A method and system for predicting charging demand of agricultural machinery in cold regions in rural microgrid

Abstract

The present application relates to the field of smart grid technology, and specifically to a method and system for predicting charging demand of cold-region agricultural machinery in a rural microgrid, the method comprising: obtaining the total charging power of all charging piles at each sampling time of each day as the observed value at each sampling time, and obtaining the corresponding predicted value in combination with the Monte Carlo simulation algorithm; based on the trend characteristics of the observed value in the local time window, combined with the difference between the observed value and the fitted value in the local time window, obtaining the load interference degree at each sampling time; based on the data fluctuation of multiple sampling times before each sampling time, obtaining the deviation degree at each sampling time, combining the load interference degree to obtain the noise evaluation factor at each sampling time, to obtain the adaptive Kalman gain, and obtain the optimal predicted value at each sampling time and the predicted correction value at the future time. The present application can improve the prediction accuracy of the charging load and ensure the accuracy of the demand prediction.

Description

Technical Field

[0001] The present application relates to the field of smart grid technology, and in particular to a method and system for predicting charging demand for cold-region agricultural machinery vehicles in a rural microgrid. Background Art

[0002] Microgrid is a small power generation and distribution system, mainly composed of distributed power sources, energy storage devices, energy conversion devices, loads, monitoring and protection devices, etc. It can be connected to the external power grid, disperse electricity demand, and improve energy self-sufficiency. It is an important part of the current smart grid. Since cold-region villages have fertile land and sparse population, they are very suitable for the development of microgrids. In order to adapt to the development of new energy and meet the farming needs of vast land, electric agricultural machinery is generally used in rural areas. Electric agricultural machinery has the characteristics of high battery density, fast power consumption, and high charging requirements. Therefore, it is necessary to predict the charging needs of electric agricultural machinery, so as to better realize the capacity expansion, dispatching operation and peak load regulation capacity construction in microgrids.

[0003] When the Monte Carlo algorithm is used to predict the charging load of electric agricultural machinery, the prediction results deviate greatly from the true values ​​due to the influence of pseudo-random numbers generated by computer simulation, so it is necessary to make prediction corrections based on the prediction results. When the Kalman filter algorithm is used for correction, the same noise evaluation factor is often used, which will cause the cumulative effect of correction errors, causing the correction value of the prediction results to gradually deviate from the true charging load sequence. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a method and system for predicting the charging demand of agricultural machinery in cold regions in a rural microgrid. The technical solutions adopted are as follows:

[0005] The embodiment of the present application provides a method for predicting charging demand of agricultural machinery in cold regions in a rural microgrid, comprising the following steps:

[0006] The total charging power of all charging piles at each sampling time of each day is obtained as the observed value at each sampling time, and the corresponding predicted value is obtained by combining the Monte Carlo simulation algorithm;

[0007] The multiple sampling moments before each sampling moment constitute a local time window of each sampling moment, and the trend evaluation coefficient of each sampling moment is calculated based on the trend characteristics of the observed values ​​in the local time window. The load interference degree of each sampling moment is obtained by combining the difference between the observed value and the fitted value in the local time window;

[0008] Analyze the difference between the observed value and the predicted value corresponding to the multiple sampling moments before each sampling moment, obtain the deviation of each sampling moment based on the range and average level of the difference at the multiple sampling moments before each sampling moment, and obtain the noise assessment factor of each sampling moment in combination with the load interference degree;

[0009] The adaptive Kalman gain at each sampling moment is obtained according to the noise evaluation factor at each sampling moment to obtain the optimal prediction value at each sampling moment. The prediction correction value at the future moment is obtained by combining the prediction value at the future moment and the distance between the future moment and the current sampling moment.

[0010] Preferably, the trend evaluation coefficient calculation formula at each sampling moment is:

[0011] Where B represents the trend assessment coefficient at the current sampling moment, Δτ represents the intensity of the trend term after time series decomposition of all observations in the local time window at the current sampling moment, exp() represents an exponential function with a natural constant as the base, and ΔL represents the slope of the corresponding fitting straight line for all observations in the local time window at the current sampling moment.

