Enterprise energy economic management optimization method and system based on artificial intelligence

Abstract

The invention relates to the field of data prediction, in particular to an enterprise energy economic management optimization method and system based on artificial intelligence. Comprising the steps of obtaining historical electric energy data, and determining reference data and target data; calculating the time difference between the target data and the reference data, determining the duration of an effective analysis section, and obtaining an initial effective analysis section; the distribution concentration degree of each initial analysis effective section is calculated, and analysis effective sections are determined; lagging item representative points in the effective section are obtained and analyzed, and representative point activeness corresponding to each lagging item representative point is calculated; lagging difference weights corresponding to the lagging item representative points are determined, and optimized lagging item representative points are obtained; determining the number of additional points of representative points of each lag term; and outputting predicted electric energy data through an ARIMA algorithm. According to the invention, the prediction accuracy of the electric energy data can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data prediction, in particular to an enterprise energy economic management optimization method and system based on artificial intelligence. BACKGROUND

[0002] The "energy economy" of an enterprise generally refers to the comprehensive economic activities related to costs, benefits, resource utilization, and energy management in the process of energy consumption and use of the enterprise. The core purpose is to optimize energy use, improve energy utilization efficiency, reduce energy costs, and reduce energy waste while meeting production needs. The main content of enterprise energy economy includes: energy cost, which is the cost of energy procurement, consumption and management of the enterprise. It usually includes the procurement cost and related operating cost of energy such as electricity, natural gas, water, steam, etc.; energy use efficiency, which is the energy consumption per unit of production or unit output, reflecting the efficiency of energy use. Energy utilization rate can be improved through measures such as improving production process, equipment maintenance, and energy-saving technology. Corresponding energy economic management optimization can be achieved by analyzing and predicting energy data based on artificial intelligence, ensuring reasonable energy allocation, reducing economic benefits, and rational use of energy.

[0003] In the prior art, in order to realize the management optimization of enterprise energy economy in the aspect of electric energy, the enterprise electricity data is collected and analyzed, and the power consumption is analyzed through a prediction model (specifically ARIMA). Because the electricity price is generally high during the peak period of electricity consumption, the economic benefit is low, so the electricity consumption during the peak period of electricity consumption is generally adjusted according to the energy consumption result, and is allocated to other periods with low power consumption, so as to improve the economic benefit of electric energy management.

[0004] There is a major problem in the process of predicting and analyzing enterprise electric energy data through the ARIMA model: the enterprise electricity consumption is complex. For example, the electricity consumption of an enterprise can be divided into: daily variation, weekly variation, and weekly variation, etc. Among them, the daily variation can be manifested as high power consumption during the working period and low demand during the non-working period; the weekly variation can be manifested as dynamic electricity consumption change during weekdays, but non-dynamic electricity consumption change during weekends, which significantly reduces the stability of the electricity consumption data; further, seasonal changes, such as air conditioner use in summer, also greatly increase electricity consumption, while in winter it is relatively low. Therefore, the data showing unstable changes is predicted through the ARIMA model, which may result in inaccurate prediction results due to the lack of stability of the predicted data. SUMMARY

[0005] The present application provides an enterprise energy economic management optimization method and system based on artificial intelligence to solve the existing problems.

[0006] An enterprise energy economic management optimization method based on artificial intelligence provided by the present application adopts the following technical scheme:

[0007] An embodiment of the present application provides an enterprise energy economic management optimization method based on artificial intelligence, which comprises the following steps:

[0008] Obtaining historical electric energy data, and determining reference data and target data from the historical electric energy data respectively;

[0009] Calculating the time difference of the target data and the reference data, determining the analysis effective segment length according to the time difference, and dividing the historical electric energy data according to the analysis effective segment length to obtain an initial analysis effective segment;

[0010] Calculating the distribution concentration of each initial analysis effective segment, and determining the initial analysis effective segment with a distribution concentration greater than a preset concentration threshold as an analysis effective segment;

[0011] Obtaining a lag term representative point in the analysis effective segment, and calculating the representative point activity corresponding to each lag term representative point;

[0012] According to the representative point activity, determining the lag difference weight corresponding to the lag term representative point to obtain an optimized lag term representative point;

[0013] According to the corresponding relationship between the preset lag difference weight and the additional point quantity, determining the additional point quantity of each lag term representative point;

[0014] Taking the optimized lag term representative point as the input of the sample data of the ARIMA algorithm, taking the additional point quantity of each lag term representative point as the insertion sample point quantity of the lag difference of the ARIMA algorithm, and running the ARIMA algorithm, and outputting the predicted electric energy data through the ARIMA algorithm.

[0015] Optionally, the reference data and the target data are determined from the historical electric energy data, specifically comprising:

[0016] The data in the nth period in the historical electric energy data is determined as the reference data, wherein the historical electric energy data is obtained by dividing the electric energy data according to a preset time period, and the time axes of the historical electric energy data in different preset time periods are the same;

[0017] The data in the nth-j period in the historical electric energy data is determined as the target data, wherein n and j are positive integers, n is greater than j, and the data in the nth-j period is not the data in the first period.

