Concentration completion method and training method of concentration completion model

Abstract

The invention provides a concentration completion method and a training method of a concentration completion model, which can be applied to the technical field of in-situ leaching mining. The concentration complementing method comprises the following steps: performing feature extraction on leachate concentration data acquired within a preset time to obtain a leachate concentration feature tensor; aiming at the mth data element in the M discrete metal concentration characteristic tensors, embedding the mth data element into a characteristic part, representing that the metal concentration data has missing, in the metal concentration characteristic tensors to obtain an mth target metal concentration characteristic tensor; according to the extraction and injection unit feature tensor, the liquid extraction amount feature tensor, the liquid injection amount feature tensor, the time sequence compression feature vector and the mth target metal concentration feature tensor, obtaining an embedded evaluation value corresponding to the mth data element; and according to the M embedded evaluation values, determining the complemented target metal concentration data at the position where the metal concentration data is missing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of in-situ leaching mining, and more particularly to a concentration completion method and a training method of a concentration completion model. BACKGROUND

[0002] With the increase of consumption of mineral resources, some mineral resources are increasingly scarce. How to realize efficient development and rational utilization of mineral resources has become a problem to be solved in the mining industry. As an important mining method without goaf, in-situ leaching mining technology has been widely used in mining practice in recent years due to its advantages in reducing production cost, shortening construction period, reducing surface disturbance and improving resource utilization.

[0003] In related technologies, the missing of leaching liquid metal concentration data is often completed by simple difference or average completion method, and the accuracy of the completion result is not high, which further affects the scientificity of the management decision of the mine resources. SUMMARY

[0004] In view of the above problems, the present application provides a concentration completion method and a training method of a concentration completion model.

[0005] According to a first aspect of the present application, a concentration completion method is provided, comprising: performing feature extraction on leaching liquid concentration data obtained within a preset time to obtain a leaching liquid concentration feature tensor, wherein the leaching liquid concentration feature tensor comprises a pumping and injection unit feature tensor, a pumping volume feature tensor, an injection volume feature tensor, a timestamp feature tensor, a metal concentration feature tensor and a discrete metal concentration feature tensor, the metal concentration feature tensor has a feature representing missing metal concentration data, and the discrete metal concentration feature tensor comprises M discrete metal concentration data; performing global time sequence compression processing on the timestamp feature tensor to obtain a time sequence compression feature vector; embedding an mth data element in the M discrete metal concentration feature tensors into the feature representing the missing metal concentration data in the metal concentration feature tensor to obtain an mth target metal concentration feature tensor; obtaining an embedding evaluation value corresponding to the mth data element according to the pumping and injection unit feature tensor, the pumping volume feature tensor, the injection volume feature tensor, the time sequence compression feature vector and the mth target metal concentration feature tensor; and determining target metal concentration data after completion at the position where the metal concentration data is missing according to M embedding evaluation values.

[0006] According to an embodiment of this application, the above-mentioned global temporal compression processing of the timestamp feature tensor to obtain a temporal compressed feature vector includes: dividing the timestamp feature tensor into N sub-timestamp feature tensors; for the nth of the N sub-timestamp feature tensors; performing a discrete cosine transform on the nth sub-timestamp feature tensor based on a preset nth discrete cosine transform basis function corresponding to the nth sub-timestamp feature tensor to obtain the nth sub-temporal compressed feature vector; and concatenating the N sub-temporal compressed feature vectors to obtain a temporal compressed feature vector.

[0007] According to an embodiment of this application, obtaining the embedding evaluation value corresponding to the m-th data element based on the injection unit feature tensor, the injection volume feature tensor, the injection volume feature tensor, the time-series compression feature vector, and the m-th target metal concentration feature tensor includes: performing attention calculation on the timestamp feature tensor based on the injection volume feature tensor, the injection volume feature tensor, and the time-series compression feature vector to obtain a processed timestamp feature tensor; performing matrix multiplication on the injection unit feature tensor, the injection volume feature tensor, the injection volume feature tensor, and the processed timestamp feature tensor to obtain an injection-time-series coupling tensor; and performing a dot product operation on the injection-time-series coupling tensor and the m-th target metal concentration feature tensor to obtain the embedding evaluation value corresponding to the m-th target metal concentration feature tensor.

[0008] According to an embodiment of this application, the above-mentioned attention calculation on the timestamp feature tensor based on the above-mentioned extraction volume feature tensor, the above-mentioned injection volume feature tensor, and the above-mentioned time-series compression feature vector to obtain a processed timestamp feature tensor includes: performing matrix multiplication on the above-mentioned extraction volume feature tensor, the above-mentioned injection volume feature tensor, and the above-mentioned time-series compression feature vector to obtain a fused tensor; mapping the above-mentioned fused tensor to obtain attention features; and using the above-mentioned attention features to process the above-mentioned timestamp feature tensor to obtain a processed timestamp feature tensor.

[0009] According to an embodiment of this application, the above-mentioned leachate concentration data is obtained by injecting chemical leaching solution into the target mining area through an injection unit, thereby selectively leaching the useful metal components in the target mining area from the ore.

[0010] The second aspect of this application provides a training method for a concentration completion model, comprising: extracting features from sample leachate concentration data to obtain a sample leachate concentration feature tensor, wherein the sample leachate concentration feature tensor includes a sample injection unit feature tensor, a sample extraction volume feature tensor, a sample injection volume feature tensor, a sample timestamp feature tensor, a sample metal concentration feature tensor, and a sample discrete metal concentration feature tensor. The sample metal concentration feature tensor has features characterizing the presence of missing sample metal concentration data. The sample discrete metal concentration feature tensor includes K sample discrete metal concentration data, which in turn includes sample label discrete metal concentration data corresponding to the missing sample metal concentration data and multiple sample candidate discrete metal concentration data. The method further includes processing the sample timestamp feature tensor... The tensor undergoes global temporal compression to obtain a sample temporal compressed feature vector. For the k-th sample data element in the aforementioned K sample discrete metal concentration feature tensors, this k-th sample data element is embedded into the sample metal concentration feature tensor at the feature representing the missing sample metal concentration data, resulting in the k-th sample target metal concentration feature tensor. Based on the aforementioned sample injection unit feature tensor, sample extraction volume feature tensor, sample injection volume feature tensor, sample temporal compressed feature vector, and the k-th sample target metal concentration feature tensor, an embedding evaluation value is obtained for the m-th data element. Based on the target loss function and the M embedding evaluation values, a loss value is obtained, and the parameters of the concentration completion model are adjusted based on this loss value to obtain a trained concentration completion model.

