on the power consumption to obtain the power consumption after data cleaning;

[0045] The prediction unit is used to analyze the cleaned power consumption data to obtain the user's predicted power consumption;

[0046] Among them, the power consumption is cleaned and the cleaned power consumption is obtained, including:

[0047] L represents the sequence of power consumption within a set time interval; the power consumption in L is sorted in order from early to late according to the acquisition time, and N represents the total power consumption in L;

[0048] Then for the nth power consumption l in Ln , the formula for data cleaning is:

[0049]

[0050] n∈[1,N],wl n For l n The power consumption after data cleaning, wu represents the total number of control sequences, wgt i represents the influence weight of the ith control sequence, c i represents the total amount of electricity consumption in the ith control sequence, wt i,j represents the cleaning weight of the jth power consumption in the ith control sequence, l i,j Represents the jth electricity consumption in the ith comparison sequence.

[0051] When performing data cleaning, the data cleaning process is not performed based only on a fixed range of power consumption, but is performed based on multiple control sequences of different lengths. This allows the range of data involved in the data cleaning process to no longer be a fixed value. In addition, the data cleaning process is not performed based on a single range. The present invention can integrate the calculation results of different ranges, so that the results of data cleaning are more accurate.

[0052] In addition, since the power consumption at the top of the ranking may not have enough data as a reference, in the process of data cleaning of these power consumptions, the missing data can be processed by filling them with 0. For example, if the number of power consumptions in the reference sequence is 5, and the power consumption currently undergoing data cleaning is ranked third in L, then the first and second power consumptions in L are used as the fourth and fifth power consumptions in the reference sequence, and the first to third power consumptions in the reference sequence are filled with 0.

[0053] Preferably, it also includes a customer management module, which is used to manage user information;

[0054] User information includes electricity purchase records, payment records and account balances.

[0055] The customer management module is mainly used to manage users, that is, electricity sales objects. It can summarize and manage various user-related data to improve management efficiency.

[0056] Preferably, the power consumption collection module includes an IoT electric meter and a power consumption database;

[0057] The IoT meter is used to upload the user’s electricity consumption to the electricity consumption database in real time;

[0058] The electricity consumption database is used to store the electricity consumption of users.

[0059] The IoT electric meter is installed at the user's electricity consumption site, and collects electricity consumption by uploading electricity consumption (electricity consumption in a day) at a regular interval (for example, once a day).

[0060] The predicted power consumption is related to the upload interval. For example, if the upload interval is 1 day, the predicted power consumption is the power consumption for the next day.

[0061] Preferably, the cleaned power consumption data is analyzed to predict the power consumption of the user in the next cycle, including:

[0062] The cleaned electricity consumption data is input into the trained neural network model for prediction to obtain the user's predicted electricity consumption.

[0063] Preferably, the trained neural network model includes an LSTM model or a GRU model.

[0064] LSTM and GRU are good at processing time series data and can capture long-term dependencies in sequences.

[0065] LSTM is an improved recurrent neural network (RNN) structure that is specifically designed to solve the gradient vanishing and gradient exploding problems of traditional RNN when processing long sequence data. LSTM controls the flow of information by introducing three gating mechanisms (input gate, forget gate, and output gate) to better capture long-term dependencies in the sequence.

[0066] Input Gate: Controls the extent to which the current input information enters the cell state.

[0067] Forget Gate: Determines which information in the cell state needs to be forgotten.

[0068] Output Gate: controls the effect of cell state on the current output.

[0069] GRU controls the flow of information by introducing two gating mechanisms (update gate and reset gate), thereby reducing the complexity of the model and improving training efficiency.

[0070] Update Gate: Controls the degree of integration between the hidden state of the previous moment and the current input, and determines which information needs to be updated.

[0071] Reset Gate: Controls the impact of the previous hidden state on the current state and determines which information needs to be forgotten

[0072] like Figure 2 , the training process of the neural network model is roughly as follows:

[0073] Step 1: Data preprocessing:

[0074] Standardize or normalize the electricity usage data used for training and split it into training and test sets.

[0075] Convert the time series data into a form suitable for neural network input (such as taking historical electricity consumption as input and the electricity consumption at the current time point as output).