[0012] Preferably, the calculation formula for the load interference degree at each sampling moment is:

[0013] Where A represents the load interference degree at the current sampling time, W represents the number of observations in the local time window at the current sampling time, and L i and R i They represent the fitted value of the i-th sampling time on the corresponding fitting line in the local time window of the current sampling time and the observed value at the i-th sampling time, respectively. To take the absolute value.

[0014] Preferably, the difference value is the absolute value of the difference between the observed value and the predicted value corresponding to each sampling moment.

[0015] Preferably, the corresponding calculation formula of the deviation degree at each sampling moment is:

[0016] C=R×μ; wherein C represents the deviation at the current sampling moment, and R and μ represent the range and mean of all the difference values ​​in the local time window at the current sampling moment, respectively.

[0017] Preferably, the noise evaluation factor at each sampling moment corresponds to the calculation formula:

[0018] ω=ω 0 ×ε×norm(A+C); where ω represents the noise evaluation factor at the current sampling moment, ω 0represents the preset initial noise estimation value, ε represents the preset scaling factor, Norm() represents the normalization function, A and C represent the load interference degree and deviation degree at the current sampling moment respectively.

[0019] Preferably, the adaptive Kalman gain at each sampling moment further includes:

[0020] According to the predicted value, observed value and noise evaluation factor at each sampling moment, the Kalman filter algorithm is used to obtain the Kalman gain at each sampling moment.

[0021] Preferably, the optimal prediction value further includes:

[0022] According to the optimal estimated value of the previous sampling moment at each sampling moment, the observed value at each sampling moment, the predicted value and the adaptive Kalman gain, the Kalman filter algorithm is used to obtain the optimal predicted value at each sampling moment.

[0023] Preferably, the predicted correction value further includes:

[0024] For the t-th moment after the current sampling moment of the day, the optimal prediction value at the current moment, the prediction value at the t-th moment, and the Kalman gain at the t-th moment are used, and the Kalman filter algorithm is used to obtain the prediction correction value at the t-th moment;

[0025] Among them, the Kalman gain K at the tth moment t The expression of K is: t =K′ t +K o / (tt o ), where K′ t represents the Kalman gain at the tth moment of the day before the current day. If the current day is the first prediction, then K′ t =0, K o Indicates the current sampling time t o The Kalman gain.

[0026] An embodiment of the present application also provides a charging demand prediction system for cold-region agricultural machinery in a rural microgrid, the system comprising a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the steps of any one of the above methods when executing the computer program.

[0027] From the above, it can be seen that the method and system for predicting charging demand of cold-region agricultural machinery vehicles in a rural microgrid provided by the present application have at least the following beneficial effects:

[0028] The present application obtains predicted data and observed data through the load prediction module and the data acquisition module, obtains the load interference degree at each sampling moment according to the trend characteristics and volatility characteristics of the observed data at each sampling moment, reflects the actual situation and credibility of the observed data at each sampling moment, and obtains the deviation degree in combination with the difference characteristics of the observed data and the predicted data at each sampling moment, measures the correction strength when correcting the predicted data at each sampling moment, and finally obtains the noise assessment factor at a single sampling moment. The noise assessment factor changes dynamically with the change of the observed data obtained by real-time sampling, accurately assesses the observation noise, and obtains the corresponding Kalman gain, obtains the optimal prediction for the time period for which the observed data has been collected, and obtains the corresponding prediction correction value for the time period for which the observed data has not been collected, that is, the future moment.

[0029] Compared with the traditional Monte Carlo simulation algorithm, this application aims to introduce the Kalman filtering algorithm to correct the predicted data. At the same time, during the correction process, the corresponding noise assessment factor is obtained according to the distribution of the observed data and the predicted data, which can more accurately reflect the data fluctuation and obtain the corresponding correction strength. Therefore, the corrected prediction result is closer to the actual charging load demand of rural electric agricultural machinery, which is beneficial to the smart grid's allocation of electric energy to the rural microgrid and realizes the reasonable distribution of electric energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0031] Figure 1 A flowchart of the steps of a method for predicting charging demand of agricultural machinery in cold regions in a rural microgrid provided in this application;