[0011] According to an embodiment of this application, the M embedding evaluation values ​​include the label embedding evaluation value corresponding to the discrete metal concentration data of the sample label and the candidate embedding evaluation value corresponding to the multiple sample candidate discrete metal concentration data respectively. The loss value obtained based on the target loss function and the M embedding evaluation values ​​includes: obtaining the loss value based on the target loss function and the label embedding evaluation value and the candidate embedding evaluation value.

[0012] According to an embodiment of this application, before performing feature extraction on the sample leachate concentration data, the method further includes: pre-filling the sample candidate metal concentration data to obtain the sample leachate concentration data.

[0013] According to an embodiment of this application, the above-mentioned pre-filling of sample candidate metal concentration data to obtain the sample leachate concentration data includes: when there are multiple consecutive missing locations in the sample candidate metal concentration data, determining a first target filling value based on a preset number of sample candidate metal concentration data before the first consecutive missing location among the multiple consecutive missing locations; and filling the multiple consecutive missing locations using the first target filling value.

[0014] According to an embodiment of this application, the above-mentioned pre-filling of sample candidate metal concentration data to obtain the sample leachate concentration data includes: when there are multiple discontinuous missing parts in the sample candidate metal concentration data, determining a second target filling value for each of the multiple discontinuous missing parts based on a preset number of sample candidate metal concentration data before the discontinuous missing part; and filling the discontinuous missing values ​​using the second target filling value.

[0015] A third aspect of this application provides a concentration completion device, comprising: an extraction module for extracting features from leachate concentration data acquired within a preset time period to obtain a leachate concentration feature tensor, wherein the leachate concentration feature tensor includes an injection unit feature tensor, an injection volume feature tensor, an injection volume feature tensor, a timestamp feature tensor, a metal concentration feature tensor, and a discrete metal concentration feature tensor, wherein the metal concentration feature tensor contains features characterizing the absence of metal concentration data, and the discrete metal concentration feature tensor includes M discrete metal concentration data; and a compression module for performing global temporal compression processing on the timestamp feature tensor to obtain a temporal compressed feature vector. The embedding module is used to embed the m-th data element from the M discrete metal concentration feature tensors into the feature representing the missing metal concentration data in the metal concentration feature tensor, thereby obtaining the m-th target metal concentration feature tensor. The evaluation module is used to obtain the embedding evaluation value corresponding to the m-th data element based on the injection unit feature tensor, the injection volume feature tensor, the injection volume feature tensor, the time-series compression feature vector, and the m-th target metal concentration feature tensor. The completion module is used to determine the completed target metal concentration data at the location where the missing metal concentration data is located based on the M embedding evaluation values.

[0016] A fourth aspect of this application provides a training apparatus for a concentration completion model, comprising: a sample extraction module for extracting features from sample leachate concentration data to obtain a sample leachate concentration feature tensor, wherein the sample leachate concentration feature tensor includes a sample injection unit feature tensor, a sample extraction volume feature tensor, a sample injection volume feature tensor, a sample timestamp feature tensor, a sample metal concentration feature tensor, and a sample discrete metal concentration feature tensor; the sample metal concentration feature tensor has features characterizing the existence of missing sample metal concentration data; the sample discrete metal concentration feature tensor includes K sample discrete metal concentration data, the K sample discrete metal concentration data including sample label discrete metal concentration data corresponding to the missing sample metal concentration data and multiple sample candidate discrete metal concentration data; and a sample compression module for compressing the sample timestamp feature tensor. A global temporal compression process is performed to obtain a sample temporal compression feature vector. A sample embedding module is used to embed the k-th sample data element from the k-th discrete metal concentration feature tensor into the feature tensor representing missing metal concentration data, thus obtaining the k-th target metal concentration feature tensor. A sample evaluation module is used to obtain a sample embedding evaluation value for the m-th data element based on the sample injection unit feature tensor, the sample extraction volume feature tensor, the sample injection volume feature tensor, the sample temporal compression feature vector, and the k-th target metal concentration feature tensor. An adjustment module is used to obtain a loss value based on the target loss function and the M embedding evaluation values, and adjust the parameters of the concentration completion model based on the loss value to obtain a trained concentration completion model.

[0017] A fifth aspect of this application provides an electronic device comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0018] A sixth aspect of this application also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implement the steps of the above-described method.

[0019] A seventh aspect of this application also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method.

[0020] According to the embodiments of this application, by constructing a multi-dimensional tensor from the leachate concentration data, and integrating the injection unit, injection volume, extraction volume, timestamp, metal concentration, and discrete metal concentration, a complete profile of the high-dimensional leachate concentration data is obtained. Then, by performing global temporal compression processing on the timestamp, long temporal dependencies are captured, thereby reducing the complexity of subsequent embedding calculations. For the m-th data element in the M discrete metal concentration feature tensors, the m-th data element is embedded into the missing position, thereby generating the corresponding m-th target metal concentration feature tensor to obtain the embedding evaluation value corresponding to the m-th data element. This enables the determination of the missing data in the metal concentration data, which can restore the integrity of the leachate concentration data, improve the accuracy of the completion results, and thus improve the stability of the mining process and the efficiency of resource recovery, and enhance the scientific nature of mine resource management decisions. Attached Figure Description

[0021] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0022] Figure 1 The illustration shows an application scenario of the concentration completion method and apparatus, and the training method and apparatus for the concentration completion model according to embodiments of this application;