[0076] Step 2: Build and train the model:

[0077] Build an LSTM or GRU model and set the appropriate number of layers and neurons.

[0078] Use the training set for training and adjust hyperparameters such as learning rate, batch size, etc.

[0079] Training process:

[0080] Input the data into the model for forward propagation and calculate the loss.

[0081] Update the model parameters via back-propagation.

[0082] Use gradient clipping to prevent gradient explosion.

[0083] Step 3: Validate the model:

[0084] Use the test set to verify the performance of the model and check for overfitting or underfitting.

[0085] Preferably, the set time interval is [T s ,T e ], T s is the starting time of the time interval, T e is the time to start cleaning the electricity consumption data, T s and T e The time length between them is the set time length.

[0086] Preferably, the set length of time comprises one year.

[0087] By acquiring historical electricity consumption data over a long period of time, the prediction result can be made more accurate. In the present invention, the set time length can also be other lengths such as half a year, which can be adjusted as needed.

[0088] Preferably, the control sequence is determined as follows:

[0089] Calculate the total number of control sequences wu based on L;

[0090] The total amount of electricity consumption in the i-th control sequence is

[0091] The first The Nth power consumption is taken as the power consumption in the ith comparison sequence, and the power consumption in the ith comparison sequence is sorted in order from early to late according to the acquisition time.

[0092] The number of control sequences of the present invention is not fixed, but can be based on l n The calculated number of control sequences is obtained by calculating the characteristics of the surrounding power consumption. Therefore, the number of calculated control sequences can be made more in line with actual needs, avoiding setting a fixed number of control sequences, and effectively reducing the probability of too many or too few control sequences.

[0093] Preferably, the total number of control sequences wu is calculated based on L, comprising:

[0094] Wu is calculated using the following formula:

[0095]

[0096] MK indicates the number of presets, l k represents the n-MK+kth power consumption in L, ndif represents the number of power consumptions in the set nuc that meet the abnormal value monitoring conditions; nuc is the set of n-MKth to n-1th power consumptions in L; MK is the total number of power consumptions in nuc, prval is the maximum number of control sequences set, l max It represents the maximum power consumption in nuc, and η is the control value.

[0097] The total number of control sequences of the present invention ranges from 1 n The calculation is based on the severity of the change in the surrounding power consumption and the amount of power consumption that meets the abnormal value monitoring conditions. Since these two different types of data are taken into account, it can be obtained in l n The more drastic the change in the surrounding power consumption is, the smaller the value of wu is. In this way, the power consumption obtained after data cleaning can more effectively retain the data change characteristics. In addition, since the number of power consumption under the abnormal value monitoring condition is also introduced, the larger the number is, the larger the value of wu is. In this way, more control sequences of different ranges can be used to compare l n Perform data cleaning to enhance data cleaning efforts and reduce the impact of outliers.

[0098] Furthermore, the preset number may be 20, the control value may range from [0.3, 0.8], and the maximum number of control sequences may be 10.

[0099] Furthermore, the control value is 0.6.

[0100] Furthermore, for the power consumption q in nuc, lq is used to represent the value obtained after data cleaning of the power consumption q. This means that q meets the outlier monitoring conditions.

[0101] Abnormal value monitoring is based on the degree of change in the values ​​before and after data cleaning. The greater the degree of change, the greater the possibility that the electricity consumption before data cleaning is abnormal data.

[0102] Preferably, wgt i The calculation is done using the following formula:

[0103]

[0104] num i is the total amount of electricity consumption contained in the ith control sequence, num max Indicates the maximum value of the total amount of electricity consumption included in the comparison sequence.

[0105] The present invention wgt i It is related to the amount of electricity consumption contained in the control sequence. The larger the amount, the larger the range of the control sequence. Therefore, when performing data cleaning, the impact on the result should be smaller. Because if the range of the control sequence is larger, it contains more data that is too far away from the acquisition time of the current data cleaning process. The more far the acquisition time is different, the smaller the reference value. Therefore, the weight of the present invention can make the control sequence with a smaller average acquisition time difference have a larger corresponding impact weight, which is conducive to obtaining more effective data cleaning results.

