Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
Applicant’s arguments filed 8/6/2026 with respect to claim(s) 1-20 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-17, and 19-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Lingras et al. (US 20200151836 A1) in view of Dempster et al. (ROCKET: Exceptionally fast and accurate time series classification using random convolutional kernels, 29 Oct 2019) and Lu et al. (US 20210313065 A1).
As to claim 1 Lingras teaches a method, comprising: (see Lingras paragraph [0047] "Referring to FIGS. 3-6, the steps for generating an AI-based building energy model 600 for a client building 860 according to one embodiment of the application are described in the following.", and see Lingras paragraph [0107] "FIG. 8 is a flow chart illustrating operations 800 of modules (e.g., 331) within a data processing system 300 (e.g., a building analytics system 3000 and/or an energy profile dataset and clustering system 4000 within a building energy modeling system, a predictive building control system, a building energy management system, a utility demand response control system, a control system, a supervisory control and data acquisition (“SCADA”) system, and/or an energy management system (“EMS”)) for generating an AI-based building energy model 600 for a client building 860")
for receipt of time-series data indicative of energy consumption associated with a type of building of a plurality of different types of buildings and a climatic zone from a plurality of climatic zones: (see Lingras paragraph [0024] "in terms of clustering of buildings, an individual building is treated as belonging to a cluster determined by core variables including but not limited to building type (e.g., large office, post-1980 construction or secondary school, pre-1980 construction), climate zone specifications, and building envelope parameters", and see Lingras paragraph [0030] "the goal of the clustering system 2000 is to deliver the closest possible building model 740 for a client building 860 (i.e., a building operated and/or owned by a client) based on that building's historical energy consumption data. The underlying assumption of clustering based on historical energy data is that energy data represents in and of itself an identifiable pattern unique to that type of building.", and see Lingras paragraph [0034] "When the client has high-resolution temporal data available (hourly or sub-hourly), it is converted in a special “rough pattern” (or “rough sets” as referred to in machine-learning literature) format consisting of three statistical measures (i.e., mean, maximum and minimum values) for a representative day of each month.", and see Lingras paragraph [0049] "(1) Building type 1010: Large office, post-1980 construction; Medium office, post-1980 construction; Medium office, pre-1980 construction; Small office, post-1980 construction; Primary school, post-1980 construction; and, Secondary school, post-1980 construction; (2) Climate zone specification 1020: 1A (e.g., like Miami, Fla.); 2A (e.g., line Houston, Tex.); 2B (e.g., like Phoenix, Ariz.); 3A (e.g., like Atlanta, Ga.); 3B—Coast (e.g., like Los Angeles, Calif.); 3B (e.g., like Las Vegas, Nev.); 3C (e.g., like San Francisco, Calif.); 4A (e.g., like Baltimore, Md.); 4B (e.g., like Albuquerque, N.M.); 4C (e.g., like Seattle, Wash.); 5A (e.g., like Chicago, Ill.); 5B (e.g., like Boulder, Colo.); 6A (e.g., like Minneapolis, Minn.); 6B (e.g., like Helena, Mont.); 7 (e.g., like Duluth, Minn.); and, 8 (e.g., like Fairbanks, Ak.)")
according to the type of building and the climatic zone; (see Lingras paragraph [0052] "The purpose of the clustering logic is to split the energy profile database 200 into a number of groups 260 representing an aggregate of buildings with similar characteristics.", and see Lingras paragraph [0056] "energy profiles are separated into separate groups based on known attributes about the buildings (such as climate zone specification 1020, building type 1010, etc.). For example, in FIG. 2, climate zone specifications 1020 are used as a grouping parameter, thus all the buildings are first split into 16 climate specification groups (for example), before proceeding with the use of clustering algorithms.", and see Lingras paragraph [0063] "Given that group-splitting was manually performed using the climate zone specification parameter (e.g., using 16 climate zones), which was then followed by clustering each subgroup into 10 clusters, a matrix with 16 rows by 10 columns (climate zone x cluster numbers) is obtained. For each element of the matrix, the discrepancy calculation is performed, and the element with the lowest value in the matrix is indicative of the best cluster match.")
