Power grid load prediction method and system based on multi-dimensional data

Abstract

The invention relates to a multi-dimensional data-based power grid load prediction method and system. The method comprises the steps of obtaining historical load data of a user side; calculating the correlation between the historical load data and the historical meteorological data; extracting meteorological data of which the correlation is greater than a preset value; taking the meteorological data of which the correlation is greater than a preset value and the corresponding historical load data as a training sample set; inputting the training sample set into a neural network for training to obtain a power grid load prediction model; and predicting the load of a target user side by using the power grid load prediction model to obtain a prediction result. According to the invention, by calculating the correlation between the historical load data and the meteorological data, the meteorological factors highly related to the load change can be screened out, so that the prediction precision of the model is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid load prediction, and in particular to a power grid load prediction method and system based on multi-dimensional data. Background Art

[0002] In the power system, load forecasting is an important part of ensuring the safety, stability and economic operation of the power grid. Accurate load forecasting can help power companies reasonably arrange power generation plans, optimize power grid dispatching, reduce operating costs and reduce power waste. With the rapid development of renewable energy and the increasing complexity of power demand, traditional load forecasting methods have been unable to adapt to the needs of modern power grids. Existing methods mostly rely on historical load data for single time series analysis, ignoring the impact of multi-dimensional factors on load changes. Summary of the invention

[0003] To solve the above problems, an object of the embodiments of the present invention is to provide a method and system for power grid load prediction based on multi-dimensional data.

[0004] A power grid load forecasting method based on multi-dimensional data, comprising:

[0005] Step 1: Obtain historical load data on the user side;

[0006] Step 2: Calculate the correlation between historical load data and historical meteorological data;

[0007] Step 3: Extract meteorological data with correlation greater than a preset value;

[0008] Step 4: The meteorological data with correlation greater than a preset value and its corresponding historical load data are used as a training sample set;

[0009] Step 5: Input the training sample set into the neural network for training to obtain a power grid load prediction model;

[0010] Step 6: Use the power grid load prediction model to predict the load on the target user side to obtain a prediction result.

[0011] Preferably, in step 1, it includes:

[0012] The VMD algorithm is used to preprocess the historical load data on the user side to obtain the preprocessed historical load data on the user side.

[0013] Preferably, in step 3, calculating the correlation between the historical load data and the historical meteorological data includes:

[0014] Using the formula:

[0015]

[0016] Calculate the correlation between historical load data and historical meteorological data.

[0017] Preferably, in step 4, the training sample set includes: historical load data, ambient temperature, rainfall probability and humidity.

[0018] Preferably, in step 4, the perceived temperature is used instead of the ambient temperature to regenerate the training sample set.

[0019] The present invention also provides a power grid load forecasting system based on multi-dimensional data, comprising:

[0020] A historical load data acquisition module is used to acquire historical load data on the user side;

[0021] A correlation calculation module is used to calculate the correlation between historical load data and historical meteorological data;

[0022] A meteorological data screening module, used to extract meteorological data with a correlation greater than a preset value;

[0023] A training sample set building module, used to use meteorological data with a correlation greater than a preset value and its corresponding historical load data as a training sample set;

[0024] A training module, used for inputting the training sample set into a neural network for training to obtain a power grid load prediction model;

[0025] The load prediction module is used to use the power grid load prediction model to predict the load on the target user side to obtain a prediction result.

[0026] The present invention also provides an electronic device, comprising a bus, a transceiver, a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory and the processor are connected via the bus, and wherein the computer program, when executed by the processor, implements the steps in the above-mentioned method for predicting power grid load based on multi-dimensional data.

[0027] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps in the above-mentioned method for predicting power grid load based on multi-dimensional data are implemented.

[0028] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0029] The present invention relates to a power grid load forecasting method based on multidimensional data. Compared with the prior art, the present invention can screen out meteorological factors that are highly correlated with load changes by calculating the correlation between historical load data and meteorological data, thereby improving the forecasting accuracy of the model.

