AI-based method and system for determining the production plan of flexible DC support capacitors

Abstract

The method and system for determining the production plan of flexible DC support capacitors based on AI provided by the present invention relate to the field of artificial intelligence technology. In the present invention, first, the capacitor performance semantic vector corresponding to the target capacitor performance data is deeply mined, and the capacitor performance depth vector is output; secondly, the production plan semantic vector corresponding to the target capacitor performance data is deeply mined, and the production plan depth vector is output; then, the capacitor performance depth vector and the production plan depth vector are merged and mined, and the production plan mining vector is output; finally, the target capacitor production plan corresponding to the target capacitor performance data is generated according to the production plan mining vector. Based on the above content, the problem of relatively low reliability of capacitor production plan determination existing in the prior art can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an AI-based method and system for determining a production plan for a flexible DC support capacitor. Background Art

[0002] Flexible DC Transmission (FDC) technology has become one of the important innovations in the current global energy transmission field. It has shown significant advantages in the process of transmitting large-scale clean energy, especially wind energy, solar energy, energy storage and pumped storage. With the acceleration of the transformation of the global energy structure, the proportion of clean energy continues to rise, and the traditional AC power grid is facing more and more challenges. Flexible DC transmission technology has become an important part of the future energy network due to its unique advantages in efficient and stable transmission of large-scale clean energy.

[0003] Secondly, in the flexible DC transmission system, the support capacitor is one of the core components of the converter valve and a key hub device for energy interaction. The functions of the support capacitor include:

[0004] Stabilize converter valve operation: In a flexible DC transmission system, the main task of the converter valve is to realize the conversion and regulation of DC current, while the support capacitor provides the necessary power reserve and stability support. By regulating the current of the converter valve, the support capacitor helps to ensure the smooth operation of voltage and current during the power conversion process, thereby improving the stability of the entire system;

[0005] Energy interaction and energy storage function: As an energy storage element, the support capacitor can store and release energy during the power transmission process, which not only helps to balance the power supply and demand, but also effectively cope with the voltage fluctuations in the system and ensure the efficient and stable operation of the transmission system. In the process of accessing renewable energy sources with large fluctuations such as wind power and photovoltaic power, the support capacitor can alleviate the impact of power fluctuations on converter valves and other equipment;

[0006] Improve the reliability of converter valves: The operational reliability of supporting capacitors directly affects the stability and safety of the entire converter valve and flexible DC transmission system. If supporting capacitors malfunction or fail, the converter valves may not be able to effectively convert electrical energy, thus affecting the stability of power transmission and even causing system shutdown. Therefore, the design and maintenance of supporting capacitors are crucial to ensure their reliable operation under long-term and high-load conditions.

[0007] Based on this, it is necessary to produce supporting capacitors whose performance meets actual needs. However, in the prior art, staff generally determine the production plan based on relevant experience. In this way, since the staff are restricted by the limitations of experience and knowledge, the reliability of the determined production plan is relatively low, and the actual performance of the produced supporting capacitors is difficult to meet the actual needs of users. Summary of the invention

[0008] In view of this, the purpose of the present invention is to provide a method and system for determining a production plan for flexible DC support capacitors based on AI, so as to improve the problem of relatively low reliability in determining the capacitor production plan in the prior art.

[0009] To achieve the above object, the present invention adopts the following technical solution:

[0010] A method for determining a production plan of a flexible DC support capacitor based on AI, comprising:

[0011] Deeply mine the capacitor performance semantic vector corresponding to the target capacitor performance data, and output the capacitor performance depth vector, wherein the target capacitor performance data is used to describe at least one performance of the target flexible DC support capacitor to be produced, and the performance includes capacitance value, maximum operating voltage, equivalent series resistance and filtering performance;

[0012] Deeply mining the production plan semantic vector corresponding to the target capacitor performance data, and outputting a production plan deep vector, wherein the production plan semantic vector is formed by semantically mining the production plans corresponding to capacitor performance data similar to the target capacitor performance data;

[0013] Performing combined mining based on the capacitor performance depth vector and the production plan depth vector, and outputting a production plan mining vector;

[0014] Based on the production plan mining vector, a target capacitor production plan corresponding to the target capacitor performance data is generated, wherein the target capacitor production plan is at least used to reflect at least one process parameter of the target flexible DC support capacitor in the production process, and the process parameters include electrode-related parameters, insulation layer-related parameters, welding process parameters and electrolyte-related parameters.

[0015] In a preferred embodiment of the present invention, in the above-mentioned AI-based flexible DC support capacitor production plan determination method, the step of merging and mining the capacitor performance depth vector and the production plan depth vector to output the production plan mining vector includes:

[0016] Perform at least one stage of combined mining based on the capacitor performance depth vector and the production plan depth vector, and output the capacitor performance mining vector corresponding to the combined mining of the at least one stage, wherein in the combined mining of each stage, perform focused feature mining on the capacitor performance depth vector and the production plan depth vector loaded in the current stage to obtain the production plan depth vector mined in the current stage, and perform focused feature mining on the production plan depth vector mined in the current stage and the capacitor performance depth vector loaded in the current stage to obtain the capacitor performance depth vector mined in the current stage, when the current stage does not belong to the last stage, the production plan depth vector and the capacitor performance depth vector mined in the current stage are used as the capacitor performance depth vector and the production plan depth vector loaded in the next stage, and when the current stage belongs to the last stage, the capacitor performance depth vector mined in the current stage is used as the capacitor performance mining vector corresponding to the combined mining of the at least one stage;

[0017] A production solution mining vector corresponding to the target capacitor performance data is determined based on the capacitor performance mining vector.

[0018] In a preferred embodiment of the present invention, in the above-mentioned method for determining a flexible DC support capacitor production plan based on AI, the at least one stage includes multiple stages;

[0019] The step of performing at least one stage of combined mining according to the capacitor performance depth vector and the production plan depth vector, and outputting the capacitor performance mining vector corresponding to the combined mining of the at least one stage includes:

[0020] For the combined mining of the first stage among the multiple stages, focused feature mining with a first direction is performed based on the capacitor performance depth vector and the production solution depth vector loaded in the first stage, and the production solution depth vector mined in the first stage is output, wherein the capacitor performance depth vector and the production solution depth vector loaded in the first stage are the initial capacitor performance depth vector and the production solution depth vector, and the focused feature mining with the first direction means that the semantic information of the capacitor performance depth vector loaded in the first stage is integrated into the semantic information of the production solution depth vector loaded in the first stage;

[0021] For the combined mining of the first stage among the multiple stages, focusing feature mining with a second direction is performed based on the production solution depth vector mined in the first stage and the capacitor performance depth vector loaded in the first stage, and the capacitor performance depth vector mined in the first stage is output, wherein the focusing feature mining with the second direction refers to fusing the semantic information of the production solution depth vector mined in the first stage into the semantic information of the capacitor performance depth vector loaded in the first stage;

[0022] For each stage other than the second stage in the multiple stages, performing merge mining of the current stage based on the production solution depth vector and the capacitor performance depth vector mined in the previous stage, and outputting the production solution depth vector and the capacitor performance depth vector mined in the current stage;

[0023] After outputting the production solution depth vector and the capacitor performance depth vector mined in the last stage, the capacitor performance depth vector mined in the last stage is used as the capacitor performance mining vector corresponding to the combined mining of the at least one stage.