selecting a specialized trained functional neural network (FNN) model corresponding to the classification group, (see Lingras paragraph [0066] "Once the energy profile match 720 has been identified using the energy profile database 200, a corresponding physical building model 740 for that energy profile is fetched from the building models database 100", and see Lingras paragraph [0093] "Step K14: Use prediction techniques such as regression, neural networks, regression trees, random forests, support vector regression to accurately emulate the thermodynamic profiles: (a) The prediction models will be typically developed using weather parameters and setpoints as input; (b) Separate modules may be developed for different time periods; and, (c) There may be a further categorization of prediction models depending on external temperature.", and see Lingras paragraph [0103] "Step C5: Identify the buildings from the knowledge cloud that best match the weather and energy profiles of the new building.", and see Lingras paragraph [0104] "Step C6: Create a decision module that combines the decision trees of the buildings from the knowledge cloud identified in step C5.", and see Lingras paragraph [0110] "selecting a cluster from the set of clusters 270; and, selecting the energy profile 720 in the cluster that is a closest match to that of the client building 860 using the respective representative pattern 250 for the cluster.")
to a short-term energy consumption forecast (see Lingras paragraph [0067] "default control strategies (e.g., pre-set thermostat configurations for different times of day, etc.)", and see Lingras paragraph [0072] "The AI-based building energy model 600 may be used to predict building state and energy consumption rapidly and in real-time.", and see Lingras paragraph [0073] "the rapidly-available predictions of the AI-based building energy model 600 may be used in optimization routines to find the least energy-consumption strategies using various inputs (e.g., an upcoming weather forecast, etc.).", and see Lingras paragraph [0094] "Step K15: Use the prediction modules from step K14 to determine optimum setpoint schedule using evolutionary optimization techniques such as genetic algorithms.", and see Lingras paragraph [0097] "Build a decision tree that predicts the setpoint schedule scenario based on external conditions.")
supplying the time-series data of the classification group to the selected specialized FNN model to obtain the short-term energy consumption forecast. (see Lingras paragraph [0072] "The first input training dataset module I1 790 includes the application of machine learning algorithms and the use of datasets of building dynamics to generate an AI-based building energy model 600 using the physical energy simulation software module 550. The AI-based building energy model 600 may be used to predict building state and energy consumption rapidly and in real-time.", and see Lingras paragraph [0093] "(a) The prediction models will be typically developed using weather parameters and setpoints as input", and see Lingras paragraph [0105] "Step C7: The decision module from step C6 is made available to the client for managing the setpoints for the new building.")
Lingras does not explicitly teach "executing random convolutional kernel (RCK) on the time-series data to generate a classification group of the time-series data", "the specialized trained FNN model comprising a plurality of continuous layers having continuous neurons configured to map time-series data derived functions", and "using integral operations over bivariate parameter functions; and"
However, Dempster teaches
executing random convolutional kernel (RCK) on the time-series data to generate a classification group of the time-series data (see Dempster section [1] "We show that state-of-the-art classification accuracy can be achieved using a fraction of the time required by even these recent, more scalable methods, by transforming time series using random convolutional kernels, and using the transformed features to train a linear classifier. We call this method Rocket (for RandOm Convolutional KErnel Transform).", and see Dempster section [3] "Rocket transforms time series using a large number of random convolutional kernels, i.e., kernels with random length, weights, bias, dilation, and padding. The transformed features are used to train a linear classifier.", and see Dempster section [3.2] "Each kernel is applied to each input time series, producing a feature map. The convolution operation involves a sliding dot product between a kernel and an input time series.", and see Dempster section [3.3] "The transformed features are used to train a linear classifier. Rocket can, in principle, be used with any classifier. We have found that Rocket is very effective when used in conjunction with linear classifiers (which have the capacity to make use of a small amount of information from each of a large number of features).", and see Dempster section [3.3] "(A ridge regression model is trained for each class in a ‘one versus rest’ fashion, with L2 regularization.)")