[0030] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1 A flow chart of a power grid load forecasting method based on multi-dimensional data provided by the present invention;

[0033] Figure 2 A schematic diagram of a reference load characteristic curve provided by the present invention;

[0034] Figure 3 A schematic diagram of a periodic load characteristic curve provided by the present invention;

[0035] Figure 4 A schematic diagram of a trend load characteristic curve provided by the present invention;

[0036] Figure 5 A schematic diagram of a peak load characteristic curve provided by the present invention;

[0037] Figure 6 A load forecasting flow chart provided by the present invention;

[0038] Figure 7 A schematic diagram of VMD load decomposition provided by the present invention;

[0039] Figure 8 A schematic diagram of a meteorological data sample set provided by the present invention;

[0040] Fig. 9 A schematic diagram of the MAPE and RMSE indicators provided by the present invention;

[0041] Fig.10 A schematic diagram of the training error provided by the present invention;

[0042] Fig.11 This is a comparison chart of the evaluation indicators provided by the present invention. DETAILED DESCRIPTION

[0043] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0044] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0045] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0046] See also Figure 1 , a power grid load forecasting method based on multi-dimensional data, comprising:

[0047] Step 1: Obtain historical load data on the user side;

[0048] Since the data on the electricity consumption side is relatively large and complex, the present invention adopts two channels for access and comprehensive use. On the one hand, the load data on the electricity consumption side of the system application side is obtained through the front-end page functions of the electricity consumption information collection, SG186 marketing and other systems, and data export tools. On the other hand, the business middle station of the provincial company's Internet department applies for the acquisition of marketing business-related data, including comprehensive data such as power data uploaded by user meters, user meter multipliers, user types, and user-executed electricity prices. In the business middle station, the relevant data total addition calculation formula is added according to actual needs, and the calculation results are migrated to the specified available location through RPA robots or JAVA scripts. Finally, by comparing with the front-end exported data, more accurate and reliable load data on the electricity consumption side is obtained.

[0049] In addition to load data, weather, maintenance plan and other data are also essential external data in load forecasting. Meteorological data uses the grid data of the Meteorological Bureau, including real-time hourly meteorological data within the day and meteorological data for the next 168 hours released twice a day. The maintenance plan-related data also uses existing object-oriented plan data, mainly from the maintenance plan online compilation system of the control cloud platform. Through this system, weekly and monthly maintenance plans can be generated into object-oriented data and provided to the load forecasting function module to realize automatic docking of maintenance plan data.

[0050] The present invention collects relevant data such as historical load, grid connection mode, weather, maintenance plan, etc. required for load forecasting, and actively conducts research on the load characteristics of the new district grid. The present invention determines the starting voltage level for load forecasting from the 10 kV side, and after thoroughly analyzing the 10 kV load characteristics, gradually combines from low to high to complete the load forecast of the grid above 10 kV. According to the above principles, the present invention divides 658 10 kV loads into four categories: benchmark load, periodic load, trend load and peak load, providing a characteristic basis for allocating different forecasting resources for different types of load forecasting in the later stage.

[0051] The benchmark load is mainly industrial load, with the following main features: Figure 2 The load value changes of this type are less affected by external factors such as holidays and weather, and the load extreme value range is relatively fixed. If there are no special matters, the forecasting difficulty of this type of load change trend is relatively low, and it can be regarded as the benchmark load value during the forecast period.

[0052] The periodic loads are mainly schools, heating and other loads with significant periodic changes. The main characteristics are as follows: Figure 3 This type of load value has obvious seasonal or cyclical changes. The load extremes in different long-term periods vary greatly, and the load extremes in a typical period vary little. This type of load change trend needs to be deeply bound to the long-term time parameters.

[0053] Trend loads are mainly general industry and commerce, with the main characteristics such as Figure 4 The trend load value has the characteristics of obvious changes over time and weather. At the same time, the individual load extreme value is relatively fixed, but the overall load extreme value has no obvious regularity. This type of load is the key part of the load forecasting task, and the correlation between its value change and time, events, meteorology and other factors is the key task of load accurate forecasting research.

[0054] Peak loads are mainly residential loads, with the following main features: Figure 5The change in peak load values ​​has obvious regional and time characteristics. The difference between this type of load and trend load is that there are implicit trigger conditions. For example, when the temperature is greater than 32°C, residents or commercial complexes in the area will start related cooling loads on a large scale. This type of load needs to be given priority consideration when predicting regional maximum values ​​during the peak summer and winter periods.

