Module and method

Abstract

The invention relates to a module and a method for flexibly carrying out tasks relating to the provision of information relating to a technical system provided with system software. The module (1) according to the invention comprises an input interface (2) for inputting tasks, analysis software (3) for analysing tasks input by means of the input interface, an interface (4) to an LLM-based system for providing outputs initiated by a prompt and generated by means of Internet information, and an interface (5) to the system software (6). The analysis software (3) is designed to analyse input tasks relating to the provision of information relating to the technical system provided with system software in order to determine which information can be provided by the system software (6) for carrying out the particular task, and to access the LLM-based system for information that cannot be provided by means of the system software (6) and is required for carrying out the particular task in order to provide outputs initiated by a prompt and generated by means of Internet information.

Description

[0001] 202404079

[0002] 1

[0003] Description

[0004] Module and procedure

[0005] The invention relates to a module and a method for carrying out tasks relating to the provision of information on a technical system equipped with system software.

[0006] With the ever-increasing pace of technical and regulatory changes and developments, it is becoming increasingly challenging for more complex technical systems to provide system-related data in the required timeliness and according to the needs and wishes of the user.

[0007] An example of a complex technical system is energy distribution systems or their components. These are subject to requirements that vary regionally and must be configured and operated in accordance with local regulations. For example, energy audits must be conducted according to standards that not only differ by country but can also change over the system's operating lifetime (e.g., DIN EN 16247-1, ISO 50001 in Germany, ASHREA process or ISO 50002 in the USA).

[0008] The invention aims to make the provision of information relating to a technical system equipped with system software more flexible.

[0009] The problem is solved by a module according to claim 1 and a method according to claim 13. Advantageous further developments are specified in the dependent claims.

[0010] The process steps listed below in the description of the module according to the invention are each, individually or in combination, also part of a process according to the invention.

[0011] The module according to the invention is designed to perform tasks related to the provision of information concerning a technical system equipped with system software (a "technical system" is a physical system that serves a technical purpose, e.g., a system for energy supply or distribution). It comprises an input interface for entering tasks and analysis software for [202404079]

[0012] 2

[0013] Analysis of tasks entered via the input interface, an interface to an LLM-based system for providing outputs initiated by a prompt (or dialog) and generated using Internet information ("LLM" stands for "large language model", i.e., refers to ChatGPT or a ChatGPT-like system), and an interface to the system software.

[0014] The analysis software is designed to analyze input tasks related to the provision of information concerning a technical system equipped with system software, determining which information required for completing the respective task can be provided by the system software. For information required for completing the respective task that cannot be provided by the system software, the software utilizes the LLM-based system for providing outputs generated from internet information, initiated by a prompt or dialog.

[0015] It may be possible to analyze recorded tasks to determine the type of information source from which information is needed or obtained: a) only from the LLM-based system, b) only from the system software, c) or from both.

[0016] Task processing then proceeds according to its type; that is, only the respective information source (type a) or b)) or both information sources (type c)) are used to complete the task. For example, different processing algorithms may be used depending on the task. By accessing two different information sources (LLM-based system or system software) depending on the task, flexible and efficient task processing is possible. In particular, comprehensive maintenance of the data stored locally on the system and queried via the system software is eliminated.

[0017] Tasks typically involve providing information to the user, but are not necessarily limited to this. The module can also be designed to perform a parameter change in the system software as part of a task. This change would then be confirmed or communicated to the user.

[0018] The module can be implemented as a combination of hardware and software or as a purely software module. In the form of a purely software solution, it could be, for example, 202404079.

[0019] 3

[0020] The download or update will be made available and will run on the computing resources (e.g., CPU or MCU) of the technical system.

[0021] The term "interface" should be understood functionally and does not initially imply anything about the architecture or structure of the software, but rather expresses the establishment of access to information sources external to the module; that is, the module is designed to access the LLM-based system or the system software. Of course, such interfaces can also be explicitly programmed.

[0022] The term "an interface to an LLM-based system" also encompasses the constellation of multiple LLM-based systems, each of which can be accessed. The analysis system may select an LLM-based system. The term "prompt" should be interpreted as accessing one or more LLM-based systems in a suitable manner, and specifically includes the case of a dialog conducted for this purpose.

[0023] In one configuration of the module, the types of editable tasks are defined (typically, these tasks involve providing information in some form, but it may also be possible, for example, to modify system parameters via a task). For tasks that do not correspond to a defined type, an output may be provided indicating that editing that task type is not possible. Restricting the task type to specific categories (e.g., creating an energy audit, querying general technical information, querying system-specific information, calculating system-specific values, etc.) may be advisable for legal reasons (e.g., potential copyright infringement).