[0032] Figure 2 A schematic diagram of the prediction correction process provided for this application;

[0033] Figure 3 This is a schematic diagram of the prediction correction effect provided in this application. DETAILED DESCRIPTION

[0034] In order to further explain the technical means and effects adopted by this application to achieve the predetermined invention purpose, the following is a detailed description of the charging demand prediction method and system for cold-region agricultural machinery vehicles in a rural microgrid proposed by this application, in combination with the accompanying drawings and preferred embodiments, as well as its specific implementation method, structure, features and effects. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0035] Unless otherwise specified and limited, terms such as "comprises", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a circuit structure, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such articles or devices. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the article or device including the element. In addition, the term "and\or" used herein includes any and all combinations of one or more related listed items. All technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of this application.

[0036] The following is a detailed description of a method and system for predicting charging demand for cold-region agricultural machinery vehicles in a rural microgrid provided by the present application in conjunction with the accompanying drawings.

[0037] See also Figure 1 , which shows a flowchart of a method for predicting charging demand of cold-region agricultural machinery vehicles in a rural microgrid provided by an embodiment of the present application, including the following steps:

[0038] Step 100: Obtain the total charging power of all charging piles at each sampling time every day as the observed value at each sampling time, and obtain the corresponding predicted value in combination with the Monte Carlo simulation algorithm.

[0039] Taking Q, a rural area in a cold region, as an example, the area uses a "wind, solar, heat and storage" integrated energy system to build a microgrid system. In order to better balance the power supply in the smart grid and the internal power demand, the charging load of electric agricultural machinery is predicted. The charging load of electric agricultural machinery in the area is closely related to the current number of agricultural machinery in the area.

[0040] Therefore, no data prediction is performed within the initial week of data collection. The total charging power of all charging piles in the current area is obtained through the new energy vehicle monitoring platform, that is, the sum of the charging power of all charging piles at each sampling time is used as the observation value at each sampling time. The day is divided into 1440 minutes using a 24-hour system. The charging load collected every day constitutes the daily charging load observation sequence, which is recorded as the observation sequence.

[0041] The number of electric agricultural machinery in the current area is counted, recorded as N, and then the Monte Carlo simulation algorithm is used to obtain the charging load forecast data corresponding to each day, where the setting conditions of the Monte Carlo simulation algorithm are: the number of simulations is set to 1000 times, the number of electric agricultural machinery vehicles: the number of vehicles in the first prediction is 80% of N, and 80% of the charging trips on the previous day are determined as the number of vehicles in other predictions. In this embodiment, the battery capacity of all vehicles is about 50KWh, the charging power of the slow charging pile is set to 7KW, and the charging power of the fast charging pile is set to 40KW. The starting charging time is mainly concentrated at 11:00 and 18:00, and the standard deviation of the starting charging time is a random number between about 1 and 2h. The corresponding charging load forecast sequence for each day is thus obtained, which is recorded as the forecast sequence for each day. It should be noted that in order to keep consistent with the time scale of the observation sequence, the charging load forecast sequence is also in minutes.

[0042] Step 200: Multiple sampling moments before each sampling moment constitute a local time window of each sampling moment, calculate the trend evaluation coefficient of each sampling moment based on the trend characteristics of the observed values ​​in the local time window, and combine the difference between the observed value and the fitted value in the local time window to obtain the load interference degree of each sampling moment.

[0043] In order to better configure the electric energy resources of microgrids under smart grids, it is necessary to predict the charging demand of electric agricultural machinery, which has the largest electricity demand in microgrids. Due to the influence of Monte Carlo pseudo-random numbers in the prediction sequence, the prediction results deviate from the actual results. In the observation sequence, the charging demand of charging piles does not all come from electric agricultural machinery, and the new energy vehicle monitoring platform does not include all charging piles in the area. Therefore, there are also observation errors in the observation sequence. It is necessary to correct the prediction data by combining the prediction sequence and the observation sequence.