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the invention of Lingras to use the random convolutional kernel transform of Dempster to classify the building energy time-series data, because Dempster's method achieves state-of-the-art classification accuracy at a fraction of the computational expense and scales linearly to large datasets, making classification of Lingras's growing energy profile database faster and more accurate, and the substitution of one known time-series classification technique for another would have yielded the predictable result of grouping buildings by their energy consumption patterns, see Dempster Abstract and section [1].
Lingras as modified by Dempster does not explicitly teach "the specialized trained FNN model comprising a plurality of continuous layers having continuous neurons configured to map time-series data derived functions"
However, Lu teaches
the specialized trained FNN model comprising a plurality of continuous layers having continuous neurons configured to map time-series data derived functions (see Lu paragraph [0022] "FLM is a popularly used method in functional data analysis (FDA), which deals with data in the form of functions. FLM can be used to analyze data measured over time", and see Lu paragraph [0022] "the proposed FDNN first fits a series of basis functions to each layer respectively. The series of basis functions can model high-dimensional omic data and complex disease phenotypes, considering their underlying structure. The FDNN further builds multiple hidden layers via functional linear models with functional coefficients as weights for the hidden nodes. The multi-layer functional neural network can capture the complex relationship between omic predictors and disease phenotypes.", and see Lu paragraph [0029] "Then, additional D−1 hidden layers can be built recursively with possibly different functional coefficients as shown in Eq. (2.7). X.sup.(d)=σ.sup.(d)(α.sub.0.sup.(d)+∫α.sup.(d)X.sup.(d−1)dt.sup.(d−1)), 1<d≤D.", and see Lu paragraph [0032] "omic data from an individual can be analyzed by the trained FDNN to determine a likelihood of a condition.", and see Lu paragraph [0058] "In the real world, no true functions but rather discrete points are recorded at t.sub.i,j=1, . . . , p.")
using integral operations over bivariate parameter functions; and (see Lu paragraph [0026] "Ŷ.sub.i(s.sub.ij)=Z.sub.iθ+α.sub.0(s.sub.ij)+∫α(s.sub.ij,t)G.sub.i(t)dt, (2.2) where α(s, t) is a bivariate function, and α.sub.0(s) is a function which plays the role as an intercept.", and see Lu paragraph [0029] "When the output is a vector, α.sub.0.sup.(d) is a univariate function and α.sup.(d) is a bivariate function.", and see Lu paragraph [0030] "Compared with DNN, matrix multiplication can be substituted based on the weight matrix W.sup.(d) with integration based on the functional coefficient α.sup.(d) in the FDNN model. The key difference between DNN and FDNN lies that the weights α.sup.(d) and biases α.sub.0.sup.(d) are functions.")
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to further modify Lingras as modified by Dempster to implement the per-group prediction model as the functional neural network of Lu, because Lingras invites the use of neural networks as its prediction technique and Lu's FNN reduces the number of parameters, accounts for the structure of data measured over time, avoids overfitting, and captures non-linear effects, thereby providing more accurate predictions with the predictable result of a trained per-group forecasting model, see Lu paragraph [0030].
As to claim 2, Lingras as modified by Dempster and Lu teaches the method of claim 1,
related to the different types of buildings and the plurality of climatic zones (see Lingras paragraph [0043] "the combination of all values of the parameters 1010, 1020, 1030, 1990 may produce an initial database 100 having 14,400 building models. These models may be simulated using EnergyPlus™ software (for example) 150 to produce energy consumption profiles 200 when combined with weather data 2000", and see Lingras paragraph [0045] "By simulating a building model at various geographical locations (using different weather stations for the same model), and using multiple years of weather data for the simulations, a much larger energy profile database may be obtained due to the addition of these two extra parameters.", and see Lingras paragraph [0072] "The first training dataset generation module T1 770 may use various climatic conditions with various building control strategies in order to generate a cohesive representation of the client building's dynamics under various internal and external conditions.")
to a short-term energy consumption forecast model configured to provide the short-term energy consumption forecast. (see Lingras paragraph [0071] "The physical energy simulation software module 150 then generates and outputs an AI-based building energy model 600 (which may be referred to as a building response model (“BRM”)).", and see Lingras paragraph [0076] "the AI-based building energy model 600 (or BRM) is a rapid-prediction model that has been trained on the datasets generated through a physical simulation software module 550 using the adjusted building model 760.")