[0018] Optionally, the time difference of the target data and the reference data is calculated, specifically comprising:

[0019] Acquire peak power consumption time of target data and reference data, and obtain target time sequence and reference time sequence, wherein, the target data is data of the n-jth period in historical power data, and the reference data is data of the nth period in historical power data;

[0020] Calculate absolute value of difference between the mth element in the target time sequence and the mth element in the reference time sequence, and determine the mth time difference of target data and reference data;

[0021] Acquire the mth time difference of each time of target data and reference data and sum them up, and obtain the jth period time difference of target data and reference data;

[0022] Acquire the jth period time difference of target data and reference data respectively and sum them up, and obtain the time difference of target data and reference data, wherein, the data of the n-jth period to the n-1th period in historical power data is taken as target data respectively;

[0023] Optionally, the analysis effective segment length is determined according to the time difference, and specifically includes:

[0024] The peak arrangement of target data and reference data is calculated according to the time difference;

[0025] The period length between the period number of target data and the period number of reference data with the highest peak arrangement is determined as the analysis effective segment length.

[0026] Optionally, the peak arrangement of target data and reference data is calculated according to the time difference, and specifically includes:

[0027] The data of the n-jth period to the n-1th period in historical power data is taken as target data respectively, the mean value of the data corresponding to the peak power consumption time in the target data is determined as the target power peak mean value, and the mean value of the data corresponding to the peak power consumption time in the reference data is obtained, and the reference power peak mean value is obtained;

[0028] The peak arrangement is calculated according to the time difference, the target power peak mean value and the reference power peak mean value.

[0029] Optionally, the distribution concentration of each initial analysis effective segment is calculated, and specifically includes:

[0030] Short-time Fourier transform is performed on each initial analysis effective segment to obtain frequency domain results;

[0031] The distribution concentration is calculated according to the maximum frequency component energy value and the total energy value in the frequency domain results.

[0032] Optionally, the lag representative point in the analysis effective segment is acquired, and specifically includes:

[0033] The effective segment is fitted by using a least square method to obtain an effective segment curve;

[0034] The maximum point and the minimum point in the effective segment curve are determined as the hysteresis representative points.

[0035] Optionally, the representative point activity corresponding to each hysteresis representative point is calculated, and the calculation specifically includes:

[0036] The hysteresis representative points are fitted by using a least square method to obtain a continuous curve;

[0037] The slope interval of the continuous curve is obtained, and the slope interval is divided according to a preset slope length to obtain a slope interval;

[0038] The activity of each slope interval is calculated, and the activity of the slope interval corresponding to the slope of the hysteresis representative point is determined as the representative point activity corresponding to the hysteresis representative point.

[0039] Optionally, the activity of each slope interval is calculated, and the calculation specifically includes:

[0040] The average of the number of hysteresis representative points of each analysis effective segment in the pth slope interval is determined as the average number of representative points of the pth slope interval;

[0041] The median of the slope of the pth slope interval is obtained;

[0042] The absolute value of the difference between the number of hysteresis representative points of the uth analysis effective segment in the pth slope interval and the number of hysteresis representative points of the (u-1)th analysis effective segment in the pth slope interval is calculated to obtain the representative point number difference of the uth analysis effective segment, wherein the uth analysis effective segment is not the first analysis effective segment;

[0043] The representative point number difference is inverted to obtain the number difference reciprocal of the uth analysis effective segment;

[0044] The number difference reciprocals of each analysis effective segment are obtained and summed to obtain the number difference reciprocal of the pth slope interval;

[0045] The activity of the pth slope interval is calculated according to the average number of representative points of the pth slope interval, the median of the slope of the pth slope interval, and the number difference reciprocal of the pth slope interval;

[0046] The activity of each slope interval is obtained.

[0047] The application provides an enterprise energy economic management optimization system based on artificial intelligence, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor.

[0048] The technical scheme of the application has the beneficial effects that:

[0049] The application predicts and analyzes enterprise energy (electricity) data to obtain a prediction result. In the predicted energy data, the time period with relatively concentrated electricity consumption is analyzed and controlled for electricity use, and is reasonably dispersed to different time periods. The problem of low economic benefits caused by high electricity prices in the time period with concentrated electricity consumption is solved. That is, the electricity consumption peak load shifting method is used to maximize economic benefits.

[0050] In the prediction process of enterprise electricity data by the prediction model, the instability of the enterprise electricity data under different time conditions may cause deviation of the prediction result. The periodic trend analysis of the enterprise electricity data is performed, the trend analysis result is taken as a lag difference item, and the lag difference method is combined to eliminate the instability and improve the prediction accuracy.