[0044] Firstly, according to the number of charging vehicles on the new energy vehicle monitoring platform of the previous day, the charging vehicles of electric agricultural machinery on the previous day were determined. Specifically, the charging vehicles with a charging time greater than 2.5h were defined as an electric agricultural machinery. The number of charging vehicles was determined, and the Monte Carlo simulation algorithm was used to obtain the prediction sequence for the day.

[0045] The new energy vehicle monitoring platform is used to count the charging power of all charging piles on the day in real time with a sampling interval of 1 minute to obtain the corresponding observation sequence. The data distribution of the observation sequence is used to correct the data of the prediction sequence. When using Kalman filtering for data correction, the noise evaluation factor determines the credibility of the observation data. Therefore, the size of the noise evaluation factor should be adjusted in real time based on the dynamic distribution of the observation data to ensure that the actual observation situation can be reflected in a timely manner.

[0046] Based on the real-time observation sequence collected on the day, there will be more noise points in the observation sequence due to the randomness of charging pile plugging and unplugging and the fluctuation of charging power. Since the charging demand of electric agricultural machinery in cold rural areas has certain regular characteristics, such as the morning agricultural machinery output is over at 11 o'clock, so a short charging is generally carried out at 11 o'clock to ensure the output demand in the afternoon. Similarly, after the electric agricultural machinery power output ends at 18 o'clock, it will be fully charged to meet the demand of the next day, so the observation sequence of the day has certain trend characteristics.

[0047] In order to measure the trend characteristics of each sampling moment in the observation sequence, multiple sampling moments before each sampling moment constitute the local time window of each sampling moment. In this embodiment, taking the current sampling moment as an example, W sampling moments are intercepted in the past as the local time window of the current sampling moment. In this embodiment, W is set to 60. If the number of sampling moments in the local time window of the day is less than W, the sampling moments in the previous day can be intercepted to ensure that there are W sampling moments in the local time window.

[0048] Based on the above analysis, based on the trend of the data in the local time window at each sampling moment, the trend evaluation coefficient B at each sampling moment is obtained. Taking the current sampling moment as an example, the expression of the trend evaluation coefficient is specifically:

[0049] In the formula, B represents the trend evaluation coefficient at the current sampling moment, Δτ represents the trend item strength after time series decomposition of all observations in the local time window at the current sampling moment, wherein, in this embodiment, STL (time series decomposition algorithm) is used to decompose all observations in the local time window, and the method of calculating the trend item strength by using trend item components and trend item strength formula after decomposition is the prior art, exp() represents an exponential function with a natural constant as the base, ΔL represents the slope of the corresponding fitting straight line of all observations in the local time window at the current sampling moment, and the least squares method is used for fitting in this embodiment.

[0050] Among them, STL decomposition, trend strength and least square method are all well-known technologies and will not be described in detail in this embodiment.

[0051] If the trend characteristics of the observed data points within the local time window determined by the current sampling time (the total charging power of the charging pile obtained by the monitoring platform) are not obvious, the value of the trend intensity obtained at this time is small, indicating that the current time is not the time for charging the electric agricultural machinery. The charging load at this time may be caused by charging fluctuations of other models. At the same time, if the slope of the fitted straight line is a negative value, it indicates that the current total charging power is gradually decreasing, that is, the total charging power value decreases. At this time, it is more susceptible to external interference, making the value of the trend evaluation coefficient larger.

[0052] The trend evaluation coefficient B is used to analyze the possibility that the observed data may be disturbed, and the load interference degree A at each sampling time is obtained by further combining the difference between the observed value and the fitted value in the local time window:

[0053] Where A represents the load interference degree at the current sampling time, W represents the number of observations in the local time window, and L i and R i They represent the fitted value of the i-th sampling time on the corresponding fitting line in the local time window of the current sampling time and the observed value at the i-th sampling time, respectively. To take the absolute value.

[0054] If the data point distribution trend in the local time window at the sampling moment is small, it indicates that the data point has a weaker anti-interference ability and is more susceptible to interference. At this time, the corresponding trend evaluation coefficient value is larger. At the same time, the deviation between the actual observation value and the straight line fitting value in the local time window is larger, indicating that the local time window observation data of the data point is more seriously disturbed by the outside world, and the load interference value of the data point is larger. On the contrary, if the data point corresponds to a data point with a strong trend, it indicates that the data point is just located in the main charging interval of the electric agricultural machinery. At this time, due to the large demand for charging power of the electric agricultural machinery, the trend is strong and the anti-interference ability is strong, and the load interference value is smaller.

