Method for checking the plausibility of a forecast
Patent Information
- Application Number
- EP2023707030
- Authority / Receiving Office
- EP · EP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2022-03-23
- Filing Date
- 2023-02-20
- Publication Date
- 2024-11-27
AI Technical Summary
In local energy markets, energy providers face risks and additional costs due to inaccurate energy forecasts, especially during disruptive events like snow on PV systems or wind failures, as they must cover energy shortfalls from higher-level networks, leading to inefficient energy allocation and increased costs.
A method and device for checking the plausibility of energy forecasts by calculating a 'confidence rating' based on influencing variables, using a processor to compare the calculated probability with a limit value, and issuing warnings or revisions to energy producers, thereby improving forecast compliance and reducing costs.
Enhances the robustness of local energy markets by early detection of disruptive events, improves bid quality, reduces costs, and optimizes reserve capacities, allowing for better coordination and increased local resource efficiency.
Smart Images

Figure 1.1
Abstract
Description
[0001] Description
[0002] Procedure for checking the plausibility of a forecast
[0003] The invention relates to a method for checking the plausibility of a forecast and a device for checking the plausibility.
[0004] In a local overall energy system, e.g. in a city district or a municipality, there are n different energy subsystems, e.g. buildings, each with different energy technology systems, e.g. energy generation, consumption or storage systems. The most efficient allocation, i.e. assignment and distribution, of the energy generated and consumed in the overall system as well as the exchange on the respective energy grid represents a central optimization / coordination problem. However, since each energy subsystem only has knowledge of its own energy technology systems, a central coordination platform is required for the optimized exchange between the energy subsystems and the energy grid while adhering to the technical restrictions of the energy grid. Such a central coordination platform can, for example, be described in DE102018221156A1 (Siemens Aktiengesellschaft) 10.06.2020, will be provided by a Local Energy Market (= LEM).
[0005] However, participation in energy trading on the LEM, such as day-ahead or intraday trading, also carries risks. Incorrect forecasts of the amount of energy to be provided may result in additional costs for the energy provider: if the energy provider can only provide the LEM with a smaller amount of energy than it previously offered, it must purchase and pay for the energy difference from other sources, such as a higher-level grid.
[0006] This problem is particularly aggravated by "disruptive" events, when the amount of energy available decreases unexpectedly due to errors or external influences, e.g. snow on a PV system, a failure of a CHP plant or calm wind at a wind turbine (PV = photovoltaics; CHP = combined heat and power plant).
[0007] The present invention is based on the object of improving a forecast with regard to the amount of energy available in a local energy market.
[0008] The object is achieved by a method having the features of independent patent claim 1 and by a device having the features of independent patent claim 11. Advantageous embodiments and further developments of the invention are specified in the dependent patent claims.
[0009] The method according to the invention is used to check the plausibility of a forecast from a first energy producer; the forecast states which quantity of energy a first energy generation plant of the first energy producer will provide in a local energy market in a defined period of time. In the following, in accordance with the terminology used in a LEM, such a forecast is also referred to as a “bid”. According to the method, a first data set containing information about the forecast from the first energy producer is transmitted to a processor that is available to the local energy market for the plausibility check. According to the method, one or more further data sets containing information about influencing factors that could influence the quantity of energy that can be provided are transmitted to the processor. The processor then calculates a probability that the forecast can be met.This probability can therefore be characterized with the term "confidence rating". This calculation of the probability is carried out on the basis of the information provided on the influencing variables. The processor then compares the calculated probability with a predetermined limit value. If the calculated probability falls below the limit value, the processor causes a corresponding notification to be transmitted to the first energy producer. The method according to the invention and / or one or more functions, features and / or steps of the method according to the invention and / or its embodiments can be computer-aided.
[0010] The device for plausibility checking according to the invention is designed to carry out the method.
[0011] A fundamental idea of the present invention is that a higher-level instance, i.e. the device for plausibility checking, hereinafter also referred to as the "monitoring system", checks the plausibility of the incoming forecasts. One aspect of the invention is therefore a central unit ("intelligence") of the LEM, which assigns a "confidence rating" to incoming energy supply offers. This monitoring system, which forms this central institution of the LEM, processes the incoming bids and information relevant for the plausibility check, such as plant information, weather information, measured power, in order to determine the extent to which the incoming bids can probably be met ("confidence rating"). The output thus results in a probability between 0 and 1, where 1 means that the bid can be met with a forecast probability of 100%.Subsequently, if the first energy producer to submit the forecast falls below one of the threshold values, which could be 0.5, for example, the energy producer will receive a warning that he should revise his bid.