[0051] In the general electricity data, the periodic term extracted from the lag difference part is removed, and the periodic term is taken as a predicted electricity sequence. In the prediction process, there are special electricity consumption conditions of enterprises, for example, the production impulse period, and the corresponding electricity data consumption is high. At this time, if the lag difference interval is not changed, a certain overfitting phenomenon will be caused, and the prediction accuracy will be reduced. The lag difference is dynamically adjusted to realize the rolling prediction mode, eliminate the overfitting problem and improve the prediction accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical schemes in the embodiments of the application or the prior art, the drawings needed to be used in the embodiment or prior art description will be briefly introduced. Obviously, the drawings in the following description only some embodiments of the application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0053] Figure 1 A flowchart of an enterprise energy economic management optimization method based on artificial intelligence provided by an embodiment of the application;

[0054] Figure 2 A structural diagram of an enterprise energy economic management optimization system based on artificial intelligence provided by an embodiment of the application. DETAILED DESCRIPTION

[0055] In order to further clarify the technical means and effects taken by the present application to achieve the predetermined inventive purpose, the specific implementation, structure, characteristics and effects of a kind of enterprise energy economic management optimization method based on artificial intelligence according to the present application are described in detail as follows in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0057] The specific scheme of the enterprise energy economic management optimization method based on artificial intelligence provided by the present application is described in detail below in combination with the drawings.

[0058] The present application provides an enterprise energy economic management optimization method and system based on artificial intelligence, please refer to Figure 1 , which shows a flowchart of an enterprise energy economic management optimization method based on artificial intelligence provided by one embodiment of the present application, which includes the following steps:

[0059] S101, obtain historical power data, and determine reference data and target data from the historical power data respectively.

[0060] In this embodiment, the reference data and the target data are determined from the historical power data, specifically including:

[0061] The data in the nth period in the historical power data is determined as the reference data, wherein the historical power data is obtained by dividing the power data according to the preset time period, and the time axis of the historical power data in different preset time periods is the same; the data in the n-j period in the historical power data is determined as the target data, wherein n and j are positive integers, n is greater than j, and the data in the n-j period is not the data in the first period.

[0062] Exemplarily, the collection method of power data in this embodiment can be as follows: multi-period collection: according to the working time, power peak and valley characteristics of the enterprise, data collection is carried out in different time periods (such as weekdays, weekends, seasonal fluctuations, etc.), to ensure covering peak period and valley period. Multi-seasonal collection: collect data in different seasons (such as spring, summer, autumn, winter) to capture the impact of seasonal changes on energy consumption. The preset time period of historical power data can be natural day, natural month and natural year, etc. without specific limitation here.

[0063] The time axis of the historical electric energy data in different preset time periods is the same, indicating that the corresponding time points of the historical electric energy data in different preset time periods are the same. For example, the data of the first day is collected at 8 o'clock, 9 o'clock and 10 o'clock respectively, so the data collection time of each subsequent day is also 8 o'clock, 9 o'clock and 10 o'clock, and if the preset time period is a natural month or a natural year, the time axis is also the same.

[0064] For example, the collected electric energy data is as follows: [2025-01-01 08:00 energy (electric energy) consumption is 104 kWh], [2025-01-01 08:30 energy (electric energy) consumption is 150104 kWh], [……], [2025-01-04 09:00 energy (electric energy) consumption is 103 kWh], [……].

[0065] Optionally, after the electric energy data is collected, a preprocessing operation can be performed on the electric energy data, and the preprocessing operation can include:

[0066] The missing data is filled by using an interpolation method (linear interpolation) to fill the values before and after.

[0067] The statistical method (Z-score) is used to identify and remove outliers, for example, if the electric energy consumption data in a certain period is abnormal (for example, the value suddenly becomes 1000 kWh within one hour), it can be detected and removed by Z-score (exceeding a threshold such as 3). The moving average method is used to smooth the data and reduce the noise caused by short-term fluctuations. For example, the input data is: [50, 55, 53, 60, 57, 45, 52], and the output smoothed data is: [54, 55, 56, 55, 56, 54, 55]. The seasonal decomposition (such as STL decomposition) is used to decompose the data into trend, seasonality and residual parts, and the seasonal fluctuations and long-term trends are removed to reduce the influence of noise. The difference method (such as first-order difference) is used to remove the long-term trend in the data. The historical electric energy data is obtained after the preprocessing operation.

[0068] In a specific embodiment, the historical electric energy data can be the total electricity consumption in a time period (every half hour), and the unit is kWh. The electric energy sequence corresponding to the historical electric energy data is: wherein represents the electric energy sequence, represents the total electric energy in the time period .

[0069] S102, calculate the time difference between the target data and the reference data, determine the analysis effective segment length according to the time difference, and divide the historical electric energy data according to the analysis effective segment length to obtain an initial analysis effective segment.

