Intelligent pumping method, device, equipment and medium based on oil well production prediction

Abstract

Embodiments of the present disclosure disclose an intelligent oil pumping method and device based on oil well production prediction, equipment and medium. A specific embodiment of the method comprises: in response to receiving an oil pumping signal sent by an intelligent oil pumping device, obtaining a historical production data sequence and oil well parameter structured data; constructing a production prediction vector; constructing a production prediction matrix; performing production probability prediction processing on the production prediction matrix and the historical production data sequence; performing data screening processing on the historical production data sequence; performing production prediction processing on the production prediction matrix and the target production data sequence; generating an oil well production prediction data sequence; and controlling the intelligent oil pumping device to perform an intelligent oil pumping task. The embodiment can reduce the hardware wear rate of the intelligent oil pumping device and reduce invalid power consumption.

Description

Technical Field

[0001] Embodiments of this disclosure relate to the field of computer technology, and more specifically to intelligent pumping methods, apparatus, equipment, and media based on oil well production prediction. Background Technology

[0002] Intelligent oil pumping is a method of intelligently controlling the start and stop of the pumping unit based on the oil well's production. Currently, the common approach to intelligent oil pumping is to control the pumping equipment to pump oil at fixed time intervals.

[0003] However, when using the above-mentioned method for intelligent oil extraction, the following technical problems often arise: Relying solely on fixed time intervals to control the oil pumping equipment can easily lead to the equipment running idle, resulting in ineffective power consumption and equipment wear.

[0004] The information disclosed in this background section is only intended to enhance the understanding of the background of the present disclosure concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0006] Some embodiments of this disclosure provide intelligent pumping methods, apparatuses, electronic devices, and computer-readable media based on well production prediction to address one or more of the technical problems mentioned in the background section above.

[0007] In a first aspect, some embodiments of this disclosure provide an intelligent pumping method based on oil well production prediction. The method includes: in response to receiving a pumping signal sent by an intelligent pumping device, acquiring a historical production data sequence and structured oil well parameter data; for each future time point in a preset future time point sequence, constructing a production prediction vector corresponding to the future time point based on the historical production data sequence, the structured oil well parameter data, and the future time point; constructing a production prediction matrix based on the constructed production prediction vectors; performing production probability prediction processing on the production prediction matrix and the historical production data sequence to obtain a production probability data sequence; performing data filtering processing on the historical production data sequence to obtain a target production data sequence; performing production prediction processing on the production prediction matrix and the target production data sequence to obtain an oil well production data sequence; generating an oil well production prediction data sequence based on the production probability data sequence and the oil well production data sequence; and controlling the intelligent pumping device to perform an intelligent pumping task based on the oil well production prediction data sequence.

[0008] Secondly, some embodiments of this disclosure provide an intelligent pumping device based on oil well production prediction. The device includes: an acquisition unit configured to acquire historical production data sequences and structured oil well parameter data in response to receiving a pumping signal sent by an intelligent pumping device; a first construction unit configured to construct a production prediction vector corresponding to each future time point in a preset future time point sequence, based on the historical production data sequence, the structured oil well parameter data, and the future time point; a second construction unit configured to construct a production prediction matrix based on the constructed production prediction vectors; and a production probability prediction unit configured to... The production prediction matrix and the historical production data sequence are processed to obtain a production probability data sequence; a data filtering unit is configured to process the historical production data sequence to obtain a target production data sequence; a production prediction unit is configured to process the production prediction matrix and the target production data sequence to obtain an oil well production data sequence; a generation unit is configured to generate an oil well production prediction data sequence based on the production probability data sequence and the oil well production data sequence; and a control unit is configured to control the intelligent pumping equipment to perform intelligent pumping tasks based on the oil well production prediction data sequence.

[0009] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.

[0010] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

[0011] The above embodiments of this disclosure have the following beneficial effects: The intelligent oil pumping method based on well production prediction in some embodiments of this disclosure can reduce the waste of power resources and reduce ineffective wear and tear on equipment. Specifically, the reason for ineffective power consumption and equipment wear is that relying solely on fixed time intervals to control the oil pumping equipment can easily lead to idling, resulting in ineffective power consumption and equipment wear. Based on this, the intelligent oil pumping method based on well production prediction in some embodiments of this disclosure firstly acquires historical production data sequences and structured well parameter data in response to receiving a pumping signal from the intelligent oil pumping equipment. This yields initial data. Secondly, for each future time point in a preset future time point sequence, a production prediction vector corresponding to the future time point is constructed based on the aforementioned historical production data sequence, the aforementioned structured well parameter data, and the aforementioned future time point. This yields a production prediction vector. Then, a production prediction matrix is ​​constructed based on the constructed production prediction vectors. This yields a production prediction matrix. Finally, production probability prediction processing is performed on the aforementioned production prediction matrix and the aforementioned historical production data sequence to obtain a production probability data sequence. Therefore, the production probability of the oil well at various future time points can be predicted. Then, the historical production data sequence is filtered to obtain the target production data sequence. This removes zero values ​​from the historical production data sequence. Next, the production prediction matrix and the target production data sequence are processed to obtain the oil well production data sequence. This yields the oil well's production value. Then, based on the production probability data sequence and the oil well production data sequence, an oil well production prediction data sequence is generated. This provides the oil well's production at various future time points. Finally, based on the oil well production prediction data sequence, the intelligent pumping equipment is controlled to perform intelligent pumping tasks. This allows determination of whether the intelligent pumping equipment needs to be activated for oil well extraction at future time points, and the pumping frequency of the intelligent pumping equipment during extraction, based on the oil well's production at each future time point. Because the production of oil wells at various future time points can be predicted in advance, and then the intelligent pumping equipment can be controlled to pump oil based on the production of oil wells at various future time points, instead of relying solely on fixed time intervals for pumping, the probability of the intelligent pumping equipment running idle due to low oil well production can be reduced. This can reduce the waste of power resources and ineffective equipment wear caused by the idling of intelligent pumping equipment. Attached Figure Description

[0012] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0013] Figure 1 This is a flowchart of some embodiments of the intelligent pumping method based on oil well production prediction according to the present disclosure; Figure 2 This is a schematic diagram of the structure of some embodiments of an intelligent oil pumping device based on oil well production prediction according to the present disclosure; Figure 3 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0014] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0015] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0016] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0017] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0018] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0019] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] Figure 1A flow 100 of some embodiments of the smart pumping method based on well production prediction according to this disclosure is shown. This smart pumping method based on well production prediction includes the following steps: Step 101: In response to receiving the pumping signal sent by the intelligent pumping equipment, acquire the historical production data sequence and structured data of oil well parameters.

[0021] In some embodiments, the execution entity (e.g., a computing device) of the intelligent pumping method based on well production prediction can acquire historical production data sequences and structured well parameter data in response to receiving a pumping signal sent by the intelligent pumping device.

[0022] The aforementioned intelligent pumping equipment can be any device capable of pumping oil according to instructions. For example, the aforementioned intelligent pumping equipment can be an intelligent hydraulic pumping unit. The aforementioned pumping signal can be a signal used to prompt the aforementioned execution entity to begin pumping oil. Each historical production data point in the aforementioned historical production data sequence can be a numerical value used to characterize the production of the oil well. Each historical production data point in the aforementioned historical production data sequence corresponds to a point in time. The aforementioned well parameter structured data can be a structured table used to record the parameters when pumping oil from the oil well. Each row in the aforementioned well parameter structured data can correspond to a point in time, and each column can correspond to a parameter of the oil well. The parameters corresponding to the oil well can include, but are not limited to: wellhead pressure, temperature, and whether acidizing operations have been performed. The aforementioned wellhead pressure can be the pressure measured at the wellhead. For example, the value corresponding to the aforementioned wellhead pressure can be 15.2, and the corresponding unit can be megapascals. The aforementioned temperature can characterize the temperature inside the oil well. For example, the value corresponding to the aforementioned temperature can be 44.1, and the corresponding unit can be degrees Celsius. The aforementioned whether acidizing operations have been performed can be a numerical value used to characterize whether the oil well has undergone acidizing treatment. For example, the value corresponding to whether acidizing was performed can be 1 or 0, where 1 indicates that well acidizing was performed and 0 indicates that well acidizing was not performed. The executing entity can be a server.