[0024] In a preferred embodiment of the present invention, in the above-mentioned AI-based flexible DC support capacitor production plan determination method, the combined mining of the first stage of the multiple stages, based on the capacitor performance depth vector and the production plan depth vector loaded in the first stage, performs focused feature mining with a first direction, and outputs the production plan depth vector mined in the first stage, including:

[0025] For the combined mining of the first stage among the multiple stages, transposing the production plan depth vector loaded into the first stage, and outputting the production plan transposed vector corresponding to the first stage;

[0026] Associatively aggregate the capacitor performance depth vector loaded into the first stage, the production plan depth vector loaded into the first stage, and the production plan transposed vector corresponding to the first stage, and output the production plan aggregate vector corresponding to the first stage;

[0027] Based on the production plan depth vector loaded in the first stage, a first gradient optimization is performed on the production plan aggregation vector corresponding to the first stage, and a production plan gradient optimization vector corresponding to the first stage is output;

[0028] Perform at least one vector linear processing on the gradient optimization vector of the production plan corresponding to the first stage, and output the linear vector of the production plan corresponding to the first stage;

[0029] Based on the production plan gradient optimization vector corresponding to the first stage, the production plan linear vector corresponding to the first stage is subjected to a second gradient optimization, and the production plan depth vector mined in the first stage is output.

[0030] In a preferred embodiment of the present invention, in the above-mentioned AI-based flexible DC support capacitor production plan determination method, the combined mining of the first stage of the multiple stages, based on the production plan depth vector mined in the first stage and the capacitor performance depth vector loaded in the first stage, performs focused feature mining with a second direction, and outputs the capacitor performance depth vector mined in the first stage, including:

[0031] For the combined mining of the first stage among the multiple stages, transposing the capacitor performance depth vector loaded into the first stage, and outputting the capacitor performance transposed vector corresponding to the first stage;

[0032] Associatively aggregate the production plan depth vector mined in the first stage, the capacitor performance depth vector loaded in the first stage, and the capacitor performance transposed vector corresponding to the first stage, and output the capacitor performance aggregate vector corresponding to the first stage;

[0033] Based on the capacitor performance depth vector loaded in the first stage, performing a first gradient optimization on the capacitor performance aggregation vector corresponding to the first stage, and outputting the capacitor performance gradient optimization vector corresponding to the first stage;

[0034] Perform at least one vector linear processing on the capacitor performance gradient optimization vector corresponding to the first stage, and output the capacitor performance linear vector corresponding to the first stage;

[0035] Based on the capacitor performance gradient optimization vector corresponding to the first stage, the capacitor performance linear vector corresponding to the first stage is subjected to a second gradient optimization, and the capacitor performance depth vector mined in the first stage is output.

[0036] In a preferred embodiment of the present invention, in the above-mentioned AI-based flexible DC support capacitor production plan determination method, the step of determining the production plan mining vector corresponding to the target capacitor performance data based on the capacitor performance mining vector includes:

[0037] Transposing the capacitor performance mining vector, and outputting a transposed capacitor performance mining vector corresponding to the capacitor performance mining vector;

[0038] Performing association aggregation based on the capacitor performance mining vector and the transposed capacitor performance mining vector, and outputting a capacitor performance self-aggregation vector corresponding to the capacitor performance mining vector;

[0039] Perform at least one vector linear processing on the capacitor performance self-aggregation vector, and output a linear capacitor performance vector corresponding to the capacitor performance self-aggregation vector;

[0040] Based on the capacitor performance self-aggregation vector, the linear capacitor performance vector is gradient optimized, and a production solution mining vector corresponding to the target capacitor performance data is output.

[0041] In a preferred embodiment of the present invention, in the above-mentioned method for determining a production plan of a flexible DC support capacitor based on AI, the method for determining a production plan of a flexible DC support capacitor based on AI further includes:

[0042] Determine first capacitor performance data and second capacitor performance data for training, and determine a training production plan, wherein the capacitor performance data corresponding to the training production plan is similar to the first capacitor performance data, and the capacitor performance data corresponding to the training production plan is not similar to the second capacitor performance data;

[0043] Perform semantic mining on the first capacitor performance data, the training production scheme, and the second capacitor performance data, respectively, and output a first training performance semantic vector corresponding to the first capacitor performance data, a training scheme semantic vector corresponding to the training production scheme, and a second training performance semantic vector corresponding to the second capacitor performance data;

[0044] Determine the network using the candidate production scheme, perform deep mining on the first training performance semantic vector, the training scheme semantic vector, and the second training performance semantic vector, respectively, and output a first training performance depth vector, a training scheme depth vector, and a second training performance depth vector;

[0045] Determine a network using the candidate production scheme, perform at least one stage of merged mining based on the first training performance depth vector and the training scheme depth vector, output a first training performance mining vector corresponding to the at least one stage of merged mining, and determine a first training scheme mining vector based on the first training performance mining vector;

[0046] Determine a network using the candidate production scheme, perform at least one stage of merged mining based on the second training performance depth vector and the training scheme depth vector, output a second training performance mining vector corresponding to the at least one stage of merged mining, and determine a second training scheme mining vector based on the second training performance mining vector;

[0047] Based on the first training scheme mining vector and the second training scheme mining vector, the network parameters of the candidate production scheme determination network are updated to form a target production scheme determination network, wherein the target production scheme determination network is used to generate a target capacitor production scheme corresponding to the target capacitor performance data.

[0048] In a preferred embodiment of the present invention, in the above-mentioned AI-based flexible DC support capacitor production plan determination method, the step of updating the network parameters of the candidate production plan determination network according to the first training plan mining vector and the second training plan mining vector to form a target production plan determination network includes:

[0049] Using the candidate production scheme determination network, performing scheme generation processing according to the first training scheme vector and the second training scheme vector, outputting the corresponding first capacitor production scheme and second capacitor production scheme, and respectively calculating the differences between the first capacitor production scheme and the second capacitor production scheme and the training production scheme to obtain the corresponding first production scheme difference and the corresponding second production scheme difference;

[0050] Based on the first production plan difference and the second production plan difference, the network parameters of the candidate production plan determination network are updated to form a target production plan determination network.

[0051] In a preferred embodiment of the present invention, in the above-mentioned AI-based flexible DC support capacitor production plan determination method, the step of updating the network parameters of the candidate production plan determination network based on the first production plan difference and the second production plan difference to form a target production plan determination network includes:

[0052] Calculating the difference between the first training scheme mining vector and the second training scheme mining vector to obtain a corresponding mining vector difference;

[0053] respectively calculating the differences between the first training scheme mining vector and the second training scheme mining vector and the training scheme semantic vector to obtain corresponding first scheme semantic differences and second scheme semantic differences;

[0054] Based on the first production plan difference, the second production plan difference, the mining vector difference, the first plan semantic difference and the second plan semantic difference, the network parameters of the candidate production plan determination network are updated to form a target production plan determination network.