[0070] In the embodiment, the time difference of the target data and the reference data is calculated, specifically including:

[0071] The peak power consumption time of the target data and the reference data is obtained, and the target time sequence and the reference time sequence are obtained, wherein the target data is the data of the n-jth period in the historical power data, and the reference data is the data of the nth period in the historical power data;

[0072] The absolute value of the difference between the mth element in the target time sequence and the mth element in the reference time sequence is calculated, and the mth time difference of the target data and the reference data is determined;

[0073] The mth time difference of each time of the target data and the reference data is obtained and summed, and the jth period time difference of the target data and the reference data is obtained;

[0074] The data of the n-jth period to the n-1th period in the historical power data is taken as the target data respectively, the jth period time difference of the target data and the reference data is obtained respectively, and the time difference of the target data and the reference data is summed.

[0075] The analysis effective segment length is determined according to the time difference, specifically including:

[0076] The peak arrangement of the target data and the reference data is calculated according to the time difference;

[0077] The period length between the period number of the target data with the highest peak arrangement and the period number of the reference data is determined as the analysis effective segment length.

[0078] The peak arrangement of the target data and the reference data is calculated according to the time difference, specifically including:

[0079] The data of the n-jth period to the n-1th period in the historical power data is taken as the target data, the mean value of the data corresponding to the peak power consumption time in the target data is determined as the target power peak mean value, and the mean value of the data corresponding to the peak power consumption time in the reference data is obtained, and the reference power peak mean value is obtained;

[0080] The peak arrangement is calculated according to the time difference, the target power peak mean value and the reference power peak mean value.

[0081] For example, taking a preset time period of calendar days as an example. Typical enterprise electricity consumption characteristics show significant intraday fluctuations, primarily influenced by production schedules. Electricity consumption is higher during working hours (usually from morning to evening), especially during peak production periods (such as 9:00 AM to 12:00 PM and 2:00 PM to 5:00 PM); the main electricity consumption of enterprises is often concentrated during these periods of intensive production activity. With the increase in production and office activities, electricity demand increases rapidly, especially for equipment requiring large amounts of electricity (such as machinery, cooling systems, lighting, etc.).

[0082] The weekday fluctuations include the fact that most businesses consume a large amount of electricity from Monday to Friday (weekdays), while electricity demand decreases significantly on Saturdays, Sundays (rest days), or public holidays. For example, production is typically most intensive during weekdays, especially Mondays and Wednesdays.

[0083] Therefore, the peak electricity consumption time in the power energy sequence is an important data characteristic. Thus, obtaining the... Peak electricity consumption times for a day. It should be noted that the "peak" in "peak electricity consumption times" refers to the extreme value (not the maximum or minimum) of electricity consumption within a day, i.e., the peak value. This value can exist in multiple instances, as each peak value represents an extreme change in electricity consumption data.

[0084] For example, peak electricity consumption time Indicates the first Day The peak electricity consumption time corresponding to each peak (which can also be denoted as) The differences in peak electricity consumption times across different days. The smaller (where, Indicates the first The first day The smaller the characteristic difference between peak electricity consumption times, the more regular the usage time. For example, the electrical equipment in a manufacturing enterprise is turned on around 8:00 am on different days.

[0085] The smaller the difference, the higher the distribution characteristics of the peaks over several days, and the higher the likelihood that it is an effective segment for power cycle analysis.

[0086] To calculate the time difference between target data and baseline data, we can first obtain the peak electricity consumption times of the target data and baseline data, and then obtain the target time sequence and baseline time sequence respectively; then calculate the time difference between the target data and baseline data based on the target time sequence and baseline time sequence.

[0087] The time difference between the target data and the reference data is calculated according to the target time sequence and the reference time sequence. The calculation formula used can be:

[0088]

[0089] wherein, represents the time difference, represents the number of periods different between the period of the target data and the period of the reference data, represents the number of elements in the target time sequence or the reference time sequence (the number of elements in the target time sequence and the reference time sequence is the same), represents the peak electricity time corresponding to the peak value in the i th period, represents the peak electricity time corresponding to the peak value in the i th period.

[0090] The higher the time difference between the peak electricity times, the greater the corresponding time difference. When calculating the difference between the peak electricity times, when the number of peak electricity times is not the same, for example, the target time sequence has 8 elements, and the reference time sequence only has 5 elements, then the reference time sequence is copied to 8 to complete the subtraction of the peak electricity times.

[0091] The peak arrangement is calculated according to the time difference, the target power peak mean and the reference power peak mean. The calculation formula used can be:

[0092]

[0093] wherein, represents the peak arrangement, represents the normalization method, represents the target power peak mean in the i th period in the historical power data, represents the reference power peak mean. The peak arrangement is mainly the difference between the peak means. The higher the difference between the peak means, the higher the variability of the enterprise electricity consumption corresponding to the electricity peak of the same serial number in different days, and the lower the peak arrangement, because it does not meet the segmentation effectiveness for the purpose of periodicity.