[0012] The invention increases the overall robustness of a LEM.
[0013] The early detection of disruptive events can improve the quality of bids and their subsequent compliance. This, in turn, allows for better coordination within the LEM, potentially reducing the utilization of a higher-level network. The invention allows for better estimation and more efficient use of reserve capacities, such as storage batteries.
[0014] The invention reduces the additional costs for LEM participants due to forecast errors. This, in turn, reduces the incentive or necessity to place conservative bids on the LEM, e.g., only 75% of the predicted PV generation. As a result, local resource efficiency increases, as more renewable energy is offered or traded on the LEM, thus fully exploiting its potential. The LEM thus moves away from business logic and closer to physical reality.
[0015] Advantageous embodiments and further developments of the invention are specified in the dependent claims. The method according to the invention can also be further developed in accordance with the dependent device claims, and vice versa.
[0016] According to a preferred embodiment of the invention, the probability is determined by means of a linear polynomial, e.g., according to y = ßo + ßlXl + / ?21'2 + • ■ • + ßpX p calculated. The variables x n (with n = 1, 2, ..., p) are influencing factors, e.g. outside temperature, global radiation, angle of inclination, cloud movement in the next 15 minutes. The ß n are weighting factors for the influencing variables, with n = 1, 2, ..., p; the weighting factors can be obtained from historical data using logistic regression. ß0 represents a constant that may be necessary for correctly calculating the probability Y. The result Y is the desired probability: Y=0 means that the first power generation plant fails completely and cannot supply any energy; Y=1 means that the first power generation plant can fully supply the forecasted amount of energy.
[0017] According to a preferred embodiment of the invention, the probability is calculated using a neural network with a sigmoid as the activation function of the output layer. The model parameters are determined using binary cross entropy as the cost function. The cross entropy as the cost function only serves to determine the parameters (training the model). During "ongoing operation", the neural network no longer requires this function; here, the architecture of the output layer (sigmoid) ensures that a probability Y is output. The advantage is that it is a completely non-linear system, which can also map a linear polynomial.
[0018] According to a preferred embodiment of the invention, the forecast is compared with one or more other forecasts for other energy generation plants that are comparable to the first energy generation plant. In contrast to the previous procedure, in which each energy producer creates its own forecast without any comparison between the forecasts, the forecasts are validated on the basis of comparable plants. The advantage of this is that even with "dumb" energy generation plants, i.e. without their own sensors, it is possible to check whether their forecasts are plausible based on the forecasts of other energy producers with similar energy generation plants.
[0019] According to a preferred embodiment of the invention, the spatial distribution of the first and the other energy generation plants is taken into account in order to detect weather influences. Nowadays there is a very large number of prosumers: private households, for example, with their own PV roof systems act as energy producers. The invention uses the large number of measuring points or sensors thus available and their widespread local distribution. The spatial distribution of the energy generation plants can be used to detect disruptive (local) weather influences earlier and with a higher probability. For example, external influences occurring at individual energy generation plants, such as calm winds, snowfall or large cloud fronts, can be detected at an early stage and transferred to the other energy generation plants in the LEM.In this way, the forecast quality of renewable energy generation plants can be significantly improved.
[0020] According to a preferred embodiment of the invention, the influencing variables comprise plant information of the first energy generation plant, e.g. geographical coordinates, orientation, installation location, angle of inclination of the first energy generation plant, and / or plant information of other energy generation plants.
[0021] For example, the different installation specifications of energy generation systems can be utilized. In the case of private snow-covered PV systems, for example, it is often necessary to wait for a layer of snow covering PV modules, thus impairing their power production, to either melt or slide off. Since both processes depend on both the orientation and the installation angle (steepness) of the PV modules, these phenomena can be easily transferred to other energy generation systems and / or energy generation system specifications. This enables, for example, a very accurate prediction of the point in time at which a snow-covered PV system will begin generating electrical energy again.
[0022] The LEM not only has access to the individual data of each energy producer, but also to further (measurement) data from other energy producers. When using plant information from other energy generation plants that are similar to the first energy generation plant (same product, same or similar type), the characteristic behavior of another energy generation plant before a failure of the other energy generation plants occurred can be used to predict an impending failure of the first energy generation plant. If, for example, several energy producers are connected to the same CHP plant in the LEM, the characteristic behavior that all of these CHP plants show before a failure can be used to predict a failure of a CHP plant.