[0024] According to one embodiment, the analysis software is designed to extract task-relevant information from outputs provided by the LLM-based system upon prompt initiation, and to process or prepare all task-relevant information for output. This can include combining information obtained by the LLM-based system and information obtained by the system software, and embedding it in a text template. In other words, the module is designed to access the system software for information required to complete the respective task, and to combine this information for tasks requiring information from both the LLM-based system and the system software. 202404079

[0025] 4

[0026] The processing may include formatting information for output required as part of a task. This formatting is then carried out according to the specifications of the respective task.

[0027] The module can be configured for a variety of output methods (e.g., speech output, text on screen, email with PDF attachment, or a combination thereof) and for specifying the output method according to the specific task. This specification can be included in the task definition (e.g., an option or request for a preferred output file format in the task description). The processing then involves adapting the information for output according to the specified output method.

[0028] A further development of the module according to the invention comprises a learning function in which the module is designed to save an entered task and to save the information required for completing the task. This information is obtained from output generated by the Internet via the LLM-based system, which is initiated by a prompt (or dialog), and is assigned to the task and used when the task is repeated. The module can then be configured to check whether an entered task is a saved task and, if a saved task has associated information obtained via the LLM-based system, to use this information for carrying out the task. To check whether a task is a saved task, an input dialog can be executed when the task is entered.The module can also be designed to check, before using this information to perform the task, whether it is still up-to-date (e.g., forms or regulations for an energy audit), for repetitions of a saved task with associated information obtained via the LLM-based system (depending on the task and possibly also on a constraint, e.g., the time elapsed since the original task was carried out).

[0029] The invention also relates to a device with a module according to the invention, e.g. an accessory element equipped with a processor which can be connected to the technical system (e.g. plug connection, whereby a communicative connection is also established when plugging).

[0030] The invention is explained in more detail below using exemplary embodiments and the figures shown. Figures 202404079 are shown.

[0031] 5

[0032] Fig. 1 a module according to the invention,

[0033] Fig. 2: the basic sequence of a method according to the invention,

[0034] Fig. 3a: the sequence of a method according to the invention for a first problem,

[0035] Fig. 3b: an example of an output for a task corresponding to Fig. 3a,

[0036] Fig. 4: the sequence of a method according to the invention in a second problem,

[0037] Fig. 5: the course of a method according to the invention in a third problem and

[0038] Fig. 6: the sequence of a method according to the invention in a fourth problem.

[0039] Fig. 1 schematically shows a module 1 according to the invention, which is designed to perform tasks related to providing information about a power supply system equipped with system software 6. It is provided with an input interface 2 for inputting tasks. In this example, this is a software plugin (“Voice | Chat Plugin”, which functions as an I / O interface) that enables voice input and output, thus allowing communication or chat between an operator and the module 1. The core of the module 1 is a processing software 3 (also referred to as “CORE” in the figure), which includes analysis software for analyzing tasks entered via the input interface. These tasks are entered via the input interface 2 and transmitted to the processing software 3.The processing software 3 can access an interface to an LLM-based system (e.g., ChatGPT or a system comparable to ChatGPT) for processing tasks, in order to provide outputs initiated by a prompt (or dialog) and generated using internet information. In addition, module 1 is equipped with an interface (indicated by reference 5) to the system software 6.

[0040] The analysis software is designed to analyze entered tasks relating to the provision of information about the energy supply system equipped with system software, to determine which information required for completing the respective task can be provided by the system software and 202404079

[0041] 6. To access the LLM-based system for information not provided by the system software, which is required to complete the respective task, in order to provide outputs initiated by a prompt (or dialog) and generated using Internet information.

[0042] The functionality of this analysis software, or rather this module, is explained in more detail below using examples of different types of tasks. Two different types of information sources can be used to process and complete these tasks: the LLM-based system and the energy supply system, or rather, the corresponding system software.

[0043] In principle, the tasks can be divided into three types, depending on the type of information source from which information is needed or obtained: a) only from the LLM-based system b) only from the system software c) from both

[0044] In the figures, hatching indicates which component of the module from Fig. 1 carries out the respective step (identity of the hatching of the component and the element in the flowchart).