[0094] The reciprocal of the time difference is used as the coefficient of the peak arrangement to participate in the analysis of the arrangement. The higher the time difference, the lower the electricity regularity, and the lower the corresponding arrangement.

[0095]

[0096] ​​​​​The peak arrangement of the historical electric energy data from the first period to the m-th period is normalized to obtain a peak arrangement sequence of the historical electric energy data from the first period to the m-th period. The peak arrangement of the historical electric energy data from the first period to the m-th period is normalized to obtain a peak arrangement sequence of the historical electric energy data from the first period to the m-th period. .

[0097] The period length between the period number of the target data with the highest peak arrangement and the period number of the reference data in the sequence is determined as the analysis effective segment length.

[0098] S103, calculate the distribution concentration of each initial analysis effective segment, and determine the initial analysis effective segment with the distribution concentration greater than the preset concentration threshold as an analysis effective segment.

[0099] In the embodiment, the distribution concentration of each initial analysis effective segment is calculated, specifically including:

[0100] Perform short-time Fourier transform on each initial analysis effective segment to obtain a frequency domain result.

[0101] Calculate the distribution concentration according to the maximum frequency component energy and the total energy in the frequency domain result.

[0102] Exemplarily, the data of each initial analysis effective segment is discrete, and short-time Fourier transform is selected for frequency domain acquisition. Meanwhile, short time refers to an electric energy period that can be analyzed, i.e., an effective segment, which needs to include corresponding periodic changes, i.e., to ensure that there is a periodic change that can be analyzed in a period.

[0103] Obtain an initial analysis effective segment, which covers a relatively obvious periodic trend, and obtain its frequency domain result through short-time Fourier transform. Since the original electric energy sequence is discrete, the corresponding short-time Fourier transform result is continuous.

[0104] Calculate the distribution concentration according to the maximum frequency component energy and the total energy in the frequency domain result. The calculation formula used can be:

[0105]

[0106] wherein, represents the maximum frequency component energy, represents the total energy, represents the distribution concentration.

[0107] Optionally, the preset concentration threshold can be 0.5, or other values can be set according to historical experience, and the value is set according to actual needs, which is not limited in specific numerical value.

[0108] ​​When the distribution concentration is greater than the preset concentration threshold, it is proved that there is a high proportion of periodic components in the frequency domain result corresponding to the period, and the corresponding original data has obvious periodicity.

[0109] In S104, a lag representative point in the analysis effective segment is obtained, and a representative point activity corresponding to each lag representative point is calculated.

[0110] In this embodiment, the lag representative point in the analysis effective segment is obtained, specifically including:

[0111] The analysis effective segment is curve-fitted using the least square method to obtain an effective segment curve.

[0112] The maximum point and the minimum point in the effective segment curve are determined as the lag representative point.

[0113] The representative point activity corresponding to each lag representative point is calculated, specifically including:

[0114] The lag representative point is fitted using the least square method to obtain a continuous curve.

[0115] The slope interval of the continuous curve is obtained, and the slope interval is divided according to a preset slope length to obtain a slope interval.

[0116] The activity of each slope interval is calculated, and the activity of the slope interval corresponding to the slope of the lag representative point is determined as the representative point activity corresponding to the lag representative point.

[0117] The activity of each slope interval is calculated, specifically including:

[0118] The mean value of the number of lag representative points of each analysis effective segment in the pth slope interval is determined as the representative point number mean value of the pth slope interval.

[0119] The slope median value of the pth slope interval is obtained.

[0120] The absolute value of the difference between the number of lag representative points of the uth analysis effective segment in the pth slope interval and the number of lag representative points of the u-1th analysis effective segment in the pth slope interval is calculated to obtain the representative point number difference of the uth analysis effective segment, wherein the uth analysis effective segment is not the first analysis effective segment.

[0121] The representative point number difference is inverted to obtain the number difference reciprocal of the uth analysis effective segment.

[0122] The number difference reciprocal of each analysis effective segment is obtained and summed to obtain the number difference reciprocal of the pth slope interval.

[0123] The activity of the pth slope interval is calculated according to the mean of the number of representative points of the pth slope interval, the median of the slope of the pth slope interval, and the reciprocal of the number difference of the pth slope interval.

[0124] The activity of each slope interval is obtained.

[0125] Exemplarily, in the ARIMA algorithm, the lag term refers to optimizing the original adjacent difference relationship into a lag difference relationship in the difference stage of the original data, and the difference result between the two can be used as an effective periodicity. In this embodiment, during enterprise electricity consumption, the electricity consumption in each period is relatively high and concentrated in the morning because of the high activity of productivity and production equipment. Therefore, the relatively concentrated peak value at this time can be used as one of the characteristic points (representative points) with electricity consumption changes. When productivity and production equipment are relatively saturated in the afternoon, the electricity consumption will also reach a low value. Therefore, the low point can be used as another characteristic point (lag term representative point) of electricity consumption change. At this time, the electricity value with two characteristics as the lag term difference factor is used for calculation, to improve the periodicity and reduce the prediction accuracy problem caused by unstable changes.