[0023] Figure 2 A flowchart of a concentration completion method according to an embodiment of this application is shown;

[0024] Figure 3 A flowchart of another concentration completion method according to an embodiment of this application is shown;

[0025] Figure 4 A flowchart illustrating a training method for a concentration completion model according to an embodiment of this application is shown;

[0026] Figure 5 A structural block diagram of a concentration compensation device according to an embodiment of this application is shown;

[0027] Figure 6 A structural block diagram of a training apparatus for a concentration completion model according to an embodiment of this application is shown;

[0028] Figure 7 A block diagram of an electronic device suitable for implementing a concentration completion method and a training method for a concentration completion model according to an embodiment of this application is shown. Detailed Implementation

[0029] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0030] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0031] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0032] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0033] With the continuous advancement of industrialization, the consumption of mineral resources is increasing, and some resources are becoming increasingly scarce. How to achieve efficient development and rational utilization of mineral resources has become an urgent problem to be solved in the mining industry. In-situ leaching mining technology, as an important method of mining without goaf, has been widely used in mining practice in recent years due to its advantages in reducing production costs, shortening construction cycles, reducing surface disturbance, and improving resource utilization. Its basic process flow is as follows: chemical leaching solution is injected into the ore layer through injection holes, allowing useful metal components to be selectively leached from the ore. After generating a mineral-bearing solution, it is extracted to the surface through extraction holes, and the extraction and recovery process is completed in surface facilities. In the in-situ leaching mining process, the metal concentration in the leachate is an important production indicator that directly affects mineral resource mining plans, economic benefit assessments, and mining area scheduling and management. This indicator is not only related to the stability of the mining process and resource recovery efficiency, but also a core parameter guiding the control of injection and extraction volumes and the prediction of mining area production capacity. However, in actual production, due to factors such as equipment failure, data acquisition delays, and human error, the metal concentration data of the leachate often suffers from varying degrees of missing or abnormalities, posing difficulties for resource assessment and production scheduling. Furthermore, because metal concentration is affected by multiple factors, such as the volume of injected and extracted fluids, geological structure, and ore body distribution, its variation patterns exhibit strong temporal correlation and volatility, lacking stability.

[0034] In related technologies, when completing missing data, simple interpolation or averaging methods are often used, which makes it difficult to fully explore the temporal characteristics of concentration data, resulting in low accuracy of the completion results and affecting the scientific nature of mine economic analysis and resource management decisions.

[0035] In view of this, embodiments of this application provide a concentration completion method, comprising: extracting features from leachate concentration data acquired within a preset time period to obtain a leachate concentration feature tensor, wherein the leachate concentration feature tensor includes an injection unit feature tensor, an injection volume feature tensor, an injection volume feature tensor, a timestamp feature tensor, a metal concentration feature tensor, and a discrete metal concentration feature tensor, wherein the metal concentration feature tensor contains features characterizing the absence of metal concentration data, and the discrete metal concentration feature tensor includes M discrete metal concentration data; and performing global temporal compression processing on the timestamp feature tensor. Obtain the temporal compression feature vector; for the m-th data element in the M discrete metal concentration feature tensors, embed the m-th data element into the metal concentration feature tensor at the feature location representing the missing metal concentration data, to obtain the m-th target metal concentration feature tensor; based on the injection unit feature tensor, the injection volume feature tensor, the temporal compression feature vector, and the m-th target metal concentration feature tensor, obtain the embedding evaluation value corresponding to the m-th data element; based on the M embedding evaluation values, determine the target metal concentration data after completion at the location where the metal concentration data is missing.

[0036] Figure 1 The diagram illustrates an application scenario of the concentration completion method and apparatus, and the training method and apparatus for the concentration completion model, according to embodiments of this application.

[0037] like Figure 1 As shown, application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0038] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0039] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0040] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0041] It should be noted that the concentration completion method and concentration completion model training method provided in this application embodiment can generally be executed by server 105. Correspondingly, the concentration completion device and concentration completion model training device provided in this application embodiment can generally be located in server 105. The concentration completion method and concentration completion model training method provided in this application embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the concentration completion device and concentration completion model training device provided in this application embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.

[0042] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0043] The following will be based on Figure 1 The described scene, through Figures 2-3 The concentration completion method of the disclosed embodiments is described in detail.

[0044] Figure 2 A flowchart of a concentration completion method according to an embodiment of this application is shown.

[0045] like Figure 2 As shown, the concentration completion method in this embodiment includes operations S210 to S250.

[0046] In operation S210, feature extraction is performed on the leachate concentration data acquired within a preset time to obtain the leachate concentration feature tensor.

[0047] The leachate concentration feature tensor includes the injection unit feature tensor, the extraction volume feature tensor, the injection volume feature tensor, the timestamp feature tensor, the metal concentration feature tensor, and the discrete metal concentration feature tensor. The metal concentration feature tensor contains features that characterize the absence of metal concentration data, and the discrete metal concentration feature tensor includes M discrete metal concentration data.

[0048] According to the embodiments of this application, feature extraction can be performed on leachate concentration data using a natural language processing library, such as Spacy.

[0049] According to an embodiment of this application, the above-mentioned leachate concentration data can be represented as (u, e, i, mc, t), where u represents the injection unit identifier, e represents the injection volume, i represents the extraction volume, mc represents the metal concentration, and t represents the timestamp. For example, for a 10-second leaching concentration dataset, if there are missing metal concentration data at the 4th and 6th seconds, and no missing metal concentration data at the 3rd second, it can be represented as (1, 200, 170, 30, 3). Then, the leaching concentration data at the 4th second with missing metal concentration can be represented as (1, 200, 180, X, 4), and the leaching concentration data at the 6th second with missing metal concentration can be represented as (1, 300, 250, X, 6), where X represents the missing metal concentration data corresponding to that timestamp. Corresponding to the missing metal concentration data at the 4th and 6th seconds, there are data elements 10 and 20 in the discrete metal concentration feature tensor.

[0050] In operation S220, global temporal compression is performed on the timestamp feature tensor to obtain the temporal compressed feature vector.