[0055] Step 300: Analyze the difference between the observed values ​​and the predicted values ​​corresponding to the multiple sampling moments before each sampling moment, obtain the deviation of each sampling moment based on the extreme difference and average level of the difference at the multiple sampling moments before each sampling moment, and obtain the noise assessment factor of each sampling moment in combination with the load interference degree.

[0056] By analyzing the observation sequence, the load interference degree at each sampling moment can be obtained in real time to measure the credibility of the observation data at each sampling moment. However, when using Kalman to correct the data, it is necessary to correct the prediction sequence based on the observation sequence and the prediction sequence, so it is necessary to further analyze the deviation of the prediction sequence and the observation sequence at the current sampling moment.

[0057] Since the prediction sequence is based on the data of the previous day, the prediction sequence is the complete data of the whole day, while the observation sequence is collected in real time, including part of the data of the day, and changes in real time. If the difference between the observed data and the predicted data in the local time window at the current sampling moment is large, it indicates that one or both of the predicted value and the observed value are greatly affected by the outside world, and the correction strength should be large at this time; on the contrary, if the difference between the observed data and the predicted data in the local time window at the current sampling moment is small, it indicates that the prediction is more accurate, and the correction strength is small at this time.

[0058] Therefore, the difference between the observed value and the predicted value at each sampling moment in the local time window is analyzed. In this embodiment, the absolute value of the difference between the observed value and the predicted value at each sampling moment in the local time window is taken as the difference value at each sampling moment in the local time window. Further, the deviation C of each sampling moment is obtained by combining the extreme difference and the average level of the difference values ​​of all sampling moments in the local time window of each sampling moment. The specific calculation formula in this embodiment is:

[0059] C=R×μ; wherein C represents the deviation degree at the current sampling moment, R represents the range of all the difference values ​​in the local time window at the current sampling moment, and μ represents the mean of all the difference values ​​in the local time window at the current sampling moment.

[0060] If the data change trends of the predicted data and the observed data in the local time window at the current sampling moment are more consistent, the difference value of each sampling moment in the local time window will be smaller, and thus the range and mean of all difference values ​​will be smaller, and finally the deviation degree of the current sampling moment will be smaller.

[0061] The load interference degree is obtained through the distribution of observed data, and the deviation degree is obtained through the difference between observed data and predicted data. Both reflect the correction strength of the predicted value at the current sampling time. Combining the two, the noise assessment factor ω at the current sampling time is obtained:

[0062] ω=ω 0 ×ε×norm(A+C); where ω represents the noise evaluation factor at the current sampling moment, ω 0 represents the preset initial noise estimation value, which is 10 in this embodiment, and ε represents the preset scaling factor, which is set to 2 in this embodiment, in order to achieve the scaling of the noise assessment factor and limit the range of the noise assessment factor. Norm() represents the normalization function, and A and C represent the load interference degree and deviation degree at the current sampling moment, respectively.

[0063] For the current sampling moment, if the observation sequence is affected by statistical errors and there is more noise, the deviation between the observed data and the predicted data will be greater, the noise evaluation factor obtained at this time will be greater, and the Kalman gain in the Kalman algorithm will be smaller. At this time, the correction result of the predicted data is more dependent on the predicted value. On the contrary, if the fluctuation of the observation sequence is smaller and the data authenticity is higher, the deviation between the predicted data and the observed data is smaller, the value of the noise evaluation factor is smaller, and the calculated Kalman gain is larger. At this time, the correction result of the predicted data is more dependent on the observed value.

[0064] Step 400: Obtain the adaptive Kalman gain at each sampling moment according to the noise evaluation factor at each sampling moment to obtain the optimal prediction value at each sampling moment, and obtain the prediction correction value at the future moment by combining the prediction value at the future moment and the distance between the future moment and the current sampling moment.