[0126] Therefore, in order to accurately obtain such representative points, it is necessary to analyze the frequency domain characteristics corresponding to the electricity sequence, and determine the lag term representative points according to the frequency domain characteristics and the time domain change rule. When the distribution concentration degree is greater than the preset concentration threshold, it is considered that there is a lag term representative point in the effective segment; otherwise, there is no lag term representative point, and the segment data does not have predictability. Moreover, the lag term representative point can be all maximum points and minimum points in the effective segment curve of the analysis effective segment that meets the frequency domain condition.

[0127] After the lag term representative points are determined, it is indicated that such electricity data can be used as the original data for lag difference in the ARIMA algorithm. The lag difference needs to be calculated by differentiating such data, to ensure the periodicity, remove unstable noise data, and obtain accurate prediction results.

[0128] Further, the extreme values of the electricity data in several days are used as lag term representative points. Different representative points have different meanings and importance in the prediction process. For example, the extreme value points of the daily change have a fast update speed and can reflect the real-time and latest electricity data. The extreme value points of the weekly change have a long time period, and the time line is longer than the latest electricity value. Therefore, the importance of the lag difference is low, and the same is true for the seasonal extreme value points.

[0129] Furthermore, enterprises or industries typically experience significant peak periods in their annual plans and production cycles (e.g., quarterly production peaks, year-end settlement periods). During these settlement periods, there may be substantial fluctuations in electricity usage among different representative points, with the aim of improving real-time productivity and achieving the company's annual performance targets. Therefore, to ensure the real-time nature of the forecast data, reduce the probability of false predictions, and improve the final forecast accuracy, activity analysis is performed on each representative point for each lagged term to obtain the representative point's activity level.

[0130] The least squares method is used to fit the representative points of the lag terms to obtain a continuous curve, which represents the relationship between the lag term representative points. The maximum slope value of the curve is then obtained. and its minimum slope value It is divided according to a preset slope length to obtain several slope interval sequences ( (Slope intervals) Within each slope interval, there exist a number of representative points of the lagged terms. The more representative points there are, the more active the lag term representative points are within the slope range, meaning that the corresponding electricity usage data of the enterprise under this condition has a higher variation characteristic.

[0131] Optionally, the preset slope length can be set according to the actual situation, and no specific restrictions are imposed here. A preferred embodiment is that the number of slope intervals after division is 20.

[0132] The activity level for each slope interval can be calculated using the following formula:

[0133]

[0134] in, Indicates the slope interval The activity of the point represented by the inner lag term. Indicates the number of valid segments analyzed. express In the valid segments of the analysis, the slope interval The mean number of points representing the internal lag term. Indicates the first In the slope interval of the valid analysis segment The number of points represented by the lag term within the interval. Indicates the first In the slope interval of the valid analysis segment The number of points represented by the lag term within the interval. Indicates the slope interval The median slope.

[0135] In the activity calculation formula, the product weight for the former is... The higher the value, the more the number of lag term representative points in the slope interval, and the higher the activity of the lag term representative points. The smaller the slope value, i.e., the closer to 0, the higher the activity of the corresponding lag term representative points, because the closer the slope value is to 0, the closer the corresponding enterprise electricity consumption is to the maximum or minimum value.

[0136] S105, according to the activity of the representative point, determining the lag difference weight corresponding to the lag term representative point, and obtaining the optimized lag term representative point.

[0137] In this embodiment, according to the activity of the representative point, the lag difference weight corresponding to the lag term representative point can be determined after the activity of the representative point is normalized.

[0138] In the ARIMA algorithm, the lag term representative point with higher activity can be given a larger lag difference weight. The lag difference weight means that the difference result with higher activity is given higher attention, and the number of lag difference between them is increased.

[0139] For example, for the lag term representative points a and b (adjacent and existing in the same analysis effective segment), after calculating, the activity is higher, then the electric energy data points between them are increased by one difference data c, and the lag difference result a-b is increased to a-c and a-b. The purpose is to amplify the characteristics of the active lag point corresponding to the electric energy data interval, improve the proportion of real-time and latest data contained in the prediction result, and improve the prediction accuracy.

[0140] S106, according to the corresponding relationship between the preset lag difference weight and the number of additional points, determining the number of additional points of each lag term representative point.

[0141] In a specific embodiment, the preset lag difference weight can be obtained by sorting the lag difference weight, and the number of additional points between the lag term representative points with the largest part (10%) is increased by 10, and the number of additional points can be reduced by 1 for every 10% reduction, until the number of additional points of the lag term representative point with smaller weight is 0. Alternatively, the corresponding number of additional points can be set for each lag difference weight according to the actual situation, which is not limited here.