[0023] According to a preferred embodiment of the invention, the influencing variables include weather data and / or weather forecasts. The spatial distribution of the energy generation plants can be used to detect disruptive (local) weather influences earlier and with a higher probability. For example, external influences occurring at individual energy generation plants, such as calm winds, snowfall, or large cloud fronts, can be detected early and transferred to the other energy generation plants in the LEM. This can significantly improve the forecast quality of renewable energy plants.
[0024] According to a preferred embodiment of the invention, the influencing variables include information on the amount of energy provided in the past by the first energy generation plant and / or by other energy generation plants. Thus, a forecast made for a specific weather situation can be compared with previous forecasts made for similar weather conditions.
[0025] According to a preferred embodiment of the invention, the influencing variables comprise forecasts relating to other energy generation plants, the first energy producer and / or those of other energy producers. Thus, a forecast made in a specific weather situation and / or with a specific configuration and / or state of an energy generation plant can be compared with earlier forecasts made in the same or similar weather conditions and / or with the same or similar configuration and / or state of another energy generation plant. According to a preferred embodiment of the invention, a revised second and further forecasts from the first energy producer, which are transmitted to the processor after one or more indications have been transmitted to the first energy producer, are checked again for plausibility.In this process, alerts are transmitted to the first energy producer until the probability meets or exceeds the defined threshold. The advantage is that the quality of the forecasts is continuously reviewed and tracked, thus achieving better coordination within the LEM.
[0026] According to a preferred embodiment of the invention, the verification of incoming forecasts is automated. The advantage of this is that the effort required by an LEM operator for the plausibility check is reduced.
[0027] According to a preferred embodiment of the invention, methods of "transfer learning" can be applied so that when new energy producers are connected to the LEM, relatively good prediction results can be achieved even with little training data available.
[0028] Further advantages, features and details of the invention will become apparent from the following exemplary embodiments and from the accompanying drawings, which are schematic and not to scale:
[0029] Figure 1 shows a representation of the classification of a plausibility check in a LEM;
[0030] Figure 2 shows a representation of inputs and outputs of a plausibility check; and
[0031] Figure 3 shows a representation of how a cloud area influences several energy generation plants. Figure 1 shows an energy producer 1 transmitting 10 a data set with information about a forecast to an LEM 5, wherein the forecast makes a statement about an amount of energy that a first energy generation plant 33 of the first energy producer 1 will provide in the LEM 5 in a defined period of time, e.g. from 0:00 a.m. to 12:00 a.m. of the following day. In the language of the LEM 5, such a forecast is referred to as a "bid". The data set is sent to a pre-processing device 2, where the data set is subjected to pre-processing; for example, it is checked whether the energy producer has a license to submit a forecast in the LEM 5. From there, the data set is transmitted 12 to a plausibility checking device 4, which has a processor 40.
[0032] The plausibility check device 4 also receives one or more further data sets with information on influencing variables Xi , x2, . . . , x p which could influence the amount of energy available. The plausibility check device 4 calculates, with the aid of the processor 40, in a subsequent step, based on the information provided on the influencing variables, a probability that the forecast will come true. The plausibility check device 4 compares, with the aid of the processor 40, in a subsequent step, the calculated probability with a predetermined limit value.
[0033] If the calculated probability falls below the limit value , the device for plausibility checking 4 causes the device for preprocessing 2 to transmit a corresponding indication 11 to the first energy producer 1 .
[0034] In the other case, that the calculated probability reaches or exceeds the limit value, the device for plausibility checking 4 causes the device for pre-processing 2 to further process the forecast and to pass it on to a device for order matching 3 . Figure 2 shows an illustration of the inputs and outputs of a plausibility check 4. A first input (influencing variable) Xi is information about the first energy generation plant, such as coordinates, rated power, installation specifications (e.g. orientation, installation location, angle of inclination). A second input (influencing variable) x2 is information from external providers, such as weather forecasts and weather data. A third input (influencing variable) x3 is information about the power generated in the past; both values from the first energy generation plant and values from other comparable energy generation plants in the LEM 5 are of interest.A fourth input (influencing factor) x4 is information about the forecast, i.e., the offered power. Besides the forecast for the first power generation plant, forecasts for other comparable power generation plants in LEM 5 are also of interest. The output Y of the plausibility check 4 is a probability with which the forecast can be expected to be met.