[0023] In practice, the aforementioned implementing entity can obtain various historical production data from a pre-set historical production database. This historical production database can be a pre-defined database set up by technical personnel to store the various historical production data. Each historical production data point corresponds to a specific time point. Then, these historical production data points can be arranged into a historical production data sequence according to their chronological order.

[0024] Next, the time points corresponding to the aforementioned historical production data sequence can be determined as historical time points. Then, the preset structured data acquisition information and the aforementioned historical time points can be sent to the target terminal. The structured data acquisition information can be information used to acquire structured well parameter data corresponding to each historical time point. The target terminal can be a terminal for technical personnel. Then, the data sent by the target terminal can be received as structured well parameter data.

[0025] Step 102: For each future time point in the preset future time point sequence, construct the corresponding future time point production prediction vector based on the historical production data sequence, oil well parameter structured data, and the future time point.

[0026] In some embodiments, the execution entity can construct a production prediction vector corresponding to each future time point in a preset future time point sequence, based on the historical production data sequence, the structured oil well parameter data, and the future time point itself. Each future time point in the future time point sequence can be a future time point where oil well production needs to be predicted. The number of future time points in the future time point sequence is the same as the number of time points corresponding to the historical production data sequence. The production prediction vector can be a vector representing various characteristics of the oil well at the corresponding future time point.

[0027] In addressing the aforementioned technical problems in the application scenario of offshore oilfield exploitation, the following technical issues often arise: when constructing feature vectors to predict well production at future points in time, the constructed feature vectors often lack quantification of professional knowledge in the oil well field. This leads to a disconnect between the generated production prediction vector and the actual well production, resulting in low accuracy of the predicted well production. Consequently, intelligent pumping equipment may forcibly pump oil when the well fluid supply is insufficient, easily accelerating equipment wear. Considering the following requirements for this application scenario: the harsh operating environment of offshore oilfields and extremely high equipment maintenance costs, equipment failure not only affects oilfield production but can also easily trigger cascading well shutdowns, leading to prolonged unplanned production stoppages. Therefore, it is urgent to reduce the wear rate of intelligent pumping equipment to decrease the probability of failure. We have decided to adopt the following solution: In some optional implementations of certain embodiments, the execution entity may construct a production prediction vector corresponding to the future time point based on the historical production data sequence, the structured oil well parameter data, and the future time point through the following steps: The first step is to generate time interval data based on the aforementioned future time point and the preset initial pumping time point. The initial pumping time point can be the time when the oil well is first pumped. The time interval data can be a numerical value representing the number of days between the future time point and the initial pumping time point. In practice, the executing entity can determine the number of days between the future time point and the initial pumping time point as the time interval data.

[0028] The second step is to determine the target decline index by multiplying the preset decline index by the aforementioned time interval data. This decline index can be a value pre-set by technicians to describe the change in oil well production over time. For example, the decline index could be -0.0024.

[0029] The third step involves generating the future formation pressure based on the aforementioned target decline index and the preset initial formation pressure. The initial formation pressure can be the pressure measured in the middle of the oil reservoir before the well is exploited. For example, the initial formation pressure could be 20, with the corresponding unit being megapascals (MPa). The future formation pressure can be the flowing pressure of the well at the aforementioned future time point.

[0030] In practice, firstly, the intermediate value of formation pressure can be determined by raising the target decline exponent to the power of the preset base. The preset base can be the natural constant e. For example, when the target decline exponent is 3, the intermediate value of formation pressure can be e to the power of 3. As an example, an approximate value of e, "2.71828", can be used for calculation. Secondly, the product of the initial formation pressure and the intermediate value of formation pressure can be determined as the future formation pressure.

[0031] The fourth step involves generating the future theoretical maximum production rate based on the preset theoretical initial production data, the aforementioned initial formation pressure, and the aforementioned future formation pressure. The aforementioned theoretical initial production data represents the maximum oil production that can theoretically be extracted per day before the well begins production. For example, the theoretical initial production data could be 50, corresponding to cubic meters per day. The aforementioned future theoretical maximum production rate represents the maximum oil production that the well can extract at the aforementioned future point in time.

[0032] In practice, firstly, the ratio of the aforementioned future formation pressure to the aforementioned initial formation pressure can be determined as the formation pressure ratio. Secondly, the product of the aforementioned theoretical initial production data and the aforementioned formation pressure ratio can be determined as the theoretical maximum future production.

[0033] The fifth step is to determine the ratio of the preset well flow pressure to the aforementioned future formation pressure as the theoretical instantaneous production data. The aforementioned well flow pressure can be a flow pressure set by technicians to maintain efficient well production. For example, the well flow pressure could be 5, and the corresponding unit could be megapascals (MPa).

[0034] Step 6: Based on the first preset coefficient, the second preset coefficient, the third preset coefficient, and the above-mentioned theoretical instantaneous production data, generate flow pressure production data.

[0035] The first preset coefficient can be a pre-set value. For example, the first preset coefficient can be 0.8. The second preset coefficient can be a pre-set value. For example, the second preset coefficient can be 0.2. The third preset coefficient can be a pre-set value. For example, the third preset coefficient can be 1. The flowing pressure and production data can be values ​​used to describe the relationship between the flowing pressure and production of an oil well.

[0036] In practice, firstly, the square of the aforementioned theoretical instantaneous production data can be determined as the theoretical square data. Secondly, the product of the aforementioned first preset coefficient and the aforementioned theoretical square data can be determined as the first pressure production data. Then, the product of the aforementioned theoretical instantaneous production data and the aforementioned second preset coefficient can be determined as the second pressure production data. Finally, the difference between the aforementioned third preset coefficient and the aforementioned first pressure production data and the aforementioned second pressure production data can be determined as the flowing pressure production data. For example, the flowing pressure production data can be obtained by subtracting the aforementioned first pressure production data from the aforementioned third preset coefficient, and then subtracting the aforementioned second pressure production data.

[0037] Step 7: The product of the above-mentioned theoretical maximum future output and the above-mentioned flow pressure output data is determined as the theoretical output data.

[0038] Step 8: Based on the preset time window, the aforementioned future time points, and the aforementioned historical production data sequence, determine the target historical production value and the non-zero production interval data. The aforementioned time window can be a pre-set time length by the technicians. For example, the aforementioned time window can be "1 year". The aforementioned target historical production value can be the historical production data corresponding to the target historical time point. The aforementioned target historical time point can be the time point obtained by subtracting the aforementioned time window from the aforementioned future time point. The aforementioned non-zero production interval data can be the number of days between the historical time point to be processed and the aforementioned target historical time point. The aforementioned historical time point to be processed can be the time point corresponding to the historical production data with the shortest interval to the aforementioned target historical time point and a value of "0" in the aforementioned historical production data sequence.

[0039] In practice, the difference between the aforementioned future time point and the aforementioned time window can be used to determine the target historical time point. For example, when the future time point is "2027-03-01" and the time window is "1 year", the target historical time point can be "2026-03-01". Secondly, the historical production data in the aforementioned historical production data sequence that corresponds to the aforementioned target historical time point can be used to determine the target historical production value.

[0040] Then, the time points corresponding to each historical production data point that is before the target historical time point and has a value of "0" in the historical production data sequence can be determined as the comparison time points.

[0041] Then, the historical time point to be processed can be determined as the one closest to the target historical time point among the above comparison time points.

[0042] Finally, the number of days between the aforementioned historical time points to be processed and the aforementioned target historical time points can be determined as the non-zero production interval data.

[0043] The ninth step involves analyzing the historical production data sequence based on the aforementioned time window and future time points to generate rolling quantitative indicator data corresponding to the future time points.

[0044] The aforementioned rolling quantitative indicator data can be the quantitative indicators corresponding to the historical output data of each target. Each target historical output data point can be the historical output data located before the point in time to be processed within the aforementioned historical output data sequence. The point in time to be processed can be the time point obtained by subtracting the aforementioned time window from the aforementioned future time point. The aforementioned rolling quantitative indicator data can include the average output, standard deviation of output, maximum output, minimum output, and number of zero values.

[0045] In practice, the time window can be subtracted from the aforementioned future time point to obtain a time point to be processed. This time point to be processed is the target historical time point. Next, the historical production data points in the historical production data sequence that are prior to the time point to be processed can be identified as the target historical production data points. Then, the average value of these target historical production data points can be determined as the production mean. The standard deviation of these target historical production data points can be determined as the production standard deviation. Then, the target historical production data point with the largest value among these target historical production data points can be determined as the maximum production value. Then, the target historical production data point with the smallest value among these target historical production data points can be determined as the minimum production value. Then, the number of target historical production data points with a value of 0 among these target historical production data points can be determined as the number of zero values. Finally, the production mean, production standard deviation, maximum production value, minimum production value, and number of zero values ​​can be combined to form rolling quantitative indicator data.