[0051] According to the embodiments of this application, the timestamp feature tensor can be compressed using algorithms such as discrete cosine transform, discrete Fourier transform, and wavelet transform to obtain a time-series compressed feature vector.

[0052] In operation S230, for the m-th data element in the M discrete metal concentration feature tensors, the m-th data element is embedded into the feature of the metal concentration feature tensor that represents the missing metal concentration data, so as to obtain the m-th target metal concentration feature tensor.

[0053] For example, in the above example, data element 10 and data element 20 in the discrete metal concentration feature tensor can be embedded into the leaching concentration corresponding to the leaching concentration data at the 4th second, respectively, to obtain (1, 200, 180, 10, 4) and (1, 200, 180, 20, 4), thus obtaining two target metal concentration feature tensors.

[0054] In operation S240, based on the feature tensor of the injection unit, the feature tensor of the injection volume, the feature tensor of the injection volume, the temporal compression feature vector, and the feature tensor of the m-th target metal concentration, the embedding evaluation value corresponding to the m-th data element is obtained.

[0055] In operation S250, based on M embedded evaluation values, the target metal concentration data is filled in at the locations where the metal concentration data is missing.

[0056] For example, the embedding of data elements 10 and 20 in the discrete metal concentration feature tensor for the leaching concentration data at the 4th second has an embedding evaluation value of 95 for the first data element 10 and an embedding evaluation value of 80 for the second data element 20. Since the embedding evaluation value of data element 10 is greater than that of data element 20, data element 10 can be used as the complete data corresponding to the leaching concentration data at the 4th second. In this way, the missing parts in the leaching concentration data are filled one by one to obtain the completed target metal concentration data.

[0057] According to the embodiments of this application, by constructing a multi-dimensional tensor from the leachate concentration data, and integrating the injection unit, injection volume, extraction volume, timestamp, metal concentration, and discrete metal concentration, a complete profile of the high-dimensional leachate concentration data is obtained. Then, by performing global temporal compression processing on the timestamp, long temporal dependencies are captured, thereby reducing the complexity of subsequent embedding calculations. For the m-th data element in the M discrete metal concentration feature tensors, the m-th data element is embedded into the missing position, thereby generating the corresponding m-th target metal concentration feature tensor to obtain the embedding evaluation value corresponding to the m-th data element. This enables the determination of the missing data in the metal concentration data, which can restore the integrity of the leachate concentration data, improve the accuracy of the completion results, and thus improve the stability of the mining process and the efficiency of resource recovery, and enhance the scientific nature of mine resource management decisions.

[0058] According to the embodiments of this application, the above-mentioned leachate concentration data is obtained by injecting chemical leaching solution into the target mining area through an injection unit, thereby selectively leaching the useful metal components in the target mining area from the ore.

[0059] According to an embodiment of this application, the above-mentioned global temporal compression processing of the timestamp feature tensor to obtain a temporal compressed feature vector includes: dividing the timestamp feature tensor into N sub-timestamp feature tensors; for the nth sub-timestamp feature tensor; performing a discrete cosine transform on the nth sub-timestamp feature tensor based on a preset nth discrete cosine transform basis function corresponding to the nth sub-timestamp feature tensor to obtain the nth sub-temporal compressed feature vector; and concatenating the N sub-temporal compressed feature vectors to obtain the temporal compressed feature vector.

[0060] According to an embodiment of this application, the leachate concentration data includes... Each injection unit identifier and T timestamps are used to embed the timestamp tensor. First, embed the timestamp into the tensor along the embedding dimension. Dividing it into N parts, under this partitioning method, They will be represented as [ , , , ],in, ( N is set to an integer that is divisible by K.

[0061] The nth sub-temporal compression feature vector can be calculated using the following formula (1).

[0062]

[0063] ,

[0064] st n (1)

[0065] in, This represents the compressed vector of the nth sub-timestamp. Let N represent the nth sub-timestamp feature tensor, and N represent the number of sub-timestamp feature tensors. Let represent the basis function of the nth discrete cosine transform at the j-th position. This represents the nth sub-timestamp feature tensor at the j-th position.

[0066] The temporal compression feature vector can be calculated using the following formula (2).

[0067]

[0068] ,

[0069] st n (2)

[0070] in, Represents the temporal compression feature vector. This represents the compressed feature vector of the first sub-time series. This represents the compressed feature vector of the nth sub-time series. This represents the Nth sub-temporal compression feature vector.

[0071] According to the embodiments of this application, by dividing the timestamp feature tensor into dimensions and performing discrete cosine transform on each dimension, the dimensionality can be reduced while retaining the key features in each sub-timestamp feature tensor, thereby compressing the high-dimensional time series signal into a compact time series compressed feature vector and reducing the computational burden on the hardware device.

[0072] According to an embodiment of this application, the above-mentioned method of obtaining the embedding evaluation value corresponding to the m-th data element based on the injection unit feature tensor, the injection volume feature tensor, the injection volume feature tensor, the time-series compression feature vector, and the m-th target metal concentration feature tensor includes: performing attention calculation on the timestamp feature tensor based on the injection volume feature tensor, the injection volume feature tensor, and the time-series compression feature vector to obtain the processed timestamp feature tensor; performing matrix multiplication on the injection unit feature tensor, the injection volume feature tensor, the injection volume feature tensor, and the processed timestamp feature tensor to obtain the injection-injection time-series coupling tensor; and performing dot product operation on the injection-injection time-series coupling tensor and the m-th target metal concentration feature tensor to obtain the embedding evaluation value corresponding to the m-th target metal concentration feature tensor.

[0073] According to an embodiment of this application, the above-mentioned attention calculation on the timestamp feature tensor based on the pumping volume feature tensor, the injection volume feature tensor, and the time-series compression feature vector to obtain the processed timestamp feature tensor includes: performing matrix multiplication on the pumping volume feature tensor, the injection volume feature tensor, and the time-series compression feature vector to obtain a fused tensor; mapping the fused tensor to obtain attention features; and using the attention features to process the timestamp feature tensor to obtain the processed timestamp feature tensor.