[0035] Figure 3 shows an illustration of how the spatial distribution of a first and other energy generation plants is taken into account in order to detect weather influences. During the day, a cloud region 30 moves with a westerly wind of speed v over three photovoltaic energy generation plants 31, 32, 33, which are arranged in the west-east direction of movement of the cloud region 30 at the locations 11, 12, 13. The locations 13 and 12 are at a distance 13-12 from one another and the locations 12 and 11 are at a distance 12-11 from one another. The cloud region 30 therefore reaches the three photovoltaic energy generation plants 31, 32, 33 one after the other at the times t1, t2 and t3. Since the cloud region 30 has an extension z in its direction of travel, each of the three photovoltaic energy generation plants 31, 32, 33 is covered by the cloud region 30 during a period Δt = z / v.
[0036] If the cloud area 30 does not bring precipitation or rain, the energy production of the three photovoltaic energy generation systems 31, 32, 33 will decrease at times tl, t2 and t3 due to the onset of shading and will remain at the reduced level during the period Δt before the energy production increases again after the passage of the cloud area 30 due to the direct solar radiation. From the time tl at which the energy production at the first photovoltaic energy generation system 31 decreases, the distances 12-11 and 13-12 as well as the wind speed v at which the cloud region 30 moves over the photovoltaic energy generation systems 31, 32, 33, the times t2, t3 can be calculated at which the two other photovoltaic energy generation systems 32, 33 will also experience a decrease in energy production.The same applies to the increase in energy production after the passage of the cloud region 30 .
[0037] If the cloud region 30 brings snowfall, the energy production of the three photovoltaic energy generation systems 31, 32, 33 will decrease at times t1, t2, and t3 due to the snow cover on the PV modules of the three photovoltaic energy generation systems 31, 32, 33, and will remain at this decreased level until the PV modules are freed from the snow cover, e.g., by sweeping, sliding, or melting. In the event of snowfall, at least the times t2 and t3 of the decrease in energy production can be calculated from the decrease in the first photovoltaic energy generation system 31. The times at which the PV modules are freed from snow and energy production increases again cannot be derived solely from the speed of the cloud area 30, since other influences such as the position, orientation and angle of inclination of the PV modules also play a role.
Claims
Patent claims 1. A method for verifying the plausibility of a forecast regarding the quantity of energy available in a local energy market, wherein the local energy market has a processor (40), comprising the following steps: - Transmit to the processor (40) a first data set containing information about a forecast of a first energy producer (1) which amount of energy a first energy generation plant of the first energy producer (1) will provide in a defined period; - Transmit to the processor (40) one or more additional data sets containing information on influencing variables (x2, x2, ..., x p ) which could influence the amount of energy that can be provided; - Calculate, by the processor (40), a probability (Y) that the forecast can be met, based on the information provided regarding the influencing factors; - Compare, by the processor (40), the calculated probability with a predetermined limit; - if the limit value is undershot, the processor (40) transmits a corresponding notification to the first energy producer.
2. The method of claim 1, wherein the probability (Y) is determined by means of a linear polynomial where ß n Weighting factors of the influencing variables are calculated, or by means of a neural network with sigmoid as the activation function of the output layer.
3. The method of claim 2, wherein the weighting factors ß n obtained by means of a logistic regression from historical data.
4. Method according to one of the preceding claims, wherein the forecast is compared with one or more other forecasts for other energy generation plants comparable to the first energy generation plant.
5. Method according to claim 4, wherein the spatial distribution of the first and the other energy generation plants is taken into account in order to detect weather influences.
6. Method according to one of the preceding claims, wherein the influencing factors include plant information, e.g. geographical coordinates, orientation, installation location, inclination angle of the first energy generation plant.
7. Method according to any of the preceding claims, wherein the influencing factors include weather data and / or weather forecasts.
8. Method according to one of the preceding claims, wherein the influencing factors include information on the quantity of energy provided in the past by the first energy generation plant and / or by other energy generation plants.
9. Method according to one of the preceding claims, wherein the influencing factors include forecasts relating to other energy generation plants of the first energy producer and / or of other energy producers.
10. Method according to one of the preceding claims, wherein a revised second and further forecasts of the first energy producer (1), which are transmitted to the processor (40) after one or more hints have been transmitted to the first energy producer (1), are checked again for plausibility, and as long as hints are transmitted to the first energy producer (1) until the Probability (Y) that the defined limit is met or exceeded.
11. Method according to one of the preceding claims, wherein the examination of incoming forecasts is automated.
12. Device (4) for plausibility testing, configured to perform the method according to one of the preceding claims.