[0046] Step 10: Based on the structured oil well parameter data and the future time points mentioned above, generate quantitative index data for the oil well parameters. Specifically, the quantitative index data for the oil well parameters can be a vector composed of a row of data corresponding to the specified time point in the structured oil well parameter data.

[0047] In practice, firstly, the difference between the aforementioned future time point and the aforementioned time window can be determined as the time point to be processed. Secondly, the corresponding parameters in the structured data of the aforementioned oil well parameters at the aforementioned time point to be processed can be determined as individual parameter data. Each parameter data corresponds to a column name. These column names can include, but are not limited to, wellhead pressure, temperature, and whether acidizing operations are being performed. Finally, the parameter data corresponding to each column name can be combined into a vector according to a preset column name order as quantitative index data for oil well parameters. The column name order can be the order in which the column names corresponding to the parameter data are arranged. For example, the column name order could be {wellhead pressure, temperature, whether acidizing operations are being performed}.

[0048] Step 11: Based on the above theoretical production data, the above target historical production value, the above non-zero production interval data, the above rolling quantitative index data, and the above oil well parameter quantitative index data, construct the production prediction vector corresponding to the above future time points.

[0049] In practice, the theoretical production data, the target historical production value, the non-zero production interval data, the rolling quantitative index data, and the oil well parameter quantitative index data can be combined into a vector as a production prediction vector according to a preset arrangement order. The preset arrangement order can be a sequence pre-set by technical personnel. For example, the preset arrangement order could be {theoretical production data, target historical production value, non-zero production interval data, production mean, production standard deviation, maximum production value, minimum production value, number of zero values, oil well parameter quantitative index data}.

[0050] The above technical solution and its related content, combined with step 108, serve as an inventive point of this disclosure, solving the problem of "aggravated equipment wear". Factors leading to increased equipment wear often include: when constructing feature vectors to predict oil well production at future time points, the constructed feature vectors often lack quantification of professional knowledge in the oil well field, resulting in a disconnect between the generated production prediction vector and the actual oil well production. This leads to low accuracy in predicting oil well production based on the production prediction vector, and consequently, situations where intelligent pumping equipment forcibly pumps oil when the oil well's fluid supply is insufficient, easily exacerbating equipment wear. Solving these factors can reduce the rate of equipment wear. To achieve this effect, this disclosure first generates time interval data based on the aforementioned future time point and a preset initial pumping time point. This yields the time interval between the future time point and the initial pumping time point. Second, the product of a preset decline index and the aforementioned time interval data is determined as the target decline index. Then, based on the target decline index and a preset initial formation pressure, the future formation pressure is generated. This yields the formation pressure corresponding to the future time point. Then, based on the preset theoretical initial production data, the aforementioned initial formation pressure, and the aforementioned future formation pressure, the future theoretical maximum production is generated. Thus, the remaining production of the oil well at a future time point can be generated according to the time interval. Then, the ratio of the preset oil well flowing pressure to the aforementioned future formation pressure is determined as the theoretical instantaneous production data. Thus, the theoretical instantaneous production data can be obtained. Then, based on the first preset coefficient, the second preset coefficient, the third preset coefficient, and the aforementioned theoretical instantaneous production data, flowing pressure production data is generated. Thus, the flowing pressure of the oil well can be obtained. Then, the product of the aforementioned future theoretical maximum production and the aforementioned flowing pressure production data is determined as the theoretical production data. Thus, the oil production that the oil well can extract at a future time point can be obtained. Then, based on the preset time window, the aforementioned future time point, and the aforementioned historical production data sequence, the target historical production value and non-zero production interval data are determined. Thus, the target historical production value and non-zero production interval data can be obtained. Then, based on the aforementioned time window and the aforementioned future time point, the aforementioned historical production data sequence is processed through data parsing to generate rolling quantitative index data corresponding to the aforementioned future time point. Thus, the rolling quantitative index data can be obtained. Then, based on the structured data of the oil well parameters and the aforementioned future time points, quantitative index data of the oil well parameters is generated. Thus, the quantitative index data of the oil well parameters can be obtained. Finally, based on the aforementioned theoretical production data, the aforementioned target historical production values, the aforementioned non-zero production interval data, the aforementioned rolling quantitative index data, and the aforementioned quantitative index data of the oil well parameters, a production prediction vector corresponding to the aforementioned future time points is constructed. Thus, the production prediction vector can be obtained.Because it allows for the quantification of professional knowledge in the oil well field through various formulas, and the generation of theoretical oil production that a well can extract at a future time point through these formulas, this calculated oil production is then incorporated into the production prediction vector. Based on the production prediction vector, the final production output of the oil well is predicted, and finally, the intelligent pumping equipment is controlled to pump oil based on the prediction results. This reduces the probability of discrepancies between the production prediction vector and the actual production output of the oil well, improves the accuracy of the prediction results, reduces the number of times the intelligent pumping equipment forcibly pumps oil when the oil well's fluid supply is insufficient, and slows down the wear and tear on the intelligent pumping equipment.

[0051] Step 103: Construct a production forecast matrix based on the constructed production forecast vectors.

[0052] In some embodiments, the execution entity may construct a production forecast matrix based on the constructed production forecast vectors. The production forecast matrix may be a matrix composed of the constructed production forecast vectors.

[0053] In practice, the aforementioned implementing entity can arrange the production forecast vectors corresponding to the aforementioned future time point sequence into a matrix as a production forecast matrix, according to the chronological order of the various future time points in the sequence.

[0054] For example, when the future time sequence is {"2027-07-16", "2027-07-17"}, the production forecast vector corresponding to "2027-07-16" can be used as the first row of the matrix, and the production forecast vector corresponding to "2027-07-17" can be used as the second row of the matrix, thereby generating a matrix as the production forecast matrix.

[0055] Step 104: Perform production probability prediction processing on the production prediction matrix and historical output data sequence to obtain the production probability data sequence.

[0056] In some embodiments, the execution entity may perform production probability prediction processing on the production prediction matrix and the historical production data sequence to obtain a production probability data sequence. Each production probability data point in the production probability data sequence can be a numerical value representing the probability that an oil well will produce oil at a corresponding future time point. Each production probability data point in the production probability data sequence corresponds to a future time point.

[0057] In some optional implementations of certain embodiments, the execution entity may perform production probability prediction processing on the production prediction matrix and the historical output data sequence through the following steps to obtain a production probability data sequence: The first step is to binarize the aforementioned historical production data sequence to obtain a binarized production data sequence. This binarized production data sequence can be a historical production data sequence that has already undergone binarization.

[0058] In practice, for each historical production data point in the aforementioned historical production data sequence, in response to determining that the historical production data point is greater than a preset production threshold, "1" can be designated as the binarized production data corresponding to the historical production data point. The production threshold can be a pre-set value. Here, the specific setting of the production threshold is not limited. Then, in response to determining that the historical production data point is less than or equal to the production threshold, "0" can be designated as the binarized production data corresponding to the historical production data point.

[0059] The second step involves inputting the aforementioned production forecast matrix and the aforementioned binarized output data sequence into a pre-trained production probability prediction model to obtain a production probability data sequence. Each production probability data point in the production probability data sequence corresponds to a future time point. The aforementioned production probability prediction model can be a neural network model that takes the production forecast matrix and the binarized output data sequence as input and the production probability data sequence as output.

[0060] In some optional implementations of certain embodiments, the above-described production probability prediction model can be trained through the following steps: The first step is to obtain a sample set. Each sample in the sample set includes a sample production prediction matrix, a sample binarized output data sequence, and a sample production probability data sequence. The sample production prediction matrix can be the production prediction matrix used to train the production probability prediction model. The sample binarized output data sequence can be the binarized output data sequence used to train the production probability prediction model. The sample production probability data sequence can be the production probability data sequence used to train the production probability prediction model.

[0061] The second step, based on the sample set, is to perform the following steps: The first sub-step involves inputting the production prediction matrix and the binary output data sequence of each sample in at least one sample in the sample set into the feature processing layer of the initial neural network to obtain the historical time-series feature matrix corresponding to each sample. This historical time-series feature matrix can be a matrix obtained after performing linear transformation, concatenation, and linear rectification on the production prediction matrix and the binary output data sequence.