[0055] See also Figure 6 The data preparation stage is the foundation of artificial intelligence technology. It mainly completes the collection, aggregation, and verification of relevant data, and performs some preprocessing according to actual needs to provide data services for subsequent work links.

[0056] See also Figure 7 The present invention first conducts characteristic analysis based on the 10kV line load data, uses the VMD algorithm to split the historical load according to the characteristics, and allocates prediction resources in combination with the benchmark, periodic, trend and peak load characteristics completed in the early stage to improve the efficiency of data processing and prediction calculation.

[0057] Variational Mode Decomposition (VMD) is an adaptive, completely non-recursive method of modal variation and signal processing. This technology has the advantage of being able to determine the number of modal decompositions. Its adaptability is reflected in determining the number of modal decompositions of a given sequence according to actual conditions. In the subsequent search and solution process, it can adaptively match the optimal center frequency and limited bandwidth of each mode, and can achieve effective separation of the intrinsic mode function (IMF), frequency domain division of the signal, and then obtain the effective decomposition components of the given signal, and finally obtain the optimal solution to the variational problem. It solves the problems of modal aliasing and high-frequency signal loss in empirical mode decomposition (EMD). Its calculation formula is:

[0058]

[0059] Where K represents the number of modes to be decomposed (a positive integer), δ(t) represents the Dirac function, and {u k} represents the kth modal component, {w k} represents the frequency center of the kth component.

[0060] After performing variational mode decomposition on the historical load data, the historical load data can be divided into periodic load, peak load, trend load and benchmark load. During training, the preset parameters in the training process need to be adjusted according to the historical load data.

[0061] Step 2: Calculate the correlation between historical load data and historical meteorological data;

[0062] Step 3: Extract meteorological data with correlation greater than a preset value;

[0063] Step 4: The meteorological data with correlation greater than a preset value and its corresponding historical load data are used as a training sample set;

[0064] In practical applications, the present invention combines historical load and meteorological data, and uses algorithms such as the Pearson coefficient to carry out correlation analysis of load change influencing factors. It can be seen from Table 1 below that the correlation coefficient of the historical load value one day before the forecast date is the largest, followed by the first two days, and so on. At the same time, temperature, rainfall, humidity and daily time constitute the top four load influencing factors. Based on this result, the present invention includes the load data, temperature, rainfall probability and humidity historical data seven days before the forecast date into the model training sample set.

[0065] Table 1

[0066]

[0067]

[0068] Step 5: Input the training sample set into the neural network for training to obtain a power grid load prediction model;

[0069] After the training is completed, the present invention puts the formed model into the actual operating environment. When carrying out actual prediction, the load, weather and other data of the past 7 days are first made into a sample set for the model to carry out calculation parameter analysis. After the analysis is completed, the model automatically brings in weather forecast and other information to carry out load forecasting work, and outputs relevant forecast data after completion. After the load forecast is completed, it is still necessary to continuously carry out error analysis to continuously adjust and optimize the performance of the forecast model.

[0070] like Figure 9-10 As shown, the present invention traces back the entire link by evaluation and carries out optimization and improvement work, so as to achieve the purpose of iterative optimization of the model. Commonly used evaluation indicators are MAPE (mean absolute percentage error) and RMSE (root mean square error), where MAPE is used to reflect the average difference between the predicted value and the actual value. The lower the value of MAPE, the better the ability of the model to predict the value. RMSE is used to measure the average size of the error between the model predicted value and the actual observed value. A smaller RMSE value indicates that the difference between the predicted value of the model and the actual value is smaller, that is, the degree of fit of the model is better.

[0071] It should be noted that the present invention can also optimize the prediction model through the following method to improve the accuracy of load prediction.

[0072] 5.1 Finding Similar Days

[0073] Through the analysis of actual meteorological and load data, it was found that the load forecasting results using only meteorological temperature as the main factor had large deviations. Through actual load analysis and relevant data review, the perceived temperature was introduced as the main factor to find an algorithm for optimizing the load model on historical similar days.