[0142] S107, taking the optimized lag term representative point as the input of the sample data of the ARIMA algorithm, taking the number of additional points of each lag term representative point as the number of inserted sample points of the lag difference of the ARIMA algorithm, and running the ARIMA algorithm, and outputting the predicted electric energy data through the ARIMA algorithm.

[0143] ​After S107, the embodiment can further include obtaining a new analysis effective segment (newly generated electric energy data), taking the new analysis effective segment as historical data, recalculating the lag difference weight of the lag representative point, obtaining a new lag representative point and a new lag difference weight of the lag representative point, using the new lag representative point and the new lag difference weight of the representative point for prediction, obtaining predicted electric energy data, and achieving the purpose of rolling prediction.

[0144] Illustratively, according to the predicted electric energy data, first, the peak and valley periods of electricity consumption need to be identified, the production and electricity consumption time is reasonably adjusted, and the load demand is balanced. Through optimization of power dispatching, high energy consumption equipment in low price period is preferentially arranged, and excessive consumption in peak period is avoided. At the same time, by using load management strategies such as peak clipping and valley filling and demand response, the energy expenditure of enterprises in the power peak period is reduced, thereby effectively reducing the overall energy cost.

[0145] In summary, in the embodiment of the present application, the prediction result is obtained by predicting and analyzing the enterprise energy (electric energy) data. In the predicted energy data, by judging and analyzing, the time period with relatively concentrated electricity consumption is regulated and controlled for electric energy use, and is reasonably dispersed to different time. The problem of low economic benefit caused by high electricity price in the period with concentrated electricity consumption is solved. That is, by using the peak clipping and valley filling method of electric energy consumption, the economic benefit maximization is realized.

[0146] The present application also proposes an enterprise energy economic management optimization system based on artificial intelligence, please refer to Figure 2 which shows the structure diagram of an enterprise energy economic management optimization system based on artificial intelligence provided by an embodiment of the present application, the system comprises a data acquisition module 101, a data processing module 102 and a prediction analysis module 103.

[0147] The data acquisition module 101 is used for acquiring historical electric energy data, and determining reference data and target data from the historical electric energy data respectively;

[0148] The data processing module 102 is used for calculating the time difference of the target data and the reference data, determining the analysis effective segment length according to the time difference, and dividing the historical electric energy data according to the analysis effective segment length to obtain an initial analysis effective segment; calculating the distribution concentration degree of each initial analysis effective segment, and determining the initial analysis effective segment with the distribution concentration degree greater than a preset concentration threshold as an analysis effective segment; obtaining the lag representative point in the analysis effective segment, and calculating the representative point activity corresponding to each lag representative point; determining the lag difference weight corresponding to the lag representative point according to the representative point activity, obtaining the optimized lag representative point; and determining the number of extra points of each lag representative point according to the corresponding relationship between the preset lag difference weight and the number of extra points;

[0149] The prediction analysis module 103 is configured to input the optimized lag representative points into the ARIMA algorithm as sample data, input the additional point number of each lag representative point into the ARIMA algorithm as the inserted sample point number of the lag difference, and run the ARIMA algorithm, and output predicted power data through the ARIMA algorithm.

[0150] It should be noted that the system provided in the above embodiments is only exemplified by the division of the above functional modules. In actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the above-described functions. In addition, the enterprise energy economic management optimization system and the enterprise energy economic management optimization method based on artificial intelligence provided in the above embodiments belong to the same concept, and the specific implementation process is described in the method embodiment, which will not be described here.

[0151] It should be noted that the above sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0152] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments.

[0153] The above only describes the preferred embodiments of the present application and does not limit the present application. Any modification, equivalent replacement, improvement, etc. within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. An artificial intelligence-based method for optimizing enterprise energy economic management, characterized in that, include: Acquire historical power data, and determine baseline and target data from the historical power data respectively; Calculate the time difference between the target data and the baseline data, determine the effective analysis period based on the time difference, and divide the historical power data according to the effective analysis period to obtain the initial effective analysis period; Calculate the distribution concentration of each initial effective segment and determine the initial effective segments of the analysis that have a distribution concentration greater than a preset concentration threshold as effective segments of the analysis; Obtain representative points of lagged terms in the effective segment of the analysis, and calculate the activity of the representative points corresponding to each lagged term representative point; Based on the activity of the representative points, the lag difference weights corresponding to the lag term representative points are determined, and the optimized lag term representative points are obtained. Based on the pre-defined correspondence between the lag difference weights and the number of extra points, determine the number of extra points for each lag term representing a point; The optimized hysteresis representative points are used as the input of the sample data for the ARIMA algorithm. The number of additional points for each hysteresis representative point is used as the number of insertion sample points for the hysteresis difference of the ARIMA algorithm. The ARIMA algorithm is then run to output the predicted power data.