[0062] The aforementioned feature processing layer can be a neural network layer that takes the sample production prediction matrix and the sample binarized production data sequence as input and the historical time-series feature matrix as output. This feature processing layer may include an embedding layer, a projection layer, a concatenation function, a mapping layer, and a linear rectified function.

[0063] The aforementioned embedding layer can be a neural network layer that takes a sequence of binarized sample output data as input and an output density matrix as output. For example, the embedding layer can be nn.Embedding. The aforementioned density matrix can be a matrix composed of various density vectors. The number of rows in the density matrix can be the length of the sequence of binarized sample output data, and the number of columns can be a preset length. Each density vector can be a sample binarized output data that is mapped to a feature vector of a preset length. The preset length can be a pre-defined value. The specific setting of the preset length is not limited here.

[0064] The aforementioned projection layer can be a fully connected layer that takes the sample production prediction matrix as input and the projected prediction matrix as output. The aforementioned projected prediction matrix can be a sample production prediction matrix of a preset length whose corresponding number of columns are mapped to the aforementioned number of columns.

[0065] The concatenation function described above can be a function that takes the dense production matrix and the projected prediction matrix as input and outputs the concatenated prediction matrix. For example, the concatenation function could be the `concat()` function. The concatenated prediction matrix described above can be the matrix obtained by concatenating the dense production matrix and the projected prediction matrix. For example, when both the dense production matrix and the projected prediction matrix have shapes of (m, n), the shape of the concatenated prediction matrix can be (m, 2n).

[0066] The aforementioned mapping layer can be a linear layer that takes a concatenated prediction matrix as input and a mapped prediction matrix as output. The aforementioned mapped prediction matrix can be a concatenated prediction matrix whose corresponding column number is mapped to a preset mapping value. The number of rows in the aforementioned mapped prediction matrix can be the length of the aforementioned binary output data sequence, and the corresponding number of columns can be the preset mapping value. The aforementioned preset mapping value can be a pre-defined numerical value. Here, the specific setting of the aforementioned preset mapping value is not limited.

[0067] The aforementioned linear rectifier function can be a function that takes a mapping prediction matrix as input and a historical time series feature matrix as output. The aforementioned historical time series feature matrix can be the matrix obtained by inputting the mapping prediction matrix into the linear rectifier function.

[0068] The second sub-step involves inputting the historical temporal feature matrix corresponding to each of the at least one sample into the feature encoding layer of the initial neural network to obtain the historical context encoding matrix corresponding to each sample. The historical context encoding matrix can be a historical temporal feature matrix that has undergone deep feature extraction. The shape of the historical context encoding matrix is ​​the same as that of the historical temporal feature matrix. The feature encoding layer can be a neural network that takes the historical temporal feature matrix as input and the historical context encoding matrix as output. The feature encoding layer can be a Mamba network.

[0069] The third sub-step involves inputting the historical context encoding matrix and the sample production probability data sequence corresponding to each of the at least one sample into the autoregressive decoding layer of the initial neural network to obtain the decoding state matrix corresponding to each sample. The decoding state matrix can be the matrix obtained after decoding the historical context encoding matrix and the sample production probability data sequence.

[0070] The aforementioned autoregressive decoding layer can be a neural network layer that takes a historical context encoding matrix as input and a decoded state matrix as output. This autoregressive decoding layer may include a pooling layer, a pre-decoding algorithm, and a decoder.

[0071] The aforementioned pooling layer can be an attention pooling layer that takes the historical context encoding matrix as input and the global encoding vector as output. The aforementioned global encoding vector can be a vector with a length equal to the aforementioned preset mapping value, obtained after attention pooling of the aforementioned historical context encoding matrix.

[0072] The above-described decoding pre-processing algorithm can be an algorithm that takes the global encoding vector and the sample production probability data sequence as input, and the decoder input matrix as output. The processing steps of the above-described decoding pre-processing algorithm are as follows: First, for each sample production probability data point in the aforementioned sample production probability data sequence, the sample production probability data can be input into a feature embedding network to obtain a vector of length equal to the preset mapping value, which serves as the sample production probability vector. The feature embedding network can be `nn.Embedding`. Second, the sample production probability vector and the global encoding vector can be input into a feature concatenation function to obtain the output vector, which serves as the encoded concatenation vector. The feature concatenation function can be the `concat()` function. Then, the encoded concatenation vector can be input into a linear layer to obtain an encoded concatenation vector mapped to the preset mapping value, which serves as the input vector.

[0073] Finally, the obtained input vectors can be stacked into a matrix using a stacking function, which serves as the decoder input matrix. For example, the first input vector can be placed in the first row of the matrix, the second input vector in the second row, and so on, resulting in a matrix that can be used as the decoder input matrix. The stacking function can be any function capable of stacking vectors into a matrix. For example, the stacking function could be `np.vstack()`.

[0074] The decoder described above can be a neural network that takes the decoder input matrix as input and the decoder state matrix as output. The decoder can be a Mamba network. The decoder state matrix can be the decoder input matrix after processing by the Mamba network. The decoder state matrix has the same shape as the decoder input matrix.

[0075] The fourth sub-step involves inputting the decoding state matrix corresponding to each of the at least one of the above samples into the probability output layer to obtain the production probability data sequence.

[0076] The aforementioned probability output layer can be a neural network layer that takes a decoded state matrix as input and produces a sequence of probability data as output. This probability output layer can include a linear output layer and a sigmoid function.

[0077] The aforementioned linear output layer can be a linear layer that takes the decoded state matrix as input and the decoded output vector as output. The decoded output vector can be a vector obtained by mapping each row of the decoded state matrix to a numerical value. The length of the decoded output vector is the same as the length of the aforementioned sample binarized output data sequence.

[0078] The sigmoid function described above can be a function that takes the decoded output vector as input and the generated probability data sequence as output. The sigmoid function can map each value in the decoded output vector to a value between 0 and 1.

[0079] The fifth sub-step involves, for each of the at least one sample mentioned above, performing a data comparison between the production probability data sequence corresponding to that sample and the production probability data sequences of the samples included in the sample, to obtain a data comparison result. The data comparison result can be the loss value corresponding to the production probability data sequence.

[0080] The sixth sub-step involves determining the initial neural network as a production probability prediction model in response to the determination that at least one data comparison result corresponding to the at least one sample satisfies a preset data comparison condition. The data comparison condition can be that the number of each target data comparison result is greater than a preset threshold. The preset threshold can be a pre-set value. Here, the specific setting of the preset threshold is not limited. Each target data comparison result can be a data comparison result greater than the preset comparison threshold. The preset comparison threshold can be a pre-set value. Here, the specific setting of the preset comparison threshold is not limited.

[0081] In some optional implementations of certain embodiments, the execution entity may perform data comparison between the production probability data sequence corresponding to the sample and the sample production probability data sequence included in the sample through the following steps to obtain the data comparison result: The first step is to perform the following steps for each future time point corresponding to the above production probability data sequence: The first sub-step is to determine the production probability data corresponding to the aforementioned future time point in the above production probability data sequence as the target production probability data.

[0082] The second sub-step is to determine the sample production probability data corresponding to the aforementioned future time points in the above sample production probability data sequence as the target sample production probability data.

[0083] The third sub-step involves determining the logarithm of the aforementioned target production probability data as logarithmic probability data. This logarithm can be a base-2 logarithm. For example, when the target production probability data is 0.5, the logarithmic probability data can be the base-2 logarithm of 0.5.

[0084] The fourth sub-step involves determining the product of the target sample production probability data and the logarithmic probability data as the first data item.

[0085] The fifth sub-step involves determining the difference between the preset value and the aforementioned target production probability data as the target difference value. The preset value can be a pre-defined value. For example, the preset value can be 1.

[0086] The sixth sub-step is to determine the difference between the preset value and the target sample production probability data as the target sample difference data.

[0087] The seventh sub-step involves determining the logarithm of the target difference data as the logarithmic difference data. This logarithm can be a base-2 logarithm. For example, when the target difference data is 3, the logarithmic difference data can be a base-2 logarithm of 3.

[0088] The eighth sub-step is to determine the product of the above target sample difference data and the above difference logarithmic data as the second numerical data.

[0089] The ninth sub-step involves determining the target number by summing the first and second data items mentioned above.

[0090] The second step is to determine the average number of items as the average number of items.