[0045] Figure 2 illustrates the basic procedure. In step S21, an input or a request is made. This either represents a task or a task is derived from it. It may happen that the input is insufficient to define the task precisely (query S22). In this case, follow-up questions may be provided, or a dialogue may ensue. Follow-up questions can also be initiated in a later processing step, for which an example is provided in Figure 5. If the task is (at least provisionally) defined with sufficient precision, it is assessed whether external information is necessary, i.e., whether the task can be processed solely based on information available from the system software. In particular, a classification can be made according to which of the three types a) - c) mentioned above the task belongs to. If no external information (i.e.,If information is required (to be provided via the LLM-based system), the system software is queried for the information necessary for the task (step S24), processed (step S25), and the result is output (S26). Otherwise, the LLM-based system (e.g., ChatGPT) is used via a corresponding prompt or 202404079.

[0046] 7

[0047] The required external information is determined via a dialog (step S27). If additional internal information is needed (S28), a query is made to the system software (S29), and external and internal information are integrated according to the output task (step S30).

[0048] The individual steps do not necessarily have to proceed in the order shown, nor do they have to occur in exactly the same form. For example, step S23 might already clearly define which internal and external information is required. In that case, queries for internal and external information could be performed simultaneously, and step S28 would be unnecessary. The sequence shown in Fig. 2 was chosen for the sake of simplicity and clarity and will be used below to illustrate specific tasks. However, those skilled in the art will immediately recognize the modifications, extensions, and optimizations encompassed by the invention.

[0049] Figure 3a shows a simple example that relies solely on external information (task of type a)). The user requests the definition of the term "energy meter" (step S31). In the next step, S32, the task is classified as clarified, meaning no further questions are needed. In the next step, S33, it is classified as a task requiring external information. This information is retrieved using ChatGPT (step S34). Since no further internal information is required (step S35), an output can be generated immediately (step S36). Figure 3b shows an example of this output.

[0050] Figure 4 shows an example where no external information is required (task type b)). Query S41 concerns the rated current of a 3WA switch in the power supply system. The question is clear (step S42). Therefore, no external information is required (step S43), and querying this information from the system software (step S44) is sufficient. The result can be further processed in step S45 (e.g., inserting the queried current value into a text) and communicated to the user via written or spoken output (step S46).

[0051] Figure 5 shows a more complex example, namely the creation of an energy audit for the energy supply system. In step S51, the task of conducting an energy audit is set. With this complex task, it is likely that follow-up questions will be necessary to clarify the task (step S52). For example, in this dialogue, 202404079

[0052] Step 8 clarifies for which system the energy audit is to be carried out (the entire energy supply system or only individual, identifiable components thereof) and for which country it is to be carried out (i.e., where the energy supply system is located). Once the task has been classified as clear, it is determined that external information is required (step S53). A corresponding prompt for ChatGPT is created (step S54), or a dialogue is initiated with ChatGPT. It may become apparent that further input is needed for clarification. Accordingly, questions can be posed to the user, or a user dialogue can be initiated. Information required for completing the task, such as forms, regulations, etc., can then be obtained via ChatGPT.Since additional internal information is required for the energy audit (query S55), this information is retrieved from the system software (step S56), and an energy audit report is generated for output (step S57). This report is then sent to the user, for example, via email (output S58). The module according to the invention includes a learning function that records completed tasks. This learning function is then used when new tasks are created, allowing the system to access previously recorded information for a given task when it is repeated. For example, if the energy audit described in Fig. 5 were performed again, the information stored in the learning function for a previous processing session could be used in step S52 to identify the task (e.g., by asking a question such as "Is this an audit according to regulation xxx?") and to complete the task more efficiently.In the case of an energy audit, a further ChatGPT dialogue may still be conducted to ensure that there have been no changes to the regulations. However, there are also tasks where the type can change. This is illustrated in Fig. 6.

[0053] Figure 6 shows a task where the user requests the system consumption for the last 24 hours (step S61). This task may, and probably probably, require further clarification in step S62 (e.g., whether the consumption of the entire energy distribution system or a part of it should be determined, or regarding the limitation of the time period (which 24 hours exactly)). When processing this task for the first time, external and internal information is required (task type c)), because after the corresponding analysis (step S63), the calculation formula is obtained via ChatGPT (step S64). For the calculation using the formula, internal information on parameters measured and stored by the energy system is required (e.g., current and voltage values). Accordingly (step S65), these parameters are queried from the system software (step S66). The consumption can then be calculated (step S67) and displayed (step S68).The learning function saves the calculation formula (step S69). This will be used when the task is presented again (202404079).