[0074] The processed timestamp feature tensor can be calculated using the following formula (3).

[0075] =Att (3)

[0076] in, Let Att represent the processed timestamp feature tensor, and Att represent the attention feature. Represents the timestamp feature tensor. This represents the element-wise multiplication operation of vectors.

[0077] Attention features can be calculated using the following formula (4).

[0078] Att = sigmoid(fc( )) (4)

[0079] Where Att represents the attention feature, sigmoid represents the activation function, and fc represents the fully connected layer. Represents the injection volume characteristic tensor. Let F represent the liquid extraction volume feature tensor, and let F represent the time-series compressed feature vector.

[0080] The embedding evaluation value corresponding to the m-th target metal concentration feature tensor can be calculated using the following formula (5).

[0081] ·K (5)

[0082] in, Let K represent the embedding evaluation value corresponding to the m-th target metal concentration feature tensor, and let K represent the injection time-series coupling tensor. Let m represent the feature tensor of the m-th target metal concentration.

[0083] The injection timing coupling tensor can be calculated using the following formula (6).

[0084] K= (6)

[0085] in, Represents the feature tensor of the injection unit. Represents the injection volume characteristic tensor. This represents the characteristic tensor of the pumped volume. This represents the processed timestamp feature tensor.

[0086] According to the embodiments of this application, a fusion tensor is obtained by performing matrix multiplication on the pumping volume feature tensor, the injection volume feature tensor, and the time-series compressed feature vector. The fusion tensor is then mapped through a fully connected layer to obtain attention features. These attention features are then used to process the timestamp feature tensor to obtain a processed timestamp feature tensor. This results in each timestamp feature tensor being scaled accordingly by the corresponding attention features, thereby enhancing the concentration completion model's focus on key timestamp feature tensors and improving prediction accuracy.

[0087] Figure 3 A schematic diagram of a concentration completion method according to an embodiment of this application is shown.

[0088] like Figure 3 As shown, for N sub-timestamp feature tensors 301, a discrete cosine transform is performed on them with each preset nth discrete cosine transform basis function 302 to obtain a time-series compressed feature vector 303. Then, attention calculation is performed on them with the extraction volume feature tensor 304 and the injection volume feature tensor 305 to obtain a processed timestamp feature tensor 306. Further, the processed timestamp feature tensor 306 is multiplied by the extraction and injection unit feature tensor 307, the extraction volume feature tensor 304, and the injection volume feature tensor 305 to obtain the extraction and injection time-series coupling tensor 308. The extraction and injection time-series coupling tensor 308 is multiplied by the mth target metal concentration feature tensor 309 to obtain the embedding evaluation value 310 corresponding to the 1st to 4th target metal concentration feature tensors. Then, based on each embedding evaluation value, the data element mc1 is selected as the target metal concentration data after completion at the location where the metal concentration data is missing.

[0089] For example, Table 1 shows the evaluation results of the concentration completion method provided in this application compared with other concentration completion methods. (See Table 1 below.)

[0090] Table 1

[0091] Root mean square error Mean absolute error Determination coefficient Gradient boosting regression tree 0.246885 0.202352 0.895364 Gated recurrent unit 0.286947 0.241249 0.940003 Long short-term memory network 0.184476 0.145619 0.947741 Temporal convolution network 0.151406 0.117299 0.950647 Convolutional neural network-long short-term memory network 0.108692 0.085085 0.939663 Temporal convolution network-long short-term memory network 0.187489 0.168394 0.947860 The method of the present application 0.079847 0.068822 0.995675

[0092] As can be seen from Table 1, the concentration completion method provided in this application can achieve smaller root mean square error, mean absolute error and higher coefficient of determination compared with other concentration completion methods, indicating that the concentration completion method provided in this application has better generalization ability than existing methods.

[0093] Based on the above concentration completion method, this application also provides a training method for the concentration completion model corresponding to the concentration completion method.

[0094] Figure 4 A flowchart of a training method for a concentration completion model according to an embodiment of this application is shown.

[0095] like Figure 4 As shown, the training method of the concentration completion model in this embodiment includes operations S410 to S450.

[0096] In operation S410, feature extraction is performed on the sample leachate concentration data to obtain the sample leachate concentration feature tensor.

[0097] The sample leachate concentration feature tensor includes the sample extraction unit feature tensor, the sample extraction volume feature tensor, the sample injection volume feature tensor, the sample timestamp feature tensor, the sample metal concentration feature tensor, and the sample discrete metal concentration feature tensor. The sample metal concentration feature tensor has features that characterize the existence of missing sample metal concentration data. The sample discrete metal concentration feature tensor includes K sample discrete metal concentration data. Among the K sample discrete metal concentration data, there are sample label discrete metal concentration data corresponding to the missing sample metal concentration data and multiple sample candidate discrete metal concentration data.

[0098] In operation S420, global temporal compression is performed on the sample timestamp feature tensor to obtain the sample temporal compressed feature vector.

[0099] In operation S430, for the kth sample data element in the discrete metal concentration feature tensor of K samples, the kth sample data element is embedded into the feature of the sample metal concentration feature tensor that represents the missing sample metal concentration data, and the target metal concentration feature tensor of the kth sample is obtained.

[0100] In operation S440, based on the sample injection unit feature tensor, sample extraction volume feature tensor, sample injection volume feature tensor, sample time-series compression feature vector, and the target metal concentration feature tensor of the kth sample, the sample embedding evaluation value corresponding to the mth data element is obtained.

[0101] In operation S450, based on the target loss function, the loss value is obtained according to M embedded evaluation values, and the parameters of the concentration completion model are adjusted based on the loss value to obtain the trained concentration completion model.