[0055] On the basis of the above, the present invention also provides an AI-based flexible DC support capacitor production plan determination system, including:

[0056] Memory for storing computer programs;

[0057] A processor connected to the memory is used to execute a computer program stored in the memory to implement the above-mentioned AI-based method for determining a production plan for flexible DC support capacitors.

[0058] The method and system for determining the production plan of flexible DC support capacitors based on AI provided by the present invention firstly conducts deep mining on the semantic vector of capacitor performance corresponding to the target capacitor performance data, and outputs the capacitor performance depth vector; secondly, conducts deep mining on the semantic vector of the production plan corresponding to the target capacitor performance data, and outputs the production plan depth vector; then, conducts combined mining based on the capacitor performance depth vector and the production plan depth vector, and outputs the production plan mining vector; finally, generates the target capacitor production plan corresponding to the target capacitor performance data based on the production plan mining vector. Based on the above content, since the production plan mining vector as the basis for generating the target capacitor production plan includes not only the semantic information of the target capacitor performance data, but also the semantic information of the production plan corresponding to the capacitor performance data similar to the target capacitor performance data, that is, the semantic information of the target capacitor performance data can be reinforced based on the semantic information of the similar production plan, thereby effectively improving the semantic representation accuracy of the production plan mining vector, and then ensuring the reliability of the determined target capacitor production plan, therefore, the problem of relatively low reliability of capacitor production plan determination existing in the prior art can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings.

[0060] Figure 1 A structural block diagram of the AI-based flexible DC support capacitor production plan determination system provided in an embodiment of the present invention.

[0061] Figure 2 A schematic flow chart of a method for determining a production plan for flexible DC support capacitors based on AI provided in an embodiment of the present invention.

[0062] Figure 3 A schematic diagram of a multi-stage merge mining provided by an embodiment of the present invention.

[0063] Figure 4 A schematic diagram of focused feature mining in a first direction provided by an embodiment of the present invention.

[0064] Figure 5A schematic diagram of another multi-stage merge mining provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0066] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0067] like Figure 1 As shown, an embodiment of the present invention provides an AI-based flexible DC support capacitor production plan determination system. The AI-based flexible DC support capacitor production plan determination system may include a memory and a processor.

[0068] In detail, the memory and the processor are electrically connected directly or indirectly to achieve data transmission or interaction. For example, the memory and the processor can be electrically connected through one or more communication buses or signal lines. The processor is used to execute an executable computer program stored in the memory to implement the method for determining the production plan of flexible DC support capacitors based on AI provided in an embodiment of the present invention.

[0069] Optionally, the memory may be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), etc.

[0070] Furthermore, the processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a system on chip (SoC), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0071] Understandably, Figure 1 The structure shown is for illustration only. The AI-based flexible DC support capacitor production plan determination system may also include Figure 1 More or fewer components as shown, or with Figure 1 The different configurations shown, for example, may also include a communication unit for information exchange with other devices. In addition, the AI-based flexible DC support capacitor production solution determination system may be a server or a server cluster consisting of multiple servers.

[0072] Combination Figure 2 The embodiment of the present invention also provides an AI-based flexible DC support capacitor production plan determination method that can be applied to the above-mentioned AI-based flexible DC support capacitor production plan determination system. Among them, the method steps defined in the process related to the AI-based flexible DC support capacitor production plan determination method can be implemented by the AI-based flexible DC support capacitor production plan determination system (in the following embodiments, it can be referred to as the production plan determination system). The following will be Figure 2 The specific process shown is explained in detail.

[0073] Step S110 , deeply mining the capacitor performance semantic vector corresponding to the target capacitor performance data, and outputting a capacitor performance depth vector.

[0074] In an embodiment of the present invention, the production plan determination system can perform deep mining on the capacitor performance semantic vector corresponding to the target capacitor performance data and output a capacitor performance depth vector; that is, after obtaining the capacitor performance semantic vector corresponding to the target capacitor performance data, the capacitor performance semantic vector can be further deeply mined, so that a higher-level feature with stronger expression ability can be obtained, that is, a capacitor performance depth vector. Among them, the target capacitor performance data is used to describe at least one performance of the target flexible DC support capacitor to be produced, and the performance includes capacitance value, maximum operating voltage, equivalent series resistance and filtering performance. That is to say, the target capacitor performance data includes at least one of the capacitance value, maximum operating voltage, equivalent series resistance and filtering performance, and in other embodiments, other performances may also be included, which can be selected according to actual needs.

[0075] Step S120, deeply mining the production solution semantic vector corresponding to the target capacitor performance data, and outputting the production solution depth vector.

[0076] In an embodiment of the present invention, the production plan determination system can perform deep mining on the production plan semantic vector corresponding to the target capacitor performance data and output a production plan depth vector. That is to say, after obtaining the production plan semantic vector corresponding to the target capacitor performance data, the production plan semantic vector can be further deeply mined, so that a higher-level feature with stronger expression ability can be obtained, that is, a production plan depth vector. Among them, the production plan semantic vector is formed by semantically mining the production plan corresponding to the capacitor performance data similar to the target capacitor performance data (such as the difference between the parameters of the performance included is less than the threshold, which can be configured according to actual needs, and the corresponding threshold values ​​corresponding to different performances can also be different, that is, the actual performance represented is basically the same).

[0077] Step S130 , performing combined mining based on the capacitor performance depth vector and the production solution depth vector, and outputting a production solution mining vector.

[0078] In an embodiment of the present invention, after obtaining the capacitor performance depth vector and the production solution depth vector, the production solution determination system can perform merged mining based on the capacitor performance depth vector and the production solution depth vector, and output a production solution mining vector. In other words, the production solution mining vector can carry semantic information of two dimensions, the capacitor performance depth vector and the production solution depth vector, to achieve reinforcement of semantic information.

[0079] Step S140: generating a target capacitor production plan corresponding to the target capacitor performance data based on the production plan mining vector.

[0080] In an embodiment of the present invention, after obtaining the production plan mining vector, the production plan determination system can generate a target capacitor production plan corresponding to the target capacitor performance data based on the production plan mining vector. Among them, the target capacitor production plan is at least used to reflect at least one process parameter of the target flexible DC support capacitor in the production process, and the process parameters include electrode-related parameters (such as the surface area of ​​the electrode and the stacking method of each layer, etc.), insulation layer-related parameters (such as the thickness and material of the insulation layer), welding process parameters (welding method, temperature, time, etc.) and electrolyte-related parameters (such as the form and material of the electrolyte, etc.). In other embodiments, other process parameters may also be included, such as the thickness of the film, the ratio of the flexible plastic layer, the packaging material, the thermal conductive material, etc., which can be configured according to actual needs.