[0091] The third step is to determine the negative values ​​of the above average number of items as the data comparison results.

[0092] Step 105: Perform data filtering on the historical production data sequence to obtain the target production data sequence.

[0093] In some embodiments, the aforementioned executing entity may perform data filtering on the aforementioned historical production data sequence to obtain the target production data sequence.

[0094] The target output data sequence mentioned above can be a filtered historical output data sequence.

[0095] In practice, for each historical production data point in the aforementioned historical production data sequence, in response to determining that the historical production data is not zero, it can be designated as the target production data. Then, the designated target production data can be arranged into a target production data sequence according to the chronological order of the time points corresponding to each designated target production data point.

[0096] Step 106: Perform production prediction processing on the production prediction matrix and the above target production data sequence to obtain the oil well production data sequence.

[0097] In some embodiments, the executing entity may perform production prediction processing on the production prediction matrix and the target production data sequence to obtain an oil well production data sequence. Each oil well production data point in the oil well production data sequence may be a numerical value representing the production of an oil well, predicted for a future time point.

[0098] In some optional implementations of certain embodiments, the execution entity may perform production prediction processing on the production prediction matrix and the target production data sequence through the following steps to obtain the oil well production data sequence: The first step involves performing vector filtering on the production forecast matrix based on the target output data sequence and the preset time window, resulting in a non-zero output matrix. The time window in question is the same as the aforementioned time window. The non-zero output matrix can be the matrix obtained after filtering the production forecast matrix.

[0099] In practice, firstly, for each target output data point in the aforementioned target output data sequence, the sum of the corresponding time point and the aforementioned time window can be determined as the target time point. For example, when the time point corresponding to the target output data is "2026-03-23" and the time window is "1 year", the corresponding target time point is "2027-03-23". Then, for each row vector in the aforementioned production forecast matrix, the row vector whose corresponding future time point is the same as the aforementioned target time point can be determined as the non-zero output vector. Finally, the determined non-zero output vectors can be combined into a matrix according to the chronological order of the corresponding future time points to form the non-zero output matrix.

[0100] For example, when there are two non-zero output vectors, and the future time point corresponding to the first non-zero output vector is "2027-3-24" and the future time point corresponding to the second non-zero output vector is "2027-3-23", according to the order of the two future time points, the second non-zero output vector can be used as the first row vector of the matrix, and the first non-zero output vector can be used as the second row vector of the matrix, so as to construct a matrix as the non-zero output matrix.

[0101] The second step involves performing time-series feature extraction on the target output data sequence to obtain a time-series pattern feature matrix. This time-series pattern feature matrix can be the matrix obtained after extracting time features from the target output data sequence. The number of rows in the time-series pattern feature matrix is ​​the same as the number of rows in the non-zero output matrix.

[0102] In practice, the target output data sequence can be input into the time-series feature extraction network of a pre-trained output prediction model to obtain a time-series pattern feature matrix.

[0103] The above-mentioned production prediction model can be a neural network model that takes the target production data sequence and non-zero production matrix as input and the future production prediction data sequence as output.

[0104] The above-mentioned production forecasting model may include five layers.

[0105] The first layer can be a temporal feature extraction network. This temporal feature extraction network can be a neural network that takes the target output data sequence as input and outputs a temporal pattern feature matrix. For example, the temporal feature extraction network can be a long short-term memory network.

[0106] The second layer can be a matrix concatenation layer. This matrix concatenation layer can be a function that takes a non-zero output matrix and a temporal pattern feature matrix as input and outputs a concatenated feature matrix. For example, the matrix concatenation layer could be the `numpy.concatenate()` function.

[0107] The third layer can be a feature hiding layer. This feature hiding layer can include a linear transformation layer and a mapping function. The linear transformation layer can be a fully connected layer that takes the concatenated feature matrix as input and the linear transformation matrix as output. The linear transformation matrix can be a concatenated feature matrix whose corresponding number of columns are mapped to a first preset number of columns. The number of rows in the linear transformation matrix is ​​the same as the number of rows in the concatenated feature matrix. The first preset number of columns can be a pre-defined value. Here, the specific setting of the first preset number of columns is not limited.

[0108] The mapping function described above can be a linear rectifier function that takes a linear transformation matrix as input and a nonlinear characteristic matrix as output. The nonlinear characteristic matrix can be the matrix obtained by inputting the linear transformation matrix into the linear rectifier function.

[0109] The fourth layer can be a matrix hidden layer. This matrix hidden layer can include a dimension mapping layer and a non-linear activation function.

[0110] The aforementioned dimension mapping layer can be a fully connected layer that takes a nonlinear feature matrix as input and a linear compression matrix as output. The aforementioned linear compression matrix can be a nonlinear feature matrix whose column count is compressed to a second preset number of columns after a linear transformation. The aforementioned second preset number of columns can be a pre-defined value. Here, the specific setting of the aforementioned second preset number of columns is not limited.

[0111] The aforementioned nonlinear activation function can be a linear rectified function that takes a linear compression matrix as input and a characteristic compression matrix as output. The aforementioned characteristic compression matrix can be the matrix obtained by inputting the aforementioned linear compression matrix into the linear rectified function.

[0112] The fifth layer can be the output layer. This output layer can be a fully connected layer that takes a feature compression matrix as input and outputs a sequence of future output prediction data. Each future output prediction data point in the sequence can be a value obtained by linearly transforming the corresponding row vector in the feature compression matrix. The length of the future output prediction data sequence can be the same as the number of rows in the feature compression matrix. Each future output prediction data point in the sequence corresponds to a future time point.

[0113] The third step involves concatenating the non-zero output matrix and the time-series pattern feature matrix to obtain a concatenated feature matrix. In practice, the non-zero output matrix and the time-series pattern feature matrix can be input into the matrix concatenation layer of the output prediction model to obtain the concatenated feature matrix.

[0114] The fourth step involves performing nonlinear feature mapping on the concatenated feature matrix to obtain a nonlinear feature matrix. In practice, this concatenated feature matrix can be input into the hidden feature layer of the yield prediction model to obtain the nonlinear feature matrix.

[0115] The fifth step involves performing a dimension mapping process on the aforementioned nonlinear feature matrix to obtain a feature compression matrix. In practice, this nonlinear feature matrix can be input into the matrix hidden layer of the aforementioned yield prediction model to obtain the feature compression matrix.

[0116] The sixth step involves performing regression prediction processing on the aforementioned feature compression matrix to obtain the future output prediction data sequence. In practice, the aforementioned feature compression matrix can be input into the output layer of the aforementioned output prediction model to obtain the future output prediction data sequence.

[0117] Step 7: Based on the aforementioned future production prediction data sequence, generate an oil well production data sequence. In practice, for each future time point in the aforementioned future time point sequence, firstly, this future time point can be determined as the target future time point. Then, in response to the determination that each future time point corresponding to the aforementioned future production prediction data sequence is different from the aforementioned target future time point, a preset production value can be determined as the oil well production data. The preset production value can be 0. Then, in response to the determination that among the various future time points corresponding to the aforementioned future production prediction data sequence, there is one future time point that is the same as the aforementioned target future time point, the future production prediction data for the corresponding future time point that is the same as the aforementioned target future time point can be determined as the oil well production data.

[0118] Finally, the determined oil well production data can be arranged into an oil well production data sequence according to the chronological order of the corresponding future time points.

[0119] Step 107: Generate an oil well production prediction data sequence based on the production probability data sequence and the oil well production data sequence.

[0120] In some embodiments, the executing entity may generate an oil well production prediction data sequence based on the production probability data sequence and the oil well production data sequence. Each oil well production prediction data point in the oil well production prediction data sequence may be a numerical value representing the production of the oil well at a corresponding future point in time.

[0121] In some optional implementations of certain embodiments, the aforementioned execution entity may generate an oil well production prediction data sequence based on the aforementioned production probability data sequence and the aforementioned oil well production data sequence through the following steps: The first step is to perform the following steps for each future time point corresponding to the above production probability data sequence: The first sub-step involves identifying the production probability data corresponding to the aforementioned future time points in the above production probability data sequence as the production probability data to be processed.

[0122] The second sub-step involves identifying the oil well production data corresponding to the aforementioned future time points in the above oil well production data sequence as the oil well production data to be processed.

[0123] The third sub-step involves determining the product of the aforementioned production probability data to be processed and the aforementioned oil well production data to be processed as the oil well production prediction data.