[0074] 5.1.1: Calculate body temperature

[0075] Body temperature T e It refers to the temperature people feel after considering the combined effects of factors such as ambient temperature, humidity and wind speed. It can quantitatively express people's feelings about cold and heat, and is also an evaluation index of human comfort. Its calculation formula is:

[0076]

[0077] Where T represents temperature, H R represents humidity and υ represents wind speed.

[0078] 5.1.2: Calculate similar days

[0079] In load forecasting, the meteorological data of 10 days before the current time and 15 days before and after the same day last year are used as the data pool to calculate the perceived temperature of each day, and the two days most similar to the perceived temperature of the next two days are found as similar days for the next two days.

[0080] The similarity value algorithm calculates the absolute percentage error (MAPE) of two days, which is a statistical indicator to measure the degree of error between the predicted value and the actual value. The calculation formula of MAPE is:

[0081] MAPE = (1 / n) × Σ | (actual value - predicted value) / actual value | × 100%

[0082] Here, n represents the number of data points.

[0083] 5.1.3: The analysis results of adding body temperature to find similar days and using temperature alone to find similar days show that the prediction results obtained using body temperature are significantly improved, see Fig.11 shown.

[0084] 5.2 Meteorological data interpolation

[0085] The data obtained from the control cloud is 24 points a day, but the model prediction requires 288 points a day, so the data is interpolated. Interpolation refers to interpolating a continuous function based on discrete data so that this continuous curve passes through all given discrete data points. Using the interpolation method, the approximate value of the function at other points can be estimated based on the value of the function at a finite number of points. The algorithm is as follows:

[0086]

[0087] Among them, X1, Y1 and X2, Y2 are the coordinates of the known data points, X is the value of the independent variable to be estimated, and Y is the corresponding estimated dependent variable value.

[0088] 5.3 Other optimization measures

[0089] On weekdays, the load data of 40 working days from the current time is extracted as the basic data for model training; on holidays, the load data of 40 days from the current time is extracted as the basic data for model training.

[0090] The above historical load data is used as a sample, and the load data for the next two days is predicted by the algorithm based on the forecast indicators for the next two days (weather, whether it is a holiday, etc.). Adjustments are made to various analysis fields according to the accuracy of the model, mainly including the following aspects:

[0091] 5.3.1 Meteorological data correction

[0092] The meteorological data used in the current model calculations are historical meteorological data and forecast meteorological data. Temperature, humidity, wind speed, rainfall probability, sunshine intensity, and air pressure are obtained from the meteorological data of the control cloud as analysis indicators to participate in the model calculation.

[0093] The accuracy of the meteorological data predicted by the control cloud initially used was relatively low, resulting in poor model prediction results. We manually recorded the weather forecast data on mobile phones for many days and replaced it with the predicted weather data indicators used in the model. Observation and analysis found that the model results were relatively accurate. Therefore, after communicating with the meteorological data manufacturer, the predicted meteorological data indicators were replaced with the 1*1 grid data of the National Meteorological Administration, and the accuracy of the model results was significantly improved.

[0094] 5.3.2 Introducing the minimum factor

[0095] Through 3 years of load tracking, it was found that the daily minimum load value was highly correlated with the load forecast, so the minimum value factor was introduced to participate in the model prediction. On the day of model update, when the model prediction was made for the first time before 7 o'clock, the minimum value of the similar daily load obtained before was taken as the minimum value analysis indicator of the model calculation. When the model calculation was run after 7 o'clock, the minimum value of the actual load updated on that day was taken as the minimum value analysis indicator of the model calculation.

[0096] 5.3.3 Introducing Holiday Factor

[0097] After analyzing historical data, it is found that the load value is greatly affected by holidays. The load value fluctuates greatly on the first day of the holiday and the first day of the weekday. Therefore, when performing model calculations, it is divided into a weekday model and a holiday model. When using calculation parameters, the data with the same weekday status are matched and replaced.

[0098] 5.3.4 Forecast model adjustment

[0099] The model prediction time was originally to predict the load of the next day on the same day, but later it was adjusted to predict the data of two days, t+1 and t+2, before 8:30 am on the same day. Later, considering that the load occurring on the same day would be helpful for the prediction, it was adjusted to rolling prediction 4 times a day at 1:40, 7:40, 13:40 and 20:40. Now, because we want to use the latest forecast weather released by the control cloud at 8:00, 12:00 and 20:00 every day, the prediction time is adjusted to rolling prediction 6 times a day at 00:05, 8:05, 12:05, 16:05 and 20:05.