[0106] Preferably, wt i,j The calculation is done using the following formula:

[0107]

[0108] σb represents the maximum value of power consumption in the ith comparison sequence and l n The difference between; ord = nm, m represents l i,j The ranking value corresponding to the power consumption in L.

[0109] Based on the control sequence n During data cleaning, if l in the control sequence i,j With l n The greater the data difference and the greater the difference in the sorting values, the smaller the impact on the calculation results of the control sequence. Therefore, data cleaning processing can be performed based on the electricity consumption that is closer in time and value latitudes, which can more effectively retain the data change characteristics and effectively handle outliers.

[0110] The ranking value refers to the rank of an electricity consumption in L. Since the reference sequence is part of L, each electricity consumption in the reference sequence has a corresponding electricity consumption in L. For example, assuming that L contains 20 electricity consumptions, the electricity consumption in the reference reference sequence is the electricity consumption with a ranking value in the interval [5,10] in L. Therefore, the third electricity consumption in the reference sequence corresponds to the seventh-ranked electricity consumption in L, that is, the electricity consumption with a ranking value of 7.

[0111] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A power sales management system based on data analysis, characterized in that: Including power consumption collection module and prediction module; The power consumption collection module is used to obtain the user's power consumption; The prediction module includes a data preprocessing unit and a prediction unit; The data preprocessing unit is used to perform data cleaning on the power consumption to obtain the power consumption after data cleaning; The prediction unit is used to analyze the cleaned power consumption data to obtain the user's predicted power consumption; Among them, the power consumption is cleaned and the cleaned power consumption is obtained, including: L represents the sequence of power consumption within a set time interval; the power consumption in L is sorted in order from early to late according to the acquisition time, and N represents the total power consumption in L; Then for the nth power consumption l in L n , the formula for data cleaning is: n∈[1,N],wl n For l n The power consumption after data cleaning, wu represents the total number of control sequences, wgt i represents the influence weight of the ith control sequence, c i represents the total amount of electricity consumption in the ith control sequence, wt i,j represents the cleaning weight of the jth power consumption in the ith control sequence, l i,j Represents the jth electricity consumption in the ith comparison sequence.

2. The power sales management system based on data analysis according to claim 1 is characterized in that: It also includes a customer management module, which is used to manage user information; User information includes electricity purchase records, payment records and account balances.

3. The power sales management system based on data analysis according to claim 1 is characterized in that: The electricity consumption collection module includes an IoT meter and an electricity consumption database; The IoT electricity meter is used to upload the user’s electricity consumption to the electricity consumption database in real time; The electricity consumption database is used to store the electricity consumption of users.

4. The power sales management system based on data analysis according to claim 1 is characterized in that: Analyze the cleaned electricity consumption data and predict the user's electricity consumption in the next cycle, including: The cleaned electricity consumption data is input into the trained neural network model for prediction to obtain the user's predicted electricity consumption.

5. The power sales management system based on data analysis according to claim 4 is characterized in that: The trained neural network models include LSTM models or GRU models.

6. The power sales management system based on data analysis according to claim 1 is characterized in that: The set time interval is [T s ,T e ], T s is the starting time of the time interval, T e is the time to start cleaning the electricity consumption data, T s and T e The time length between them is the set time length.

7. The power sales management system based on data analysis according to claim 6 is characterized in that: The set time length includes one year.

8. The power sales management system based on data analysis according to claim 1 is characterized in that: The process of determining the control sequence is as follows: Calculate the total number of control sequences wu based on L; The total amount of electricity consumption in the i-th control sequence is The first The Nth power consumption is taken as the power consumption in the ith comparison sequence, and the power consumption in the ith comparison sequence is sorted in order from early to late according to the acquisition time.

9. The power sales management system based on data analysis according to claim 8, characterized in that: wgt i The calculation is done using the following formula: num i is the total amount of electricity consumption contained in the ith control sequence, num max Indicates the maximum value of the total amount of electricity consumption included in the comparison sequence.

10. The power sales management system based on data analysis according to claim 8, characterized in that: wt i,j The calculation is done using the following formula: σb represents the maximum value of power consumption in the ith comparison sequence and l n The difference between; ord = nm, m represents l i,j The ranking value corresponding to the power consumption in L.

Images