[0124] The second step is to arrange the determined production forecast data of each oil well into a sequence of oil well production forecast data. In practice, the determined production forecast data of each oil well can be arranged into a sequence according to the chronological order of the corresponding future time points.

[0125] Step 108: Based on the oil well production prediction data sequence, control the intelligent pumping equipment to perform intelligent pumping tasks.

[0126] In some embodiments, the aforementioned execution entity may control the aforementioned intelligent pumping equipment to perform intelligent pumping tasks based on the aforementioned oil well production prediction data sequence.

[0127] In addressing the aforementioned technical problems in the application scenario of pumping oil from low-yield wells, the following technical issues often arise: when controlling the intelligent pumping equipment based on the predicted well production, excessively high pumping frequency can lead to dry running or idle consumption of the equipment, exacerbating hardware wear. Considering the following requirements for this application scenario: low-yield wells have low economic returns, and dry running or idle consumption of the intelligent pumping equipment necessitates frequent maintenance, further compressing the economic returns and ultimately rendering the well unprofitable. Therefore, it is necessary to reduce the frequency of dry running or idle consumption and decrease the rate of hardware wear. We have decided to adopt the following solution: In some optional implementations of certain embodiments, the aforementioned execution entity can control the aforementioned intelligent pumping equipment to perform intelligent pumping tasks based on the aforementioned oil well production prediction data sequence through the following steps: The first step involves dividing the aforementioned oil well production prediction data sequence into multiple predicted production data sequences based on a preset partitioning window. Each predicted production data sequence represents the oil well production prediction data. Each predicted production data sequence can be a subsequence of the aforementioned oil well production prediction data sequence. The partitioning window can be a window used to partition the aforementioned oil well production prediction data sequence. For example, the partitioning window can be 3.

[0128] As an example, when the above division window is 3, every three adjacent oil well production prediction data that have not been divided in the above oil well production prediction data sequence can be divided into a sequence as a prediction production data sequence, thereby obtaining each prediction production data sequence.

[0129] The first step, for each of the above predicted output data sequences, is to perform the following steps: The first sub-step involves performing the following iterative steps for each predicted output data point in the aforementioned predicted output data sequence, based on the initial stroke data sequence: Sub-step one involves determining the initial stroke data corresponding to the predicted production data in the initial stroke data sequence as the stroke data to be processed. Each initial stroke data in the initial stroke data sequence can be a value representing the number of reciprocating motions completed by the pump of the intelligent oil pumping equipment per unit time. The length of the initial stroke data sequence is the same as the length of each predicted production data sequence in the aforementioned predicted production data sequences.

[0130] As an example, when the predicted output data is ranked first in the above predicted output data sequence, the stroke data to be processed corresponding to the predicted output data is the initial stroke data ranked first in the initial stroke data sequence.

[0131] Sub-step two involves determining the stroke coefficient by multiplying the preset empirical coefficient by the aforementioned stroke data to be processed. The empirical coefficient can be a pre-set value. Here, the specific setting of the empirical coefficient is not limited. For example, the empirical coefficient could be 0.4.

[0132] Sub-step three: The difference between the above stroke coefficient and the above predicted output data is determined as the output coefficient difference value.

[0133] Sub-step four: The square of the above-mentioned production coefficient difference is determined as the production deviation data.

[0134] Sub-step five: The square of the above-mentioned stroke data to be processed is determined as the stroke square data.

[0135] Sub-step six: Based on a preset first power coefficient, a preset second power coefficient, and the aforementioned impulse square data, power consumption data is generated. The first power coefficient can be a pre-set value. Here, the specific setting of the first power coefficient is not limited. For example, the first power coefficient can be 2. The second power coefficient can be a pre-set value. Here, the specific setting of the second power coefficient is not limited. For example, the second power coefficient can be 10. The power consumption data can be a value used to characterize the power consumed by the intelligent oil extraction device.

[0136] In practice, firstly, the product of the first power coefficient and the squared impulse data can be used to determine the intermediate power value. Secondly, the sum of the intermediate power value and the second power coefficient can be used to determine the power consumption data.

[0137] The second sub-step involves summing the generated individual production deviation data to determine the total production deviation data.

[0138] The third sub-step involves summing the generated power consumption data into a total power consumption data.

[0139] The fourth sub-step involves determining the weighted power consumption data by multiplying the sum of the power consumption data above with a preset weighting coefficient. The weighting coefficient can be a pre-set value. Here, the specific setting of the weighting coefficient is not limited. For example, the weighting coefficient can be 0.1.

[0140] The fifth sub-step involves determining the target reward value by summing the above-mentioned total production deviation data and the above-mentioned power consumption weighted data.

[0141] The sixth sub-step, in response to determining that the number of executions of the above iterative steps is less than the target number of iterations, optimizes the initial impulse data sequence based on the target reward value to obtain an optimized initial impulse data sequence as the initial impulse data sequence, updates the initial impulse data sequence, and uses the updated initial impulse data sequence to execute the above iterative steps again. The target number of iterations can be a pre-set value. For example, the target number of iterations can be 64.

[0142] In practice, for each initial stroke in the initial stroke data sequence, the sum of the initial stroke and a preset step size can be used to determine the initial stroke data, thereby updating the initial stroke data and the initial stroke data sequence. The preset step size can be a pre-defined value. For example, the preset step size can be 1.

[0143] The seventh sub-step, in response to determining that the number of executions of the above iterative steps is greater than or equal to the target number of iterations, controls the above intelligent oil pumping equipment to perform the above oil pumping task based on the determined target reward values ​​and the future time points corresponding to the above predicted production data sequence.

[0144] In practice, the target reward value with the largest value among the determined target reward values ​​can be identified as the reward value to be processed. Then, the initial stroke data sequence corresponding to the above reward value to be processed can be identified as the target initial stroke data sequence.

[0145] Then, for each target initial stroke data in the above target initial stroke data sequence, firstly, the predicted output data corresponding to the above target initial stroke data in the above predicted output data sequence can be determined as the target predicted output data.

[0146] Then, in response to determining that the target predicted yield data is greater than a preset prediction threshold, a preset start signal, the target initial stroke data, and the future time point corresponding to the target predicted yield data can be sent to the intelligent oil pumping device to control the intelligent oil pumping device to start oil pumping and to control the intelligent oil pumping device to execute the oil pumping task. The prediction threshold can be a preset value. Here, the specific setting of the prediction threshold is not limited. The start signal can be a signal used to prompt the intelligent oil pumping device to start oil pumping. For example, the start signal can be "start:on".

[0147] The above-described technical solution and its related content, as an inventive point of this disclosure, solve the problem of "accelerated equipment wear". Factors leading to accelerated equipment wear often include: when controlling intelligent pumping equipment to pump oil according to the predicted well production, excessively high stroke rates can easily cause the intelligent pumping equipment to dry-run or waste energy, easily accelerating hardware wear. Solving these factors can reduce the rate of equipment wear. To achieve this effect, this disclosure firstly, based on a preset partitioning window, performs sequence partitioning processing on the above-described oil well production prediction data sequence to obtain various predicted production data sequences. Thus, the oil well production prediction data sequence can be divided into various sub-sequences. Secondly, for each predicted production data sequence in the above-described predicted production data sequences, the following steps are performed: Firstly, based on the initial stroke data sequence, for each predicted production data in the above-described predicted production data sequence, the following iterative steps are performed: Secondly, the initial stroke data corresponding to the predicted production data in the initial stroke data sequence is determined as the stroke data to be processed. Then, the product of a preset empirical coefficient and the stroke data to be processed is determined as the stroke coefficient. Thus, the stroke coefficient can be obtained. Then, the difference between the aforementioned stroke coefficient and the aforementioned predicted output data is determined as the output coefficient difference. Thus, the output coefficient difference is obtained. Next, the square of the aforementioned output coefficient difference is determined as the output deviation data. Thus, the output deviation data is obtained. Then, the square of the aforementioned stroke data to be processed is determined as the stroke square data. Thus, the stroke square data is obtained. Then, based on the preset first power coefficient, the preset second power coefficient, and the aforementioned stroke square data, power consumption data is generated. Thus, when the stroke of the intelligent pumping equipment is the stroke data to be processed, the power consumed by the intelligent pumping equipment can be obtained. Then, the sum of the generated output deviation data is determined as the total output deviation data. Thus, the total output deviation data is obtained. Then, the sum of the generated power consumption data is determined as the total power consumption data. Thus, when the stroke of the intelligent pumping equipment is the initial stroke data sequence, the power consumed by the intelligent pumping equipment can be obtained. Then, the product of the aforementioned total power consumption data and the preset weighting coefficient is determined as the power consumption weighted data. Thus, the power consumption weighted data is obtained. Next, the sum of the above-mentioned production deviation data and the above-mentioned power consumption weighted data is determined as the target reward value. Thus, the reward value for setting the initial stroke data sequence for the above-mentioned predicted production data sequence can be obtained. Then, in response to determining that the number of executions of the above-mentioned iterative steps is less than the target number of iterations, the initial stroke data sequence is optimized based on the above-mentioned target reward value to obtain the optimized initial stroke data sequence as the initial stroke data sequence, which is then used to update the initial stroke data sequence, and the above-mentioned iterative steps are executed again using the updated initial stroke data sequence.Therefore, the initial stroke data sequence can be updated to gradually obtain the initial stroke data sequence with the highest target reward value. Finally, in response to determining that the number of executions of the above iterative steps is greater than or equal to the target number of iterations, the intelligent pumping equipment is controlled to perform the above pumping task based on the determined target reward values ​​and the future time points corresponding to the predicted production data sequence. Thus, the pumping parameters of the intelligent pumping equipment can be determined using the initial stroke data sequence with the highest target reward value, preventing the intelligent pumping equipment from running dry or idling due to insufficient fluid supply to the oil well. Furthermore, because the strokes best matching the oil well production can be gradually generated based on the predicted oil well production, and the parameters of the intelligent pumping equipment are set according to the generated strokes to control the pumping operation, the probability of the intelligent pumping equipment running dry or idling due to excessively high strokes can be reduced, thus slowing down the hardware wear rate of the intelligent pumping equipment.