[0100] Step 6: Use the power grid load prediction model to predict the load on the target user side to obtain a prediction result.

[0101] By calculating the correlation between historical load data and meteorological data, the present invention can screen out meteorological factors that are highly correlated with load changes, thereby improving the prediction accuracy of the model.

[0102] The present invention also provides a power grid load forecasting system based on multi-dimensional data, comprising:

[0103] A historical load data acquisition module is used to acquire historical load data on the user side;

[0104] A correlation calculation module is used to calculate the correlation between historical load data and historical meteorological data;

[0105] A meteorological data screening module, used to extract meteorological data with a correlation greater than a preset value;

[0106] A training sample set building module, used to use meteorological data with a correlation greater than a preset value and its corresponding historical load data as a training sample set;

[0107] A training module, used for inputting the training sample set into a neural network for training to obtain a power grid load prediction model;

[0108] The load prediction module is used to use the power grid load prediction model to predict the load on the target user side to obtain a prediction result.

[0109] The present invention also provides an electronic device, comprising a bus, a transceiver, a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory and the processor are connected via the bus, and wherein the computer program, when executed by the processor, implements the steps in the above-mentioned method for predicting power grid load based on multi-dimensional data. Compared with the prior art, the beneficial effects of the electronic device provided by the present invention are the same as the beneficial effects of the method for predicting power grid load based on multi-dimensional data described in the above-mentioned technical solution, and are not elaborated herein.

[0110] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps in the above-mentioned method for predicting power grid load based on multidimensional data are implemented. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present invention are the same as the beneficial effects of the method for predicting power grid load based on multidimensional data described in the above-mentioned technical solution, which will not be repeated here.

[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 technical solution that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for predicting power grid load based on multidimensional data, characterized in that: include: Step 1: Obtain historical load data on the user side; Step 2: Calculate the correlation between historical load data and historical meteorological data; Step 3: Extract meteorological data with correlation greater than a preset value; Step 4: The meteorological data with correlation greater than a preset value and its corresponding historical load data are used as a training sample set; Step 5: Input the training sample set into the neural network for training to obtain a power grid load prediction model; Step 6: Use the power grid load prediction model to predict the load on the target user side to obtain a prediction result.

2. A method for predicting power grid load based on multidimensional data according to claim 1, characterized in that: In step 1, it includes: The VMD algorithm is used to preprocess the historical load data on the user side to obtain the preprocessed historical load data on the user side.

3. A method for predicting power grid load based on multidimensional data according to claim 2, characterized in that: In step 3, calculating the correlation between the historical load data and the historical meteorological data includes: Using the formula: Calculate the correlation between historical load data and historical meteorological data; where x i represents historical load data, y i represents historical meteorological data, r represents correlation, and N represents the number of sampling points.

4. A method for predicting power grid load based on multidimensional data according to claim 3, characterized in that: In step 4, the training sample set includes: historical load data, ambient temperature, rainfall probability and humidity.

5. The method for predicting power grid load based on multidimensional data according to claim 3, characterized in that: In step 4, the perceived temperature is used instead of the ambient temperature to regenerate the training sample set.

6. A power grid load forecasting system based on multidimensional data, characterized in that: include: A historical load data acquisition module is used to acquire historical load data on the user side; A correlation calculation module is used to calculate the correlation between historical load data and historical meteorological data; A meteorological data screening module, used to extract meteorological data with a correlation greater than a preset value; A training sample set building module, used to use meteorological data with a correlation greater than a preset value and its corresponding historical load data as a training sample set; A training module, used for inputting the training sample set into a neural network for training to obtain a power grid load prediction model; The load prediction module is used to use the power grid load prediction model to predict the load on the target user side to obtain a prediction result.

7. An electronic device, comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, wherein: When the computer program is executed by the processor, the steps in the method for predicting power grid load based on multi-dimensional data as described in any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps in a method for predicting power grid load based on multi-dimensional data as described in any one of claims 1 to 5 are implemented.

Images