[0065] By using step 300, the noise evaluation factor corresponding to each sampling moment can be obtained in real time. Further, according to the predicted value, observed value and noise evaluation factor at each sampling moment, the Kalman filter algorithm is used to obtain the adaptive Kalman gain K at each sampling moment, thereby obtaining the optimal predicted value and subsequent predicted correction value at the observation moment. Among them, the specific calculation process of the Kalman gain is the prior art, and the noise evaluation factor is used to characterize the observed noise at each sampling moment. According to the above process of this embodiment, the noise evaluation factor at each sampling moment can be obtained to analyze the observed noise, and the adaptive Kalman gain at each sampling moment is obtained by combining the observed value and predicted value at each sampling moment through the Kalman filter algorithm.

[0066] For the optimal prediction value: According to the prediction value, observation value and adaptive Kalman gain at each sampling moment, combined with the optimal estimate value of the previous moment of each sampling moment, the optimal prediction value of each sampling moment is calculated. Specifically, for the i-th moment of the day, the observation value at the i-th moment is collected, and the prediction value at the i-th moment is obtained (obtained by Monte Carlo simulation), thereby determining the adaptive Kalman gain K at the i-th moment;

[0067] The optimal estimated value at the i-1th moment, the observed value at the i-th moment, the predicted value and the adaptive Kalman gain are combined to obtain the optimal predicted value at the i-th moment using the Kalman filter algorithm. It should be noted that in this embodiment, the optimal estimated value at the initial moment is defined as 0. The calculation process for obtaining the optimal estimated value at each sampling moment is a well-known technology and will not be elaborated in detail in this embodiment.

[0068] Furthermore, this embodiment will correct the predicted value at the future moment to ensure the accuracy of the predicted data. For the predicted correction value: according to the distance between the sampling moments, the Kalman gain of the future sampling moment is analyzed. Specifically, for the t-th moment after the i-th moment of the day (the observation value has not been collected yet, that is, the future moment), the corresponding Kalman gain is calculated. The specific expression is: K t =K′ t +K o / (tt o ), where K t represents the Kalman gain at the tth moment after the current sampling moment of the day, K′t represents the Kalman gain at the tth moment of the day before the current moment, and if the current day is the first prediction, K′ t =0, K oIndicates the current sampling time t o It should be noted that the sampling time refers to the real time scale. For example, the data is collected at the 524th minute of the day. The farther the tth moment is from the current sampling moment, the more the Kalman gain at the current sampling moment decays.

[0069] Specifically, the prediction correction flow chart in this embodiment is as follows: Figure 2 As shown, Figure 2 The process of obtaining the optimal prediction value at the i-th sampling moment is given in , and the prediction value at the i+1 moment and the corresponding prediction correction value are obtained. By recursively deducing, the optimal prediction value at each sampling moment can be obtained, and the prediction correction value at each subsequent moment can be obtained to realize the correction of the prediction data.

[0070] Specifically, the prediction correction value at the tth moment is as follows: the optimal prediction value at the i-th moment, the prediction value at the t-th moment, and the Kalman gain corresponding to the t-th moment are substituted into the Kalman filter algorithm, and the prediction correction value at the t-th moment is output, and the prediction correction values ​​at other times of the day are obtained recursively. Thus, the above process of this embodiment is used for the daily charging power to obtain the corrected prediction data.

[0071] The specific diagram of the prediction correction effect is as follows: Figure 3 As shown in the figure, since the charging power prediction is affected by the Monte Carlo simulation pseudo-random number, and the observed value is affected by the statistical error, neither can reflect the actual charging load demand of electric agricultural machinery, so it is necessary to use the observed value to correct the predicted value. Figure 3 When the observable time is 3h, the predicted value is corrected to obtain the predicted corrected value.

[0072] The prediction and correction results can reduce the impact of external interference, more accurately reflect the actual charging needs of agricultural machinery vehicles, help allocate power to microgrids in smart grids, and reduce the power load pressure of microgrids.