[0148] The above embodiments of this disclosure have the following beneficial effects: The intelligent oil pumping method based on well production prediction in some embodiments of this disclosure can reduce the waste of power resources and reduce ineffective wear and tear on equipment. Specifically, the reason for ineffective power consumption and equipment wear is that relying solely on fixed time intervals to control the oil pumping equipment can easily lead to idling, resulting in ineffective power consumption and equipment wear. Based on this, the intelligent oil pumping method based on well production prediction in some embodiments of this disclosure firstly acquires historical production data sequences and structured well parameter data in response to receiving a pumping signal from the intelligent oil pumping equipment. This yields initial data. Secondly, for each future time point in a preset future time point sequence, a production prediction vector corresponding to the future time point is constructed based on the aforementioned historical production data sequence, the aforementioned structured well parameter data, and the aforementioned future time point. This yields a production prediction vector. Then, a production prediction matrix is ​​constructed based on the constructed production prediction vectors. This yields a production prediction matrix. Finally, production probability prediction processing is performed on the aforementioned production prediction matrix and the aforementioned historical production data sequence to obtain a production probability data sequence. Therefore, the production probability of the oil well at various future time points can be predicted. Then, the historical production data sequence is filtered to obtain the target production data sequence. This removes zero values ​​from the historical production data sequence. Next, the production prediction matrix and the target production data sequence are processed to obtain the oil well production data sequence. This yields the oil well's production value. Then, based on the production probability data sequence and the oil well production data sequence, an oil well production prediction data sequence is generated. This provides the oil well's production at various future time points. Finally, based on the oil well production prediction data sequence, the intelligent pumping equipment is controlled to perform intelligent pumping tasks. This allows determination of whether the intelligent pumping equipment needs to be activated for oil well extraction at future time points, and the pumping frequency of the intelligent pumping equipment during extraction, based on the oil well's production at each future time point. Because the production of oil wells at various future time points can be predicted in advance, and then the intelligent pumping equipment can be controlled to pump oil based on the production of oil wells at various future time points, instead of relying solely on fixed time intervals for pumping, the probability of the intelligent pumping equipment running idle due to low oil well production can be reduced. This can reduce the waste of power resources and ineffective equipment wear caused by the idling of intelligent pumping equipment.

[0149] Further reference Figure 2 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of an intelligent pumping device based on oil well production prediction. These device embodiments are similar to... Figure 1 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0150] like Figure 2 As shown, an intelligent pumping device 200 based on well production prediction in some embodiments includes: an acquisition unit 201, a first construction unit 202, a second construction unit 203, a production probability prediction unit 204, a data filtering unit 205, a production prediction unit 206, a generation unit 207, and a control unit 208. The acquisition unit 201 is configured to acquire historical production data sequences and structured well parameter data in response to receiving a pumping signal from the intelligent pumping device. The first construction unit 202 is configured to construct a production prediction vector corresponding to each future time point in a preset future time point sequence, based on the historical production data sequence, the structured well parameter data, and the future time point. The second construction unit 203 is configured to construct a production prediction matrix based on the constructed production prediction vectors. The production probability prediction unit 204 is configured to process the production prediction matrix and the historical production data sequence... The system performs production probability prediction processing on the above-mentioned historical production data sequence to obtain a production probability data sequence; the data filtering unit 205 is configured to perform data filtering processing on the above-mentioned historical production data sequence to obtain a target production data sequence; the production prediction unit 206 is configured to perform production prediction processing on the above-mentioned production prediction matrix and the above-mentioned target production data sequence to obtain an oil well production data sequence; the generation unit 207 is configured to generate an oil well production prediction data sequence based on the above-mentioned production probability data sequence and the above-mentioned oil well production data sequence; and the control unit 208 is configured to control the above-mentioned intelligent oil pumping equipment to perform intelligent oil pumping tasks based on the above-mentioned oil well production prediction data sequence.

[0151] It is understandable that the units described in the device 200 are related to the reference. Figure 1 The steps in the described method correspond to each other. Therefore, the operations, features, and beneficial effects described above for the method also apply to the device 200 and the units contained therein, and will not be repeated here.

[0152] The following is for reference. Figure 3 It shows a schematic diagram of the structure of an electronic device (such as a computing device) 300 suitable for implementing some embodiments of the present disclosure. Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0153] like Figure 3As shown, the electronic device 300 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the electronic device 300. The processing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0154] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 An electronic device 300 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 3 Each box shown can represent a device or multiple devices as needed.

[0155] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 309, or installed from storage device 308, or installed from ROM 302. When the computer program is executed by processing device 301, it performs the functions defined in the methods of some embodiments of this disclosure.

[0156] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0157] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0158] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: in response to receiving a pumping signal sent by the intelligent pumping equipment, acquire historical production data sequences and well parameter structured data; for each future time point in a preset future time point sequence, construct a production prediction vector corresponding to the future time point based on the aforementioned historical production data sequence, the aforementioned well parameter structured data, and the aforementioned future time point; construct a production prediction matrix based on the constructed production prediction vectors; perform production probability prediction processing on the aforementioned production prediction matrix and the aforementioned historical production data sequence to obtain a production probability data sequence; perform data filtering processing on the aforementioned historical production data sequence to obtain a target production data sequence; perform production prediction processing on the aforementioned production prediction matrix and the aforementioned target production data sequence to obtain an oil well production data sequence; generate an oil well production prediction data sequence based on the aforementioned production probability data sequence and the aforementioned oil well production data sequence; and control the aforementioned intelligent pumping equipment to perform intelligent pumping tasks based on the aforementioned oil well production prediction data sequence.

[0159] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0160] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0161] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including an acquisition unit, a first construction unit, a second construction unit, a production probability prediction unit, a data filtering unit, a production prediction unit, a generation unit, and a control unit. The names of these units do not necessarily limit the specific unit; for example, a control unit may also be described as "a unit that controls the aforementioned intelligent oil extraction equipment to perform intelligent oil extraction tasks."