Lingras does not explicitly teach "wherein the FNN comprises a plurality of continuous layers trained to map time-series data derived functions"
However, Lu teaches
wherein the FNN comprises a plurality of continuous layers trained to map time-series data derived functions (see Lu paragraph [0022] "The FDNN further builds multiple hidden layers via functional linear models with functional coefficients as weights for the hidden nodes.", and see Lu paragraph [0028] "{W.sup.(d)(d), b.sup.(d)(d), d=1, 2, . . . , D} are coefficients which can be estimated based on performance criteria defined on the Ŷ and Y.", and see Lu paragraph [0030] "By treating the weights and biases as functions, the number of parameters can be reduced, and structure of the data can be taken into account. In addition, FDNN addresses the overfitting issue in the high-dimensional data analysis, and can be easily extended for complex phenotypes (e.g., the progression of disease measured over time and neuroimaging phenotypes).")
As to claim 3, Lingras as modified by Dempster and Lu teaches the method of claim 2,
wherein the RCK is configured to generate the classification group according to the type of building and the climatic zone from a database of class labels used to generate different classes based on class labels, (see Lingras paragraph [0042] "the scope may include five key parameters as follows: (1) Building type 1010; (2) Climate zone specification 1020; (3) Envelope: Window-to-Wall Ratio 1030; (4) Envelope: R-values 1990; and, (5) Envelope: Window U-factors 1995.", and see Lingras paragraph [0055] "two approaches may be combined: manual (supervised) grouping based on existing parameters; and, automatic (unsupervised) grouping based on clustering algorithms.", and see Lingras paragraph [0056] "energy profiles are separated into separate groups based on known attributes about the buildings (such as climate zone specification 1020, building type 1010, etc.).")
wherein the FNN is trained for each of the class labels. (see Lingras paragraph [0089] "Step K11: Find the best representative for each cluster such as ti, si, wi, hi, ei, where i=1, . . . , 5 or 10. Find a representative for each cluster combination such as ti-sj-wk-hl-en, where i,j,k,l,n=1, . . . , 5 or 10", and see Lingras paragraph [0091] "Repeat the following steps for each of the buildings from step K12 according to the ranked priority:", and see Lingras paragraph [0093] "Step K14: Use prediction techniques such as regression, neural networks, regression trees, random forests, support vector regression to accurately emulate the thermodynamic profiles")
As to claim 4, Lingras as modified by Dempster and Lu teaches the method of claim 1,
wherein the short-term energy consumption forecast is based on a selected time window from a plurality of time windows. (see Lingras paragraph [0033] "These outputs are converted into two formats to be used for matching with the client building's historical data: (1) monthly energy profiles; and, (2) month-by-hour rough profiles, and are further referred to as “energy profiles”.", and see Lingras paragraph [0034] "When the client has high-resolution temporal data available (hourly or sub-hourly), it is converted in a special “rough pattern” (or “rough sets” as referred to in machine-learning literature) format consisting of three statistical measures (i.e., mean, maximum and minimum values) for a representative day of each month.", and see Lingras paragraph [0035] "(1) Monthly energy profile for electricity consumption (12 values); (2) Monthly energy profile for thermal energy consumption (12 values); (3) Month×hour rough patterns for electricity consumption (12×24×3 values); and, (4) Month×hour rough patterns for thermal energy consumption (12×24×3 values).", and see Lingras paragraph [0093] "(b) Separate modules may be developed for different time periods")
As to claim 5, Lingras as modified by Dempster and Lu teaches the method of claim 4,
wherein the FNN is trained across the plurality of time windows. (see Lingras paragraph [0045] "By simulating a building model at various geographical locations (using different weather stations for the same model), and using multiple years of weather data for the simulations, a much larger energy profile database may be obtained due to the addition of these two extra parameters. For example, the common values for weather stations may be chosen to be the geographical locations of major cities in the United States and Canada (due to the abundant number of commercial buildings present at those locations) and 2013, 2014 and 2015 weather forecast years.", and see Lingras paragraph [0050] "Each model is matched with an appropriate weather station for its climate zone specification (located at the coordinates of the largest city) and simulated with three years of weather data, resulting in 43,200 building energy profiles.", and see Lingras paragraph [0093] "(b) Separate modules may be developed for different time periods")