[0054] Therefore, it is already present in 9 and no longer needs to be determined using ChatGPT. The task type is then no longer c) but b).

[0055] The explanations provided in the exemplary embodiment are merely illustrative. Other embodiments, modifications, and optimizations are immediately obvious to those skilled in the art. These are part of the invention concept and are included in the scope of protection.

Claims

202404079 10 Patent claims 1. Module (1) for carrying out tasks relating to the provision of information relating to a technical system equipped with system software (6), comprising - an input interface (2) for entering tasks, - an analysis software (3) for analyzing tasks entered via the input interface, - an interface (4) to an LLM-based system for providing prompt-initiated outputs generated by Internet information, and - an interface (5) to the system software (6), wherein the analysis software (3) is designed to - to analyze entered tasks that concern the provision of information relating to the technical system equipped with system software (6) to determine which information required for the completion of the respective task can be provided by the system software, and - to access the LLM-based system for the provision of prompt-initiated outputs generated by Internet information for the performance of the respective task, for information that cannot be provided by means of the system software (6).

2. Module according to claim 1, characterized in that - the analysis software (3) is trained to extract information relevant for the performance of the task from outputs provided by the LLM-based system upon initiation by a prompt, and - to process all information relevant to the execution of the task for one output.

3. Module according to claim 2, characterized in that - it is trained to access the system software (6) for information that can be provided by means of the system software (6) and is required to perform the respective task, and - the processing by the analysis software (3) for the performance of tasks requiring information from both the LLM-based system and the system software (6) includes a combination of this information. 202404079 11 4. Module according to one of the preceding claims 2 and 3, characterized in that - the processing includes formatting information for the output of information required in the course of a task, and - this formatting is carried out according to the requirements of the respective task.

5. Module according to claim 4, characterized in that - the module is designed for a variety of different output methods, - the module for defining the output procedure is designed according to the specific task, and - the processing includes an adjustment of information for output according to the specified output procedure.

6. Module according to one of the preceding claims, characterized in that - the technical system is a system for energy supply, - a task that can be performed by the module consists of conducting an energy audit for the system or a part of the system, and - for this task, information is accessed from an output provided by the LLM-based system upon initiation by a prompt.

7. Module according to one of the preceding claims, characterized in that - a task that can be performed by the module consists of answering a question and / or performing a calculation relevant to the system.

8. Module according to one of the preceding claims, characterized in that the module comprises a learning function, in the course of which the module is designed for this purpose. - to save an entered task, and - information required to complete the task, obtained from an output generated by Internet information through the LLM-based system for providing prompt-initiated output, -- to save, -- to assign to the task, and 202404079 12 -- to be used when repeating the task.

9. Module according to claim 8, characterized in that the module is designed to, - to check whether a task has been entered is a saved task, and - to use a saved task with associated information obtained via the LLM-based system to carry out the task.

10. Module according to claim 9, characterized in that the module is designed to perform an input dialog when the task is entered, in order to check whether it is a saved task.

11. Module according to claim 9 or 10, characterized in that the module is designed to check, using the LLM-based system, whether this information is still up-to-date before using this information for carrying out the task, in the case of certain repetitions of a stored task with associated information obtained by means of the LLM-based system.

12. Device comprising a module according to one of the preceding claims.

13. Procedures for carrying out tasks relating to the provision of information relating to a technical system equipped with system software, comprising - Recording tasks entered via an input interface (2), - Analysis of tasks entered via the input interface (2), - Analysis of entered tasks that concern the provision of information related to the technical system equipped with system software, to determine which information required for the completion of the respective task can be provided by the system software, and 202404079 13 - Accessing an LLM-based system to provide prompt-initiated outputs generated from Internet information for information not available through system software, required to complete the respective task.

14. Method according to claim 13, characterized in that - The recorded tasks are analyzed to determine the type of information source from which information is required or obtained: a) only from an LLM-based system, b) only from the system software, c) from both, and - the task will be processed according to its type.

15. Method according to claim 13 or 14 characterized by the fact that - a task that has been entered is saved, and - information required to complete the task, obtained from an output generated by Internet information through the LLM-based system for providing prompt-initiated output, -- saved, -- assigned to the task, and -- to be used when repeating the task.

16. Method according to claim 15, characterized in that - when a task is entered, it is checked whether it is a saved task, and - in the case of a saved task with associated information obtained via the LLM-based system, this information is used to carry out the task.

Images