[0102] For example, for the sample leachate concentration feature tensor, the sample metal concentration data at the 4th second is extracted as the sample label discrete metal concentration data, and multiple sample candidate discrete metal concentration data are randomly generated. During training, the sample label discrete metal concentration data and the multiple sample candidate discrete metal concentration data are embedded into the sample metal concentration feature tensor at the features representing missing sample metal concentration data, respectively. This yields the sample embedding evaluation value corresponding to the sample label discrete metal concentration data and the sample embedding evaluation value corresponding to the multiple sample candidate discrete metal concentration data. Then, based on the aforementioned two evaluation values, a loss value is obtained. The process is iterated until the loss value converges, thereby obtaining the trained concentration completion model.

[0103] According to the embodiments of this application, the above-mentioned M embedding evaluation values ​​include the label embedding evaluation value corresponding to the sample label discrete metal concentration data and the candidate embedding evaluation values ​​corresponding to the multiple sample candidate discrete metal concentration data respectively. Based on the target loss function, the loss value is obtained according to the M embedding evaluation values, including: obtaining the loss value based on the target loss function, according to the label embedding evaluation value and the candidate embedding evaluation value.

[0104] The loss value can be calculated according to the following formula (7).

[0105] (7)

[0106] in, Indicates the loss value. This represents the sample leachate concentration data corresponding to the discrete metal concentration data of the sample candidates. This represents the sample leachate concentration data corresponding to the discrete metal concentration data of the sample label. This represents the discrete metal concentration data of the sample candidates. This represents discrete metal concentration data for sample labels. Indicates the candidate embedding evaluation value. This indicates that the label embeds the evaluation value.

[0107] According to an embodiment of this application, before performing feature extraction on the sample leachate concentration data, the method further includes: pre-filling the sample candidate metal concentration data to obtain the sample leachate concentration data.

[0108] According to an embodiment of this application, after pre-filling the sample candidate metal concentrations, each feature value can be further mapped to the [0,1] interval using the minimum-maximum normalization method for normalization processing. Furthermore, the normalized sample candidate metal concentration data can be divided into a training set, a validation set, and a test set in a 7:1.5:1.5 ratio, and the data in the training set can be used as the sample leachate concentration data.

[0109] According to an embodiment of this application, the above-mentioned pre-filling of sample candidate metal concentration data to obtain sample leachate concentration data includes: when there are multiple consecutive missing locations in the sample candidate metal concentration data, determining a first target filling value based on a preset number of sample candidate metal concentration data before the first consecutive missing location among the multiple consecutive missing locations; and filling the multiple consecutive missing locations using the first target filling value.

[0110] According to the embodiments of this application, the above-mentioned pre-filling of sample candidate metal concentration data to obtain sample leachate concentration data includes: when there are multiple discontinuous missing points in the sample candidate metal concentration data, determining a second target filling value for each discontinuous missing point based on a preset number of sample candidate metal concentration data before the discontinuous missing point; and filling the discontinuous missing values ​​using the second target filling value.

[0111] According to embodiments of this application, the aforementioned preset number can be, for example, three. For instance, for a 20-second sample candidate metal concentration dataset, if multiple consecutive missing values ​​exist in seconds 6-8 and 10-14, the first target filling value can be the average of the sample candidate metal concentration data from seconds 3-5. For a sample candidate metal concentration dataset with missing values ​​in seconds 4 and 12, the second target filling value corresponding to second 4 can be the average of the sample candidate metal concentrations from seconds 1-3, and the second target filling value corresponding to second 12 can be the average of the sample candidate metal concentrations from seconds 9-11.

[0112] Based on the above concentration compensation method, this application also provides a concentration compensation device. The following will be combined with... Figure 5 The device is described in detail.

[0113] Figure 5 A structural block diagram of a concentration compensation device according to an embodiment of this application is shown.

[0114] like Figure 5As shown, the concentration completion device 500 of this embodiment includes an extraction module 510, a compression module 520, an embedding module 530, an evaluation module 540, and a completion module 550.

[0115] The extraction module 510 is used to extract features from the leachate concentration data acquired within a preset time period to obtain a leachate concentration feature tensor. This leachate concentration feature tensor includes a pumping unit feature tensor, a pumping volume feature tensor, an injection volume feature tensor, a timestamp feature tensor, a metal concentration feature tensor, and a discrete metal concentration feature tensor. The metal concentration feature tensor contains features indicating missing metal concentration data, and the discrete metal concentration feature tensor includes M discrete metal concentration data points. The compression module is used to perform global temporal compression processing on the timestamp feature tensor to obtain a temporal compressed feature vector. In one embodiment, the extraction module 510 can be used to execute the operation S210 described above, which will not be repeated here.

[0116] Compression module 520 is used to perform global temporal compression processing on the aforementioned timestamp feature tensor to obtain a temporal compressed feature vector. In one embodiment, compression module 520 can be used to perform the operation S220 described above, which will not be repeated here.

[0117] The embedding module 530 is used to embed the m-th data element from the M discrete metal concentration feature tensors into a feature representing the absence of metal concentration data in the metal concentration feature tensor, thereby obtaining the m-th target metal concentration feature tensor. In one embodiment, the embedding module 530 can be used to perform the operation S230 described above, which will not be repeated here.

[0118] Evaluation module 540 is used to obtain the embedding evaluation value corresponding to the m-th data element based on the above-mentioned extraction unit feature tensor, extraction volume feature tensor, injection volume feature tensor, time-series compression feature vector, and m-th target metal concentration feature tensor. In one embodiment, evaluation module 540 can be used to perform the operation S240 described above, which will not be repeated here.

[0119] The hidden data determination module 550 is used to input the multiple intrinsic mode data sequences and the gated recurrent neural network in the residual input prediction model to obtain multiple hidden state data. In one embodiment, the hidden data determination module 550 can be used to perform the operation S250 described above, which will not be repeated here.

[0120] The completion module 550 is used to determine, based on the M embedded evaluation values, the location where the missing metal concentration data exists, and then complete the target metal concentration data accordingly. In one embodiment, the completion module 550 can be used to perform the operation S250 described above, which will not be repeated here.