[0081] Based on the above content, since the production plan mining vector serving as the basis for generating the target capacitor production plan not only includes the semantic information of the target capacitor performance data, but also includes the semantic information of the production plan corresponding to the capacitor performance data similar to the target capacitor performance data, that is to say, the semantic information of the target capacitor performance data can be supplemented based on the semantic information of similar production plans, thereby effectively improving the semantic representation accuracy of the production plan mining vector, and then ensuring the reliability of the determined target capacitor production plan. Therefore, the problem of relatively low reliability of capacitor production plan determination existing in the prior art can be improved.

[0082] In the first part, it should be noted that the specific method of deep mining the capacitor performance semantic vector corresponding to the target capacitor performance data is not limited. For example, in an alternative implementation, in order to ensure that the mined capacitor performance deep vector has better feature expression ability, the above step S110 may further include:

[0083] First, the target capacitor performance data can be semantically mined to obtain a corresponding capacitor performance semantic vector, wherein the semantic mining can be obtained by performing word embedding processing through a trained word embedding model. For example, each word in the target capacitor performance data can be word embedded to obtain a word embedding vector. Then, each word embedding vector can be mean-calculated to obtain a capacitor semantic vector. For example, for "filtering performance is to filter out harmonics of 50 kHz frequency", each word embedding vector may include: "filtering": [0.1, -0.2, 0.3, 0.5, 0.2, -0.1, 0.3, -0.4, 0.1, 0.4, -0.3, ......]; "performance": [0.3, 0.1, -0.2, 0.4, 0.5, -0.3, 0.2, 0.1, 0.3, -0.2, 0.5, ......]; "for": [0.0, 0.2, -0.1, , 0.1, 0.3, 0.2, -0.2, -0.1, 0.0, 0.4, 0.2, ......]; "Filter": [0.4, -0.3, 0.1, 0.2, 0.1, 0.5, 0.4,-0.1, 0.3, -0.2, 0.0, ......]; "50": [0.0, 0.0, 0.1, -0.1, 0.0, 0.1, -0.3, 0.2,0.4, 0.1, -0.2, ......]; "kHz": [0.5, 0.4, 0.1, -0.2, 0.3, 0.2, 0.3, -0.1, -0.2, 0.4, 0.1, ......]; "Frequency": [0.2, , 0.1, 0.4, 0.3, -0.1, 0.5, -0.2, 0.1, -0.1, 0.3, -0.4, ......]; "of": [-0.1, 0.0, 0.1, 0.2, 0.3, -0.4, 0.1, 0.0, 0.2,0.1, 0.3, ......]; "harmonic": [0.3, 0.2, 0.4, 0.1, 0.0, -0.3, 0.5, -0.2, 0.3, 0.1,-0.2, ......];

[0084] Secondly, the capacitor performance semantic vector may be subjected to at least one self-attention process to achieve deep mining and obtain a corresponding capacitor performance depth vector.

[0085] In the second part, it should be noted that for step S120, the specific method of deep mining the production solution semantic vector corresponding to the target capacitor performance data is not limited. For example, in an alternative implementation, in order to ensure that the mined production solution deep vector has better feature expression ability, the above step S120 may further include:

[0086] First, a production plan corresponding to capacitor performance data similar to the target capacitor performance data can be obtained (the production plan can be a plan determined based on other methods, or a plan determined historically using the method provided by the embodiment of the present invention). Then, semantic mining can be performed on the production plan (as described above) to obtain a corresponding production plan semantic vector. For example, for "using polyester as an insulating material with a thickness of 1 mm", each word embedding vector may include: "use": [0.23, -0.45, 0.12, 0.89, -0.34, 0.50, -0.21, 0.76, -0.55, 0.62, -0.13, ......]; "polyester": [0.18, -0.34, 0.15, 0.90, -0.12, 0.67, -0.44, 0.80, -0.56, 0.39, -0.05, ......]; "as": [0.26, -0.33, , 0.22, 0.70, -0.29, 0.45, -0.31, 0.78, -0.58, 0.53, -0.11, ......]; "Insulation": [0.21, -0.40, 0.10, 0.85, -0.18, 0.55, -0.25, 0.74, -0.51, 0.49, -0.07, ......]; "Material": [0.19, -0.36,0.16, 0.78, -0.22, 0.60, -0.30, 0.72, -0.54, 0.51, -0.09, ......]; "Thickness": [0.22, -0.38, 0.12, 0.80, -0.20, 0.59, 0.01, 0.04, 0.03, 0.10, -0.01, 0.08, -0.06, 0.12, ......]; "1": [0.05, -0.10, 0.00, 0.02, -0.04, 0.08, -0.02, 0.10, -0.07,0.09, 0.01, ......]; "mm": [0.07, -0.15, 0.01, 0.04, -0.03, 0.10, -0.01, 0.08, -0.06, 0.12, 0.02, ......];

[0087] Secondly, the production plan semantic vector may be subjected to at least one self-attention process to achieve deep mining and obtain a corresponding production plan depth vector.

[0088] In the third part, it should be noted that for step S130, the specific method of merging and mining the capacitor performance depth vector and the production plan depth vector is not limited. For example, in an alternative embodiment, in order to ensure that the mined production plan mining vector has reliable feature expression capabilities, the above-mentioned step S130 may further include step S131 and step S132. The specific contents of each step are described as follows.

[0089] Step S131, performing at least one stage of combined mining according to the capacitor performance depth vector and the production plan depth vector, and outputting a capacitor performance mining vector corresponding to the at least one stage of combined mining.

[0090] In the embodiment of the present invention, at least one stage of combined mining may be performed based on the capacitor performance depth vector and the production plan depth vector, and a capacitor performance mining vector corresponding to the at least one stage of combined mining may be output. Among them, in the combined mining of each stage, the capacitor performance depth vector and the production plan depth vector loaded in the current stage are subjected to focused feature mining to obtain the production plan depth vector mined in the current stage, and the production plan depth vector mined in the current stage and the capacitor performance depth vector loaded in the current stage are subjected to focused feature mining to obtain the capacitor performance depth vector mined in the current stage. When the current stage does not belong to the last stage, the production plan depth vector and the capacitor performance depth vector mined in the current stage are used as the capacitor performance depth vector and the production plan depth vector loaded in the next stage (for example, the production plan depth vector and the capacitor performance depth vector mined in the first stage can be used as the capacitor performance depth vector and the production plan depth vector loaded in the second stage, and for another example, the production plan depth vector and the capacitor performance depth vector mined in the second stage can be used as the capacitor performance depth vector and the production plan depth vector loaded in the third stage). When the current stage belongs to the last stage, the capacitor performance depth vector mined in the current stage is used as the capacitor performance mining vector corresponding to the combined mining of at least one stage.

[0091] Step S132: determining a production solution mining vector corresponding to the target capacitor performance data based on the capacitor performance mining vector.