2. The enterprise energy economic management optimization method based on artificial intelligence according to claim 1, characterized in that, The determination of baseline data and target data from historical electricity data specifically includes: The data in the nth period of the historical power data is determined as the baseline data. The historical power data is obtained by dividing the power data according to a preset time period, and the time axis of the historical power data in different preset time periods is the same. The data in the nj-th cycle of historical electricity data is determined as the target data, where n and j are positive integers, n is greater than j, and the data in the nj-th cycle is not the data in the first cycle.

3. The enterprise energy economic management optimization method based on artificial intelligence according to claim 2, characterized in that, The calculation of the time difference between the target data and the baseline data specifically includes: The peak electricity consumption times of the target data and the reference data are obtained to obtain the target time sequence and the reference time sequence, respectively. The target data is the data of the njth period in the historical electricity data, and the reference data is the data of the nth period in the historical electricity data. Calculate the absolute value of the difference between the m-th element in the target time series and the m-th element in the reference time series, and determine it as the time difference between the target data and the reference data at the m-th time. Obtain the time difference at the m-th time of each moment between the target data and the benchmark data, and sum them to obtain the time difference at the j-th period between the target data and the benchmark data; The data from the njth cycle to the (n-1th cycle) of the historical power data are taken as target data. The time difference between the target data and the reference data in the jth cycle is obtained and summed to obtain the time difference between the target data and the reference data.

4. The enterprise energy economic management optimization method based on artificial intelligence according to claim 1, characterized in that, The determination of the effective analysis segment duration based on time differences specifically includes: Calculate the peak distribution of target data and baseline data based on time differences; The period length between the number of periods of the target data with the highest peak distribution and the number of periods of the benchmark data is determined as the effective analysis period.

5. The enterprise energy economic management optimization method based on artificial intelligence according to claim 4, characterized in that, The calculation of the peak distribution of target data and baseline data based on time differences specifically includes: The data from the njth period to the (n-1th period) of the historical electricity data are respectively used as target data. The mean value of the data corresponding to the peak electricity consumption time in the target data is determined as the mean value of the target electricity peak. The mean value of the data corresponding to the peak electricity consumption time in the benchmark data is obtained to obtain the mean value of the benchmark electricity peak. Peak distribution is calculated based on time differences, the average peak value of the target energy, and the average peak value of the reference energy.

6. The enterprise energy economic management optimization method based on artificial intelligence according to claim 1, characterized in that, The calculation of the distribution concentration of each initial analysis effective segment specifically includes: A short-time Fourier transform is performed on each effective segment of the initial analysis to obtain the frequency domain results; The distribution concentration is calculated based on the maximum energy value of the frequency components and the sum of the energy values ​​in the frequency domain results.

7. The enterprise energy economic management optimization method based on artificial intelligence according to claim 1, characterized in that, The acquisition of representative points of lag terms in the valid segment of the analysis specifically includes: The least squares method is used to perform curve fitting on the effective segment to obtain the effective segment curve; The maximum and minimum points in the effective segment curve are determined as representative points of the lag term.

8. The enterprise energy economic management optimization method based on artificial intelligence according to claim 1, characterized in that, The calculation of the representative point activity corresponding to each lag term representative point specifically includes: The least squares method is used to fit the representative points of the lag term to obtain a continuous curve; Obtain the slope range of the continuous curve and divide the slope range according to the preset slope length to obtain the slope interval; Calculate the activity of each slope interval, and determine the activity of the slope interval corresponding to the slope of the representative point of the lag term as the activity of the representative point corresponding to the lag term.

9. The enterprise energy economic management optimization method based on artificial intelligence according to claim 8, characterized in that, The calculation of the activity of each slope interval specifically includes: The average number of representative points of the lag term in the p-th slope interval of each effective segment of the analysis is determined as the average number of representative points in the p-th slope interval. Obtain the median slope of the p-th slope interval; Calculate the absolute value of the difference between the number of representative points of the lagged terms in the p-th slope interval of the u-th effective analysis segment and the number of representative points of the lagged terms in the p-th slope interval of the (u-1)-th effective analysis segment to obtain the difference in the number of representative points of the u-th effective analysis segment, where the u-th effective analysis segment is not the first effective analysis segment; The reciprocal of the difference in the number of representative points is used to obtain the reciprocal of the difference in the number of valid segments in the analysis for the uth segment. Obtain the reciprocal of the quantity difference for each valid segment and sum them to get the reciprocal of the quantity difference for the p-th slope interval; The activity level of the p-th slope interval is calculated based on the mean number of representative points in the p-th slope interval, the median slope of the p-th slope interval, and the reciprocal of the difference in the number of points in the p-th slope interval. Obtain the activity level for each slope interval.

10. An artificial intelligence-based enterprise energy economic management optimization system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of the enterprise energy economic management optimization method based on any one of claims 1-9.

Images