[0121] According to embodiments of this application, any multiple modules among the extraction module 510, compression module 520, embedding module 530, evaluation module 540, and completion module 550 can be merged into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of this application, at least one of the extraction module 510, compression module 520, embedding module 530, evaluation module 540, and completion module 550 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or implemented in hardware or firmware by any other reasonable means of integrating or packaging the circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the extraction module 510, compression module 520, embedding module 530, evaluation module 540, and completion module 550 may be implemented at least partially as a computer program module that can perform corresponding functions when the computer program module is run.

[0122] Based on the above-mentioned training method for the concentration completion model, this application also provides a training device for the concentration completion model. The following will combine... Figure 6 The device is described in detail.

[0123] Figure 6 A structural block diagram of a training apparatus for a concentration completion model according to an embodiment of this application is shown.

[0124] like Figure 6 As shown, the training device 600 for the concentration completion model in this embodiment includes a sample extraction module 610, a sample compression module 620, a sample embedding module 630, a sample evaluation module 640, and an adjustment module 650.

[0125] The sample extraction module 610 is used to extract features from the sample leachate concentration data to obtain a sample leachate concentration feature tensor. This sample leachate concentration feature tensor includes a sample injection unit feature tensor, a sample extraction volume feature tensor, a sample injection volume feature tensor, a sample timestamp feature tensor, a sample metal concentration feature tensor, and a sample discrete metal concentration feature tensor. The sample metal concentration feature tensor has features characterizing the presence of missing sample metal concentration data. The sample discrete metal concentration feature tensor includes K sample discrete metal concentration data points, which include sample label discrete metal concentration data corresponding to the missing sample metal concentration data points and multiple sample candidate discrete metal concentration data points. In one embodiment, the sample extraction module 610 can be used to perform the operation S410 described above, which will not be repeated here.

[0126] The sample compression module 620 is used to perform global temporal compression processing on the above-mentioned sample timestamp feature tensor to obtain a sample temporal compressed feature vector. In one embodiment, the sample compression module 620 can be used to perform the operation S420 described above, which will not be repeated here.

[0127] The sample embedding module 630 is used to embed the k-th sample data element in the aforementioned K sample discrete metal concentration feature tensors into the feature representing the missing sample metal concentration data in the aforementioned sample metal concentration feature tensors, thereby obtaining the k-th sample target metal concentration feature tensor. In one embodiment, the sample embedding module 630 can be used to perform the operation S430 described above, which will not be repeated here.

[0128] The sample evaluation module 640 is used to obtain the sample embedding evaluation value corresponding to the m-th data element based on the sample extraction unit feature tensor, the sample extraction volume feature tensor, the sample injection volume feature tensor, the sample time-series compression feature vector, and the target metal concentration feature tensor of the k-th sample. In one embodiment, the sample evaluation module 640 can be used to perform the operation S440 described above, which will not be repeated here.

[0129] The adjustment module 650 is used to obtain a loss value based on the target loss function and the M embedding evaluation values ​​mentioned above, and to adjust the parameters of the concentration completion model based on the loss value to obtain a trained concentration completion model. In one embodiment, the adjustment module 650 can be used to perform the operation S450 described above, which will not be repeated here.

[0130] According to embodiments of this application, any multiple modules among the sample extraction module 610, sample compression module 620, sample embedding module 630, sample evaluation module 640, and adjustment module 650 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of this application, at least one of the sample extraction module 610, sample compression module 620, sample embedding module 630, sample evaluation module 640, and adjustment module 650 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the sample extraction module 610, sample compression module 620, sample embedding module 630, sample evaluation module 640, and adjustment module 650 may be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0131] Figure 7 A block diagram of an electronic device suitable for implementing a concentration completion method and a training method for a concentration completion model according to an embodiment of this application is shown.

[0132] like Figure 7 As shown, an electronic device 700 according to an embodiment of this application includes a processor 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage portion 708 into a random access memory (RAM) 703. The processor 701 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 701 may also include onboard memory for caching purposes. The processor 701 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this application.

[0133] RAM 703 stores various programs and data required for the operation of electronic device 700. Processor 701, ROM 702, and RAM 703 are interconnected via bus 704. Processor 701 executes various operations of the method flow according to embodiments of this application by executing programs in ROM 702 and / or RAM 703. It should be noted that the programs may also be stored in one or more memories other than ROM 702 and RAM 703. Processor 701 may also execute various operations of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0134] According to embodiments of this application, the electronic device 700 may further include an input / output (I / O) interface 705, which is also connected to a bus 704. The electronic device 700 may also include one or more of the following components connected to the input / output (I / O) interface 705: an input section 706 including a keyboard, mouse, etc.; an output section 707 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 708 including a hard disk, etc.; and a communication section 709 including a network interface card such as a LAN card, modem, etc. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the input / output (I / O) interface 705 as needed. A removable medium 711, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 710 as needed so that computer programs read from it can be installed into the storage section 708 as needed.

[0135] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0136] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can 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. For example, according to embodiments of this application, the computer-readable storage medium may include ROM 702 and / or RAM 703 and / or one or more memories other than ROM 702 and RAM 703 described above.

[0137] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to enable the computer system to implement the concentration completion method provided in the embodiments of this application.

[0138] When the computer program is executed by the processor 701, it performs the functions defined in the system / apparatus of this application embodiment. According to the embodiments of this application, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0139] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 709, and / or installed from a removable medium 711. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0140] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 709, and / or installed from the removable medium 711. When the computer program is executed by the processor 701, it performs the functions defined in the system of this application embodiment. According to the embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0141] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0142] 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 application. 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 a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may 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.

[0143] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

[0144] The embodiments of this application have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of this application. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Without departing from the scope of this application, those skilled in the art can make various substitutions and modifications, all of which should fall within the scope of this application.