[0092] In the embodiment of the present invention, after obtaining the capacitor performance mining vector, the production solution mining vector corresponding to the target capacitor performance data can be determined based on the capacitor performance mining vector. In other words, the capacitor performance mining vector in the performance semantic space can be converted into the solution semantic space to obtain the corresponding production solution mining vector, so as to reliably generate the target capacitor production solution in the subsequent steps.

[0093] It can be understood that, in the above step S131, the specific manner of performing at least one stage of combined mining based on the capacitor performance depth vector and the production plan depth vector is not limited. For example, in an alternative implementation, in order to mine more reliable semantic information so that the semantic representation ability of the obtained capacitor performance mining vector is better, the at least one stage includes multiple stages, combined with Figure 3 The above-mentioned step S131 may further include step S131a, step S131b, step S131c and step S131d, and the contents of each step are described as follows.

[0094] Step S131a, for the combined mining of the first stage among the multiple stages, focus feature mining with a first direction is performed based on the capacitor performance depth vector and the production solution depth vector loaded into the first stage, and the production solution depth vector mined in the first stage is output.

[0095] In an embodiment of the present invention, for the combined mining of the first stage of the multiple stages, focused feature mining with a first direction can be performed based on the capacitor performance depth vector and the production solution depth vector loaded in the first stage, and the production solution depth vector mined in the first stage can be output. Among them, the capacitor performance depth vector and the production solution depth vector loaded in the first stage are the initial capacitor performance depth vector (such as the output of step S110) and the production solution depth vector (such as the output of step S120), and the focused feature mining with a first direction means that the semantic information of the capacitor performance depth vector loaded in the first stage is integrated into the semantic information of the production solution depth vector loaded in the first stage.

[0096] Step S131b, for the combined mining of the first stage among the multiple stages, focus feature mining with a second direction is performed based on the production plan depth vector mined in the first stage and the capacitor performance depth vector loaded in the first stage, and the capacitor performance depth vector mined in the first stage is output.

[0097] In an embodiment of the present invention, for the combined mining of the first stage of the multiple stages, focused feature mining with a second direction can be performed based on the production solution depth vector mined in the first stage and the capacitor performance depth vector loaded in the first stage, and the capacitor performance depth vector mined in the first stage is output. Among them, the focused feature mining with a second direction refers to fusing the semantic information of the production solution depth vector mined in the first stage into the semantic information of the capacitor performance depth vector loaded in the first stage. In other words, the performance semantic information can be first fused into the solution semantic information to form richer solution semantic information, and then the richer solution semantic information can be fused into the performance semantic information to obtain richer performance semantic information, that is, the capacitor performance depth vector mined in the first stage.

[0098] Step S131c, for each stage except the second stage in the multiple stages, merge mining of the current stage is performed based on the production solution depth vector and capacitor performance depth vector mined in the previous stage, and the production solution depth vector and capacitor performance depth vector mined in the current stage are output.

[0099] In an embodiment of the present invention, after the first stage of mining, for each stage other than the second stage among the multiple stages, the current stage can be merged mined based on the production solution depth vector and capacitor performance depth vector mined in the previous stage, and the production solution depth vector and capacitor performance depth vector mined in the current stage can be output. For example, the second stage can be merged mined based on the production solution depth vector and capacitor performance depth vector mined in the first stage, so as to output the production solution depth vector and capacitor performance depth vector mined in the second stage. In addition, the merged mining method can be the same as the focused feature mining in two directions corresponding to step S131a and step S131b, which will not be repeated here.

[0100] Step S131d, after outputting the production plan depth vector and the capacitor performance depth vector mined in the last stage, the capacitor performance depth vector mined in the last stage is used as the capacitor performance mining vector corresponding to the combined mining of the at least one stage.

[0101] In an embodiment of the present invention, after outputting the production plan depth vector and capacitor performance depth vector mined in the last stage, the capacitor performance depth vector mined in the last stage can be used as the capacitor performance mining vector corresponding to the combined mining of at least one stage.

[0102] It can be understood that in the above step S131a, the specific manner of performing focused feature mining with a first direction based on the capacitor performance depth vector and the production solution depth vector loaded in the first stage is not limited. For example, in an alternative implementation, in order to take into account the needs of focused mining of associated information and enrichment of semantic information, combined with Figure 4 The above step S131a may further include the following specific implementation process:

[0103] First, for the combined mining of the first stage among the multiple stages, the production plan depth vector loaded into the first stage may be transposed, such as using the row parameters in the vector as column parameters and the column parameters as row parameters, and outputting the production plan transposed vector corresponding to the first stage;

[0104] Secondly, the capacitor performance depth vector loaded in the first stage, the production plan depth vector loaded in the first stage and the production plan transposed vector corresponding to the first stage can be associated and aggregated to output the production plan aggregate vector corresponding to the first stage; for example, the capacitor performance depth vector and the production plan transposed vector can be multiplied to obtain the corresponding correlation parameter, and then the correlation parameter and the production plan depth vector can be multiplied to obtain the production plan aggregate vector, that is, semantic information fusion based on correlation is realized;

[0105] Then, based on the production plan depth vector loaded in the first stage, the production plan aggregation vector corresponding to the first stage may be subjected to a first gradient optimization, and the production plan gradient optimization vector corresponding to the first stage may be output; illustratively, the production plan depth vector loaded in the first stage and the production plan aggregation vector corresponding to the first stage may be added or averaged to obtain the corresponding production plan gradient optimization vector;

[0106] Further, the production plan gradient optimization vector corresponding to the first stage may be subjected to at least one vector linear processing, and the production plan linear vector corresponding to the first stage may be output; illustratively, the vector linear processing may refer to multiplying the production plan gradient optimization vector with a weight matrix, and then adding the multiplication result with an offset matrix, wherein the weight matrix and the offset matrix may be network parameters of the corresponding neural network, which are formed during the training process;

[0107] Finally, based on the production plan gradient optimization vector corresponding to the first stage, the production plan linear vector corresponding to the first stage can be gradient optimized for the second time, and the production plan depth vector mined in the first stage can be output; illustratively, the production plan gradient optimization vector corresponding to the first stage and the production plan linear vector corresponding to the first stage can be added or the mean calculated to obtain the corresponding production plan depth vector.