[0073] Based on the same inventive concept as the above method, an embodiment of the present application also provides a charging demand prediction system for cold-region agricultural machinery vehicles in a rural microgrid, comprising a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned methods for predicting charging demand for cold-region agricultural machinery vehicles in a rural microgrid are implemented.

[0074] It is to be understood that the sequence of the embodiments of the present application described above is for description only and does not represent the advantages and disadvantages of the embodiments. The above describes specific embodiments of the present specification. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0075] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0076] The above content is only an implementation method of the present application and is not intended to limit the scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the protection scope of the present application.

Claims

1. A method for predicting charging demand of agricultural machinery in cold regions in a rural microgrid, characterized in that: The following steps are involved: The total charging power of all charging piles at each sampling time of each day is obtained as the observed value at each sampling time, and the corresponding predicted value is obtained by combining the Monte Carlo simulation algorithm; The multiple sampling moments before each sampling moment constitute a local time window of each sampling moment, and the trend evaluation coefficient of each sampling moment is calculated based on the trend characteristics of the observed values ​​in the local time window. The load interference degree of each sampling moment is obtained by combining the difference between the observed value and the fitted value in the local time window; Analyze the difference between the observed value and the predicted value corresponding to the multiple sampling moments before each sampling moment, obtain the deviation of each sampling moment based on the range and average level of the difference at the multiple sampling moments before each sampling moment, and obtain the noise assessment factor of each sampling moment in combination with the load interference degree; The adaptive Kalman gain at each sampling moment is obtained according to the noise evaluation factor at each sampling moment to obtain the optimal prediction value at each sampling moment, and the prediction correction value at the future moment is obtained by combining the prediction value at the future moment and the distance between the future moment and the current sampling moment; The trend evaluation coefficient calculation formula at each sampling time is: In the formula, B represents the trend assessment coefficient at the current sampling time, Δτ represents the trend term intensity after time series decomposition of all observations in the local time window at the current sampling time, exp() represents an exponential function with a natural constant as the base, and ΔL represents the slope of the corresponding fitting straight line for all observations in the local time window at the current sampling time; The calculation formula of the load interference degree at each sampling time is: Where A represents the load interference degree at the current sampling time, W represents the number of observations in the local time window at the current sampling time, and L i and R i Respectively represent the fitting value of the i-th sampling time on the corresponding fitting line in the local time window of the current sampling time and the observed value at the i-th sampling time, || is the absolute value; The difference value is the absolute value of the difference between the observed value and the predicted value corresponding to each sampling time; The corresponding calculation formula of the deviation degree at each sampling time is: C = R × μ; where C represents the deviation at the current sampling moment, and R and μ represent the range and mean of all the difference values ​​in the local time window at the current sampling moment, respectively; The corresponding calculation formula of the noise evaluation factor at each sampling time is: ω=ω0×ε×norm(A+C); where ω represents the noise evaluation factor at the current sampling moment, ω0 represents the preset initial noise estimation value, ε represents the preset scaling factor, norm() represents the normalization function, and A and C represent the load interference degree and deviation degree at the current sampling moment, respectively; The adaptive Kalman gain at each sampling moment further includes: According to the predicted value, observed value and noise evaluation factor at each sampling moment, the Kalman filter algorithm is used to obtain the Kalman gain at each sampling moment. The optimal prediction value further includes: According to the optimal estimated value of the previous sampling moment at each sampling moment, the observed value at each sampling moment, the predicted value and the adaptive Kalman gain, the Kalman filter algorithm is used to obtain the optimal predicted value at each sampling moment; The forecast correction value further includes: For the t-th moment after the current sampling moment of the day, the optimal prediction value at the current moment, the prediction value at the t-th moment, and the Kalman gain at the t-th moment are used, and the Kalman filter algorithm is used to obtain the prediction correction value at the t-th moment; Among them, the Kalman gain K at the tth moment t The expression of K is: t =K t ′ +K o / (tt o ), where K t ′ represents the Kalman gain at the tth moment of the day before the current day. If the current day is the first prediction, K t ′ =0, K o Indicates the current sampling time t o The Kalman gain.

2. A charging demand prediction system for cold-region agricultural machinery in a rural microgrid, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to claim 1 are implemented.

Images