Claims

1. A concentration compensation method, characterized in that, The method includes: Feature extraction is performed on the leachate concentration data acquired within a preset time period to obtain a leachate concentration feature tensor. The leachate concentration feature tensor includes a pumping unit feature tensor, a pumping volume feature tensor, a pumping volume feature tensor, a timestamp feature tensor, a metal concentration feature tensor, and a discrete metal concentration feature tensor. The metal concentration feature tensor contains features that characterize the absence of metal concentration data. The discrete metal concentration feature tensor includes M discrete metal concentration data. The timestamp feature tensor is subjected to global temporal compression to obtain a temporal compressed feature vector; For the m-th data element in the M discrete metal concentration feature tensors, The m-th data element is embedded into the metal concentration feature tensor to represent the feature where the metal concentration data is missing, thus obtaining the m-th target metal concentration feature tensor. Based on the feature tensor of the extraction unit, the feature tensor of the extraction volume, the feature tensor of the injection volume, the temporal compression feature vector, and the feature tensor of the m-th target metal concentration, the embedding evaluation value corresponding to the m-th data element is obtained; Based on the M embedded evaluation values, the target metal concentration data is determined after filling in the missing metal concentration data at the location where the metal concentration data is missing.

2. The method according to claim 1, characterized in that, The step of performing global temporal compression processing on the timestamp feature tensor to obtain a temporal compressed feature vector includes: The timestamp feature tensor is partitioned into N sub-timestamp feature tensors. For the nth of the N sub-timestamp feature tensors; Based on the preset nth discrete cosine transform basis function corresponding to the nth sub-time stamp feature tensor, the nth sub-time stamp feature tensor is subjected to discrete cosine transform to obtain the nth sub-time series compressed feature vector. The N sub-temporal compression feature vectors are concatenated to obtain the temporal compression feature vector.

3. The method according to claim 1, characterized in that, The step of obtaining the embedding evaluation value corresponding to the m-th data element based on the feature tensor of the extraction unit, the feature tensor of the extraction volume, the feature tensor of the injection volume, the temporal compression feature vector, and the feature tensor of the m-th target metal concentration includes: Attention calculation is performed on the timestamp feature tensor based on the fluid extraction volume feature tensor, the fluid injection volume feature tensor, and the time-series compression feature vector to obtain the processed timestamp feature tensor. A matrix multiplication operation is performed on the injection unit feature tensor, the injection volume feature tensor, the injection volume feature tensor, and the processed timestamp feature tensor to obtain the injection time-series coupling tensor. Perform a dot product operation on the injection timing coupling tensor and the m-th target metal concentration feature tensor to obtain the embedding evaluation value corresponding to the m-th target metal concentration feature tensor.

4. The method according to claim 3, characterized in that, The process of performing attention calculation on the timestamp feature tensor based on the fluid extraction volume feature tensor, the fluid injection volume feature tensor, and the time-series compression feature vector to obtain the processed timestamp feature tensor includes: A matrix multiplication operation is performed on the extraction volume feature tensor, the injection volume feature tensor, and the time-series compressed feature vector to obtain a fused tensor; The fusion tensor is mapped to obtain attention features, and the timestamp feature tensor is processed using the attention features to obtain the processed timestamp feature tensor.

5. The method according to claim 1, characterized in that, The leachate concentration data is obtained by selectively leaching useful metal components from the ore after injecting chemical leaching solution into the target mining area through an injection unit.

6. A training method for a concentration completion model, characterized in that, The method includes: Feature extraction is performed on the sample leachate concentration data to obtain a sample leachate concentration feature tensor. This sample leachate concentration feature tensor includes a sample extraction / injection unit feature tensor, a sample extraction volume feature tensor, a sample injection volume feature tensor, a sample timestamp feature tensor, a sample metal concentration feature tensor, and a sample discrete metal concentration feature tensor. The sample metal concentration feature tensor has features characterizing the presence of missing sample metal concentration data. The sample discrete metal concentration feature tensor includes K sample discrete metal concentration data points, which include sample label discrete metal concentration data corresponding to the missing sample metal concentration data points and multiple sample candidate discrete metal concentration data points. The sample timestamp feature tensor is subjected to global temporal compression to obtain the sample temporal compressed feature vector; For the k-th sample data element in the discrete metal concentration feature tensor of the K samples, The k-th sample data element is embedded into the sample metal concentration feature tensor to represent the feature where the sample metal concentration data is missing, thus obtaining the k-th sample target metal concentration feature tensor; Based on the sample injection unit feature tensor, the sample extraction volume feature tensor, the sample injection volume feature tensor, the sample temporal compression feature vector, and the target metal concentration feature tensor of the kth sample, the sample embedding evaluation value corresponding to the mth data element is obtained. Based on the target loss function, a loss value is obtained according to the M embedding evaluation values, and the parameters of the concentration completion model are adjusted based on the loss value to obtain a trained concentration completion model.

7. The method according to claim 6, characterized in that, The M embedding evaluation values ​​include label embedding evaluation values ​​corresponding to the discrete metal concentration data of the sample labels and candidate embedding evaluation values ​​corresponding to the multiple candidate discrete metal concentration data of the samples, respectively. The step of obtaining a loss value based on the M embedding evaluation values ​​according to the target loss function includes: The loss value is obtained based on the target loss function, according to the label embedding evaluation value and the candidate embedding evaluation value.

8. The method according to claim 7, characterized in that, Before performing feature extraction on the sample leachate concentration data, the method further includes: The sample candidate metal concentration data is pre-filled to obtain the sample leachate concentration data.

9. The method according to claim 8, characterized in that, The process of pre-filling the candidate metal concentration data of the sample to obtain the concentration data of the sample leachate includes: In the case where there are multiple consecutive missing locations in the sample candidate metal concentration data, a first target filling value is determined based on a preset number of sample candidate metal concentration data before the first consecutive missing location among the multiple consecutive missing locations. The first target fill value is used to fill the multiple consecutive missing areas.

10. The method according to claim 8, characterized in that, The process of pre-filling the candidate metal concentration data of the sample to obtain the concentration data of the sample leachate includes: In the case where there are multiple discontinuous missing points in the candidate metal concentration data of the sample, for each of the multiple discontinuous missing points... The second target filling value is determined based on the sample candidate metal concentration data of the preset number of samples before the non-continuous missing sites; The non-continuous missing values ​​are filled using the second target filling value.

Images