[0108] It can be understood that, in the above step S131b, the specific manner of performing focused feature mining with a second direction based on the production solution depth vector mined in the first stage and the capacitor performance depth vector loaded in the first stage is not limited. For example, in an alternative implementation, in order to take into account the needs of focused mining of associated information and enrichment of semantic information, the above step S131b may further include the following specific implementation process:

[0109] First, for the combined mining of the first stage among the multiple stages, the capacitor performance depth vector loaded into the first stage is transposed (as described above), and the capacitor performance transposed vector corresponding to the first stage is output;

[0110] Secondly, the production plan depth vector mined in the first stage, the capacitor performance depth vector loaded in the first stage and the capacitor performance transposed vector corresponding to the first stage can be associated and aggregated, and the capacitor performance aggregation vector corresponding to the first stage can be output, wherein the specific method of association and aggregation can refer to the relevant description in the previous text;

[0111] Then, based on the capacitor performance depth vector loaded into the first stage, the capacitor performance aggregation vector corresponding to the first stage may be subjected to a first gradient optimization (as described above), and the capacitor performance gradient optimization vector corresponding to the first stage may be output;

[0112] Further, the capacitor performance gradient optimization vector corresponding to the first stage may be subjected to at least one vector linear processing, and the capacitor performance linear vector corresponding to the first stage may be output, wherein the specific manner of the vector linear processing may refer to the relevant description above, except that the weight matrix and the offset matrix used may be different, and the specific parameters may be formed during the training process;

[0113] Finally, based on the capacitor performance gradient optimization vector corresponding to the first stage, the capacitor performance linear vector corresponding to the first stage can be subjected to a second gradient optimization (as described above), and the capacitor performance depth vector mined in the first stage can be output.

[0114] It can be understood that, for the above-mentioned step S131, in other implementations, in order to mine more reliable semantic information, so that the semantic representation ability of the obtained capacitor performance mining vector is better, and in order to improve the overfitting problem caused by excessive attention to the semantics of the production plan due to the combined mining of multiple stages, when the at least one stage includes multiple stages, Figure 5 The above step S131 may further include the following specific implementation process:

[0115] First, for the combined mining of the first stage among the multiple stages, the capacitor performance depth vector and the production plan depth vector loaded in the first stage are subjected to difference calculation to obtain a target difference vector, and the target difference vector is weighted based on the interference parameter distribution corresponding to the first stage (i.e., the parameters at the corresponding positions are multiplied) to obtain the interference vector corresponding to the first stage; wherein the interference parameter distribution is formed in the training as the network parameter of the corresponding neural network, and each parameter in the interference parameter distribution is 1 or 0; in addition, the number of 0s can be much greater than the number of 1s, and can be specifically configured according to actual needs;

[0116] Secondly, the production solution depth vector loaded in the first stage can be added to the interference vector corresponding to the first stage to obtain a corresponding addition vector, and focusing feature mining with a first direction is performed based on the capacitor performance depth vector loaded in the first stage and the addition vector, and the production solution depth vector mined in the first stage is output;

[0117] Then, for the combined mining of the first stage among the multiple stages, focusing feature mining with a second direction is performed based on the production plan depth vector mined in the first stage and the capacitor performance depth vector loaded in the first stage, and the capacitor performance depth vector mined in the first stage is output;

[0118] Further, for each stage other than the second stage in the multiple stages, the target difference vector is weighted based on the interference parameter distribution corresponding to the current stage to obtain the interference vector corresponding to the current stage, and the production solution depth vector mined in the previous stage and the interference vector are added to obtain the corresponding addition vector, and combined mining of the current stage is performed based on the addition vector and the capacitor performance depth vector mined in the previous stage, and the production solution depth vector and the capacitor performance depth vector mined in the current stage are output;

[0119] Finally, after outputting the production solution depth vector and the capacitor performance depth vector mined in the last stage, the capacitor performance depth vector mined in the last stage is used as the capacitor performance mining vector corresponding to the combined mining of the at least one stage.

[0120] It can be understood that, in the above step S132, the specific manner of determining the production solution mining vector corresponding to the target capacitor performance data is not limited. For example, in an alternative implementation, in order to ensure that the mined production solution mining vector has a more reliable semantic representation capability, the above step S132 may further include the following specific implementation process:

[0121] First, the capacitor performance mining vector may be transposed, and a transposed capacitor performance mining vector corresponding to the capacitor performance mining vector may be output, as described above;

[0122] Secondly, based on the capacitor performance mining vector and the transposed capacitor performance mining vector, association aggregation is performed to output a capacitor performance self-aggregation vector corresponding to the capacitor performance mining vector. Exemplarily, the capacitor performance mining vector and the transposed capacitor performance mining vector may be multiplied to obtain a corresponding correlation parameter, and then the correlation parameter and the capacitor performance mining vector may be multiplied to obtain a capacitor performance self-aggregation vector.

[0123] Then, the capacitor performance self-aggregation vector is subjected to at least one vector linear processing, and a linear capacitor performance vector corresponding to the capacitor performance self-aggregation vector is output, as described above;

[0124] Finally, based on the capacitor performance self-aggregation vector, the linear capacitor performance vector is gradient optimized to output a production solution mining vector corresponding to the target capacitor performance data; illustratively, the capacitor performance self-aggregation vector and the linear capacitor performance vector can be added or the mean calculated to obtain the corresponding production solution mining vector.

[0125] Part 4: It should be noted that for step S140, the specific method of generating the target capacitor production plan corresponding to the target capacitor performance data is not limited. For example, in an alternative embodiment, the decoder included in the target production plan determination network formed by training (the training process can refer to the relevant description below. In addition, the target production plan determination network can also include an encoder, which can be used in steps S110-step S130) to process the production plan mining vector to obtain the corresponding target capacitor production plan. For example, after obtaining the production plan mining vector, the decoder can output the first word in the target capacitor production plan, and then the semantic vector of the first word and the production plan mining vector can be loaded into the decoder together, so that the decoder outputs the second word, so that each word in the target capacitor production plan can be output in sequence. Specifically, the decoder can output the probability of each candidate word based on the semantic vector currently input, for example, it can be implemented by a classification function such as softmax, and then the candidate word with the highest probability is determined as an output word.

[0126] In the fifth part, in order to ensure the effective execution of the above-mentioned steps S110 to S140, the target production plan determination network can also be trained. That is, the AI-based flexible DC support capacitor production plan determination method can also include the step of training to form the target production plan determination network. The specific implementation process of this step may include:

[0127] First, first capacitor performance data and second capacitor performance data for training may be determined, and a training production scheme may be determined, wherein the capacitor performance data corresponding to the training production scheme is similar to the first capacitor performance data (refer to the relevant description above), and the capacitor performance data corresponding to the training production scheme is not similar to the second capacitor performance data;

[0128] Secondly, the first capacitor performance data, the training production scheme and the second capacitor performance data can be semantically mined respectively to output a first training performance semantic vector corresponding to the first capacitor performance data, a training scheme semantic vector corresponding to the training production scheme and a second training performance semantic vector corresponding to the second capacitor performance data, as described above;

[0129] Then, the candidate production scheme determination network can be used to perform deep mining on the first training performance semantic vector, the training scheme semantic vector, and the second training performance semantic vector, respectively, and output a first training performance depth vector, a training scheme depth vector, and a second training performance depth vector, which can be implemented by an encoder included in the candidate production scheme determination network as described above;

[0130] Afterwards, the candidate production scheme determination network may be used to perform at least one stage of merged mining according to the first training performance depth vector and the training scheme depth vector, output a first training performance mining vector corresponding to the at least one stage of merged mining, and determine a first training scheme mining vector according to the first training performance mining vector, as described above, which may also be implemented by an encoder included in the candidate production scheme determination network;

[0131] Further, using the candidate production scheme determination network, performing at least one stage of merged mining according to the second training performance depth vector and the training scheme depth vector, outputting a second training performance mining vector corresponding to the at least one stage of merged mining, and determining a second training scheme mining vector according to the second training performance mining vector, as described above, can also be implemented by an encoder included in the candidate production scheme determination network;

[0132] Finally, the network parameters of the candidate production plan determination network can be updated based on the first training plan mining vector and the second training plan mining vector to form a target production plan determination network, wherein the target production plan determination network is used to generate a target capacitor production plan corresponding to the target capacitor performance data, that is, executing steps S110-S140.