[0162] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0163] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A smart pumping method based on oil well production prediction, comprising: In response to receiving the oil pumping signal sent by the intelligent oil pumping equipment, it acquires historical production data sequences and structured data of oil well parameters; For each future time point in the preset future time point sequence, a production prediction vector corresponding to the future time point is constructed based on the historical production data sequence, the structured data of oil well parameters, and the future time point. Based on the constructed production forecast vectors, a production forecast matrix is ​​constructed. Performing production probability prediction processing on the production prediction matrix and the historical output data sequence to obtain a production probability data sequence, wherein the process of performing production probability prediction processing on the production prediction matrix and the historical output data sequence to obtain a production probability data sequence includes: The historical production data sequence is binarized to obtain a binarized production data sequence; The production prediction matrix and the binarized output data sequence are input into a pre-trained production probability prediction model to obtain a production probability data sequence, wherein each production probability data point in the production probability data sequence corresponds to a future time point, and the production probability prediction model is trained through the following steps: Obtain a sample set, where each sample in the sample set includes a sample production prediction matrix, a sample binarized output data sequence, and a sample production probability data sequence; Based on the sample set, perform the following steps: The sample production prediction matrix and sample binarized output data sequence included in at least one sample in the sample set are input into the feature processing layer in the initial neural network to obtain the historical time series feature matrix corresponding to each sample. The historical temporal feature matrix corresponding to each of the at least one samples is input into the feature encoding layer in the initial neural network to obtain the historical context encoding matrix corresponding to each sample. The historical context encoding matrix and sample production probability data sequence corresponding to each of the at least one samples are input into the autoregressive decoding layer in the initial neural network to obtain the decoding state matrix corresponding to each sample. For each sample in the at least one sample, the decoding state matrix corresponding to each sample is input to the probability output layer to obtain the production probability data sequence; For each of the at least one samples, the production probability data sequence corresponding to each sample and the production probability data sequence of the samples included in the sample are compared to obtain the data comparison result; In response to determining that at least one data comparison result corresponding to the at least one sample meets the preset data comparison conditions, the initial neural network is determined as the production probability prediction model; The historical production data sequence is filtered to obtain the target production data sequence; Perform production prediction processing on the production prediction matrix and the target production data sequence to obtain an oil well production data sequence, wherein the process of performing production prediction processing on the production prediction matrix and the target production data sequence to obtain an oil well production data sequence includes: Based on the target output data sequence and the preset time window, the production prediction matrix is ​​subjected to vector filtering processing to obtain a non-zero output matrix. The target output data sequence is subjected to time-series feature extraction processing to obtain a time-series pattern feature matrix; The non-zero output matrix and the time-series pattern feature matrix are concatenated to obtain a concatenated feature matrix. The concatenated feature matrix is ​​subjected to nonlinear feature mapping processing to obtain a nonlinear feature matrix; The nonlinear feature matrix is ​​subjected to dimension mapping to obtain a feature compression matrix; The feature compression matrix is ​​subjected to regression prediction processing to obtain a future output prediction data sequence; Based on the future production forecast data sequence, an oil well production data sequence is generated; Based on the production probability data sequence and the oil well production data sequence, an oil well production prediction data sequence is generated, wherein each production probability data point in the production probability data sequence corresponds to a future time point, and each oil well production data point in the oil well production data sequence corresponds to a future time point; and the generation of the oil well production prediction data sequence based on the production probability data sequence and the oil well production data sequence includes: For each future time point corresponding to the production probability data sequence, perform the following steps: The production probability data corresponding to the future time point in the production probability data sequence is identified as the production probability data to be processed. The oil well production data corresponding to the future time point in the oil well production data sequence is determined as the oil well production data to be processed. The product of the production probability data to be processed and the oil well production data to be processed is determined as the oil well production prediction data; The determined production forecast data for each oil well are arranged into a sequence of oil well production forecast data. Based on the oil well production prediction data sequence, the intelligent pumping equipment is controlled to perform intelligent pumping tasks.

2. The method according to claim 1, wherein, Each production probability data point in the production probability data sequence corresponds to a future time point, and each sample production probability data point in the sample production probability data sequence corresponds to a future time point; and the data comparison between the production probability data sequence corresponding to each sample and the sample production probability data sequence included in the sample, to obtain the data comparison result, includes: For each future time point corresponding to the production probability data sequence, perform the following steps: The production probability data corresponding to the future time point in the production probability data sequence is determined as the target production probability data; The sample production probability data corresponding to the future time point in the sample production probability data sequence is determined as the target sample production probability data; The logarithm of the target production probability data is determined as logarithmic probability data; The product of the target sample production probability data and the logarithmic probability data is determined as the first numerical data item; The difference between the preset value and the target production probability data is determined as the target difference data; The difference between the preset value and the target sample production probability data is determined as the target sample difference data; The logarithm of the target difference data is determined as the logarithm of the difference data; The product of the target sample difference data and the logarithm of the difference data is determined as the second numerical data; The sum of the first item data and the second item data is determined as the target item data; The average value of the determined number of target items is used as the average number of items. The negative value of the average number of items is determined as the data comparison result.

3. An intelligent pumping device based on oil well production prediction, comprising: The acquisition unit is configured to acquire historical production data sequences and structured well parameter data in response to receiving a pumping signal from the intelligent pumping equipment. The first construction unit is configured to construct a production prediction vector corresponding to each future time point in a preset future time point sequence, based on the historical production data sequence, the structured data of oil well parameters, and the future time point. The second building unit is configured to construct a production forecast matrix based on the constructed production forecast vectors. A production probability prediction unit is configured to perform production probability prediction processing on the production prediction matrix and the historical output data sequence to obtain a production probability data sequence. The process includes: binarizing the historical output data sequence to obtain a binarized output data sequence; and inputting the production prediction matrix and the binarized output data sequence into a pre-trained production probability prediction model to obtain the production probability data sequence. Each production probability data point in the production probability data sequence corresponds to a future time point. The production probability prediction model is trained through the following steps: acquiring a sample set, where each sample includes a sample production prediction matrix, a sample binarized output data sequence, and a sample production probability data sequence; and based on the sample set, performing the following steps: processing the sample production prediction matrix and sample binarized output data sequence included in at least one sample in the sample set... The sequence is input into the feature processing layer of the initial neural network to obtain the historical time-series feature matrix corresponding to each sample; the historical time-series feature matrix corresponding to each sample in the at least one sample is input into the feature encoding layer of the initial neural network to obtain the historical context encoding matrix corresponding to each sample; the historical context encoding matrix corresponding to each sample in the at least one sample and the sample production probability data sequence are input into the autoregressive decoding layer of the initial neural network to obtain the decoding state matrix corresponding to each sample; for each sample in the at least one sample, the decoding state matrix corresponding to each sample is input into the probability output layer to obtain the production probability data sequence; for each sample in the at least one sample, the production probability data sequence corresponding to each sample and the sample production probability data sequence included in the sample are compared to obtain the data comparison result; in response to determining that at least one data comparison result corresponding to the at least one sample meets the preset data comparison condition, the initial neural network is determined as a production probability prediction model; A data filtering unit is configured to perform data filtering processing on the historical production data sequence to obtain a target production data sequence. A production prediction unit is configured to perform production prediction processing on the production prediction matrix and the target production data sequence to obtain an oil well production data sequence. The process of performing production prediction processing on the production prediction matrix and the target production data sequence to obtain the oil well production data sequence includes: performing vector filtering processing on the production prediction matrix based on the target production data sequence and a preset time window to obtain a non-zero production matrix; performing time-series feature extraction processing on the target production data sequence to obtain a time-series pattern feature matrix; performing feature concatenation processing on the non-zero production matrix and the time-series pattern feature matrix to obtain a concatenated feature matrix; performing nonlinear feature mapping processing on the concatenated feature matrix to obtain a nonlinear feature matrix; performing dimensionality mapping processing on the nonlinear feature matrix to obtain a feature compression matrix; performing regression prediction processing on the feature compression matrix to obtain a future production prediction data sequence; and generating an oil well production data sequence based on the future production prediction data sequence. A generation unit is configured to generate an oil well production prediction data sequence based on the production probability data sequence and the oil well production data sequence, wherein each production probability data in the production probability data sequence corresponds to a future time point, and each oil well production data in the oil well production data sequence corresponds to a future time point; and the generation of the oil well production prediction data sequence based on the production probability data sequence and the oil well production data sequence includes: for each future time point corresponding to the production probability data sequence, performing the following steps: determining the production probability data corresponding to the future time point in the production probability data sequence as production probability data to be processed; determining the oil well production data corresponding to the future time point in the oil well production data sequence as oil well production data to be processed; determining the product of the production probability data to be processed and the oil well production data to be processed as oil well production prediction data; and arranging the determined oil well production prediction data into an oil well production prediction data sequence. The control unit is configured to control the intelligent pumping equipment to perform intelligent pumping tasks based on the oil well production prediction data sequence.

4. An electronic device, comprising: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 2.

5. A computer-readable medium having a computer program stored thereon, wherein, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 2.

Images