[0133] It is understandable that, in the above steps, the specific manner of updating the network parameters of the candidate production scheme determination network is not limited. For example, in an alternative implementation, in order to ensure efficient and reliable training, the network parameters may be updated based on the following manner:

[0134] First, the candidate production scheme determination network can be used to perform scheme generation processing according to the first training scheme vector and the second training scheme vector respectively (for example, implemented by a decoder included in the candidate production scheme determination network), output the corresponding first capacitor production scheme and the second capacitor production scheme, and respectively calculate the differences between the first capacitor production scheme and the second capacitor production scheme and the training production scheme (which can be the edit distance between the schemes, etc.), to obtain the corresponding first production scheme difference and the corresponding second production scheme difference;

[0135] Secondly, based on the first production plan difference and the second production plan difference, the network parameters of the candidate production plan determination network can be updated to form a target production plan determination network; illustratively, the network parameters can be updated in the direction of reducing the first production plan difference and increasing the second production plan difference until preset conditions are met, such as the number of updates reaches a preset number, or the change amplitude of the difference converges.

[0136] It can be understood that the specific manner of updating the network parameters of the candidate production scheme determination network based on the first production scheme difference and the second production scheme difference is not limited. For example, in order to further improve the reliability of the target production scheme determination network formed by training, the network parameters can be updated based on the following manner:

[0137] First, the difference between the first training scheme mining vector and the second training scheme mining vector may be calculated to obtain the corresponding mining vector difference, such as the mean square error between the vectors;

[0138] Secondly, the differences between the first training scheme mining vector and the second training scheme mining vector and the training scheme semantic vector can be calculated respectively to obtain the corresponding first scheme semantic difference and second scheme semantic difference, such as the mean square error between the vectors;

[0139] Finally, based on the first production plan difference, the second production plan difference, the mining vector difference, the first plan semantic difference and the second plan semantic difference, the network parameters of the candidate production plan determination network can be updated to form a target production plan determination network; specifically, the network parameters can be updated in the direction of reducing the first production plan difference, increasing the second production plan difference, increasing the mining vector difference, reducing the first plan semantic difference and increasing the second plan semantic difference until a preset condition is met, such as the number of updates reaches a preset number, or the change amplitude of the difference (such as the difference between two adjacent differences is less than a threshold) converges.

[0140] In summary, the AI-based flexible DC support capacitor production plan determination method and system provided by the present invention firstly deeply mines the capacitor performance semantic vector corresponding to the target capacitor performance data and outputs the capacitor performance depth vector; secondly, deeply mines the production plan semantic vector corresponding to the target capacitor performance data and outputs the production plan depth vector; then, merges and mines the capacitor performance depth vector and the production plan depth vector to output the production plan mining vector; finally, generates the target capacitor production plan corresponding to the target capacitor performance data according to the production plan mining vector. Based on the above content, since the production plan mining vector as the basis for generating the target capacitor production plan includes not only the semantic information of the target capacitor performance data, but also the semantic information of the production plan corresponding to the capacitor performance data similar to the target capacitor performance data, that is, the semantic information of the target capacitor performance data can be reinforced based on the semantic information of the similar production plan, thereby effectively improving the semantic representation accuracy of the production plan mining vector, and then ensuring the reliability of the determined target capacitor production plan, therefore, the problem of relatively low reliability of capacitor production plan determination existing in the prior art can be improved.

[0141] In several embodiments provided in the embodiments of the present invention, it should be understood that the disclosed apparatus and method can also be implemented in other ways. The apparatus and method embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the apparatus, method and computer program product according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0142] In addition, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0143] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, an electronic device, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code. It should be noted that in this article, the term "include", "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements that are not explicitly listed, or also includes elements inherent to such process, method, article or device. Without more constraints, an element defined by the phrase "comprising a..." does not exclude the existence of other identical elements in the process, method, article or apparatus comprising the element.

[0144] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for determining a production plan for flexible DC support capacitors based on AI, characterized in that: The AI-based method for determining a flexible DC support capacitor production plan includes: Deeply mine the capacitor performance semantic vector corresponding to the target capacitor performance data, and output the capacitor performance depth vector, wherein the target capacitor performance data is used to describe at least one performance of the target flexible DC support capacitor to be produced, and the performance includes capacitance value, maximum operating voltage, equivalent series resistance and filtering performance; Deeply mining the production solution semantic vector corresponding to the target capacitor performance data, and outputting the production solution deep vector, wherein the production solution semantic vector is formed by semantically mining the production solution corresponding to the capacitor performance data similar to the target capacitor performance data, where the similarity refers to that the difference between the parameters of the included performance is less than a threshold value; For the combined mining of the first stage among multiple stages, focused feature mining with a first direction is performed based on the capacitor performance depth vector and the production solution depth vector loaded in the first stage, and the production solution depth vector mined in the first stage is output, wherein the capacitor performance depth vector and the production solution depth vector loaded in the first stage are the initial capacitor performance depth vector and the production solution depth vector, and the focused feature mining with a first direction means that the semantic information of the capacitor performance depth vector loaded in the first stage is integrated into the semantic information of the production solution depth vector loaded in the first stage; focused feature mining with a second direction is performed based on the production solution depth vector mined in the first stage and the capacitor performance depth vector loaded in the first stage, and the first stage mining is output. The capacitor performance depth vector obtained, wherein the focused feature mining with the second direction refers to fusing the semantic information of the production plan depth vector mined in the first stage into the semantic information of the capacitor performance depth vector loaded in the first stage; for each stage other than the first stage in the multiple stages, the combined mining of the current stage is performed based on the production plan depth vector and the capacitor performance depth vector mined in the previous stage, and the production plan depth vector and the capacitor performance depth vector mined in the current stage are output; after the production plan depth vector and the capacitor performance depth vector mined in the last stage are output, the capacitor performance depth vector mined in the last stage is used as the capacitor performance mining vector; based on the capacitor performance mining vector, the production plan mining vector corresponding to the target capacitor performance data is determined; Based on the production plan mining vector, a target capacitor production plan corresponding to the target capacitor performance data is generated, wherein the target capacitor production plan is at least used to reflect at least one process parameter of the target flexible DC support capacitor in the production process, and the process parameters include electrode-related parameters, insulation layer-related parameters, welding process parameters and electrolyte-related parameters.

2. The method for determining a flexible DC support capacitor production plan based on AI according to claim 1 is characterized in that: The step of performing a focused feature mining with a first direction based on the capacitor performance depth vector and the production solution depth vector loaded into the first stage and outputting the production solution depth vector mined in the first stage for the combined mining of the first stage of the multiple stages includes: For the combined mining of the first stage among the multiple stages, transposing the production plan depth vector loaded into the first stage, and outputting the production plan transposed vector corresponding to the first stage; Associatively aggregate the capacitor performance depth vector loaded into the first stage, the production plan depth vector loaded into the first stage, and the production plan transposed vector corresponding to the first stage, and output the production plan aggregate vector corresponding to the first stage; Based on the production plan depth vector loaded in the first stage, a first gradient optimization is performed on the production plan aggregation vector corresponding to the first stage, and a production plan gradient optimization vector corresponding to the first stage is output; Perform at least one vector linear processing on the gradient optimization vector of the production plan corresponding to the first stage, and output the linear vector of the production plan corresponding to the first stage; Based on the production plan gradient optimization vector corresponding to the first stage, the production plan linear vector corresponding to the first stage is subjected to a second gradient optimization, and the production plan depth vector mined in the first stage is output.

3. The method for determining a flexible DC support capacitor production plan based on AI according to claim 1 is characterized in that: The step of performing focused feature mining with a second direction on the combined mining of the first stage of the multiple stages and outputting the capacitor performance depth vector mined in the first stage based on the production solution depth vector mined in the first stage and the capacitor performance depth vector loaded in the first stage, comprises: For the combined mining of the first stage among the multiple stages, transposing the capacitor performance depth vector loaded into the first stage, and outputting the capacitor performance transposed vector corresponding to the first stage; Associatively aggregate the production plan depth vector mined in the first stage, the capacitor performance depth vector loaded in the first stage, and the capacitor performance transposed vector corresponding to the first stage, and output the capacitor performance aggregate vector corresponding to the first stage; Based on the capacitor performance depth vector loaded in the first stage, performing a first gradient optimization on the capacitor performance aggregation vector corresponding to the first stage, and outputting the capacitor performance gradient optimization vector corresponding to the first stage; Perform at least one vector linear processing on the capacitor performance gradient optimization vector corresponding to the first stage, and output the capacitor performance linear vector corresponding to the first stage; Based on the capacitor performance gradient optimization vector corresponding to the first stage, the capacitor performance linear vector corresponding to the first stage is subjected to a second gradient optimization, and the capacitor performance depth vector mined in the first stage is output.

4. The method for determining a flexible DC support capacitor production plan based on AI according to claim 1 is characterized in that: The step of determining the production solution mining vector corresponding to the target capacitor performance data based on the capacitor performance mining vector includes: Transposing the capacitor performance mining vector, and outputting a transposed capacitor performance mining vector corresponding to the capacitor performance mining vector; Performing association aggregation based on the capacitor performance mining vector and the transposed capacitor performance mining vector, and outputting a capacitor performance self-aggregation vector corresponding to the capacitor performance mining vector; Perform at least one vector linear processing on the capacitor performance self-aggregation vector, and output a linear capacitor performance vector corresponding to the capacitor performance self-aggregation vector; Based on the capacitor performance self-aggregation vector, the linear capacitor performance vector is gradient optimized, and a production solution mining vector corresponding to the target capacitor performance data is output.

5. The method for determining a flexible DC support capacitor production plan based on AI according to any one of claims 1 to 4, characterized in that: The AI-based method for determining a flexible DC support capacitor production plan also includes: Determine first capacitor performance data and second capacitor performance data for training, and determine a training production plan, wherein the capacitor performance data corresponding to the training production plan is similar to the first capacitor performance data, and the similarity means that the difference between the parameters of the included performance is less than a threshold value, and the capacitor performance data corresponding to the training production plan is not similar to the second capacitor performance data; Perform semantic mining on the first capacitor performance data, the training production scheme, and the second capacitor performance data, respectively, and output a first training performance semantic vector corresponding to the first capacitor performance data, a training scheme semantic vector corresponding to the training production scheme, and a second training performance semantic vector corresponding to the second capacitor performance data; Determine the network using the candidate production scheme, perform deep mining on the first training performance semantic vector, the training scheme semantic vector, and the second training performance semantic vector, respectively, and output a first training performance depth vector, a training scheme depth vector, and a second training performance depth vector; Determine a network using the candidate production scheme, perform at least one stage of merged mining based on the first training performance depth vector and the training scheme depth vector, output a first training performance mining vector corresponding to the at least one stage of merged mining, and determine a first training scheme mining vector based on the first training performance mining vector; Determine a network using the candidate production scheme, perform at least one stage of merged mining based on the second training performance depth vector and the training scheme depth vector, output a second training performance mining vector corresponding to the at least one stage of merged mining, and determine a second training scheme mining vector based on the second training performance mining vector; Based on the first training scheme mining vector and the second training scheme mining vector, the network parameters of the candidate production scheme determination network are updated to form a target production scheme determination network, wherein the target production scheme determination network is used to generate a target capacitor production scheme corresponding to the target capacitor performance data.

6. The method for determining a flexible DC support capacitor production plan based on AI according to claim 5 is characterized in that: The step of updating the network parameters of the candidate production solution determination network according to the first training solution mining vector and the second training solution mining vector to form a target production solution determination network includes: Using the candidate production scheme determination network, performing scheme generation processing according to the first training scheme vector and the second training scheme vector, outputting the corresponding first capacitor production scheme and second capacitor production scheme, and respectively calculating the differences between the first capacitor production scheme and the second capacitor production scheme and the training production scheme to obtain the corresponding first production scheme difference and the corresponding second production scheme difference; Based on the first production plan difference and the second production plan difference, the network parameters of the candidate production plan determination network are updated to form a target production plan determination network.

7. The method for determining a flexible DC support capacitor production plan based on AI according to claim 6 is characterized in that: The step of updating the network parameters of the candidate production scheme determination network based on the first production scheme difference and the second production scheme difference to form a target production scheme determination network includes: Calculating the difference between the first training scheme mining vector and the second training scheme mining vector to obtain a corresponding mining vector difference; respectively calculating the differences between the first training scheme mining vector and the second training scheme mining vector and the training scheme semantic vector to obtain corresponding first scheme semantic differences and second scheme semantic differences; Based on the first production plan difference, the second production plan difference, the mining vector difference, the first plan semantic difference and the second plan semantic difference, the network parameters of the candidate production plan determination network are updated to form a target production plan determination network.

8. A flexible DC support capacitor production plan determination system based on AI, characterized in that: The AI-based flexible DC support capacitor production plan determination system includes: Memory for storing computer programs; A processor connected to the memory is used to execute a computer program stored in the memory to implement the AI-based method for determining a flexible DC support capacitor production plan as described in any one of claims 1 to 7.

Images