As to claim 6, Lingras as modified by Dempster and Lu teaches the method of claim 1,
wherein the time-series data and the short-term energy consumption forecast are represented as periodic functions. (see Lingras paragraph [0033] "These outputs are converted into two formats to be used for matching with the client building's historical data: (1) monthly energy profiles; and, (2) month-by-hour rough profiles, and are further referred to as “energy profiles”. Monthly profiles represent the total sum of energy for each month of the year", and see Lingras paragraph [0034] "When the client has high-resolution temporal data available (hourly or sub-hourly), it is converted in a special “rough pattern” (or “rough sets” as referred to in machine-learning literature) format consisting of three statistical measures (i.e., mean, maximum and minimum values) for a representative day of each month. A representative day is obtained by taking the same hour from each day of the month, and extracting these three statistical measures from it.", and see Lingras paragraph [0035] "(3) Month×hour rough patterns for electricity consumption (12×24×3 values); and, (4) Month×hour rough patterns for thermal energy consumption (12×24×3 values).", and see Lingras paragraph [0086] "Step K8: Extract average, minimum, and maximum hourly profiles for all the twelve months for the energy usages.")
As to claim 7, Lingras as modified by Dempster and Lu teaches the method of claim 1,
wherein the time-series data comprises one or more of temperature time-series data, humidity time-series data, precipitation time-series data, or vehicle count time-series data. (see Lingras paragraph [0080] "Step K2: Extract average, minimum, and maximum hourly profiles for all the twelve months for important weather parameters such as temperature, solar radiation, wind speed, and humidity.", and see Lingras paragraph [0093] "(a) The prediction models will be typically developed using weather parameters and setpoints as input")
As to claim 8, this is directed to a computer-program embodiment that corresponds to method claim 1. See the rejection for claim 1 above, which also applies to claim 8. In addition, Lingras teaches a system having a memory, computer readable instructions, and one or more processors to perform the operations (see Lingras paragraph [0120], “the data processing system 300 may be contained in a computer software product or computer program product (e.g., comprising a non-transitory medium) … which may include a coprocessor or memory according to one embodiment of the application. This integrated circuit product may be installed in the data processing system 300.”)
As to claim 9, this is directed to a computer-program embodiment that corresponds to method claim 2. See the rejection for claim 2 above, which also applies to claim 9.
As to claim 10, this is directed to a computer-program embodiment that corresponds to method claim 3. See the rejection for claim 3 above, which also applies to claim 10.
As to claim 11, this is directed to a computer-program embodiment that corresponds to method claim 4. See the rejection for claim 4 above, which also applies to claim 11.
As to claim 12, this is directed to a computer-program embodiment that corresponds to method claim 5. See the rejection for claim 5 above, which also applies to claim 12.
As to claim 13, this is directed to a computer-program embodiment that corresponds to method claim 6. See the rejection for claim 6 above, which also applies to claim 13.
As to claim 14, this is directed to a computer-program embodiment that corresponds to method claim 7. See the rejection for claim 7 above, which also applies to claim 14.
As to claim 15, this is directed to a system or a computing device claim that corresponds to method claim 1. See the rejection for claim 1 above, which also applies to claim 15. In addition, Lingras teaches a system having a memory, computer readable instructions, and one or more processors to perform the operations (see Lingras paragraph [0120], “the data processing system 300 may be contained in a computer software product or computer program product (e.g., comprising a non-transitory medium) … which may include a coprocessor or memory according to one embodiment of the application. This integrated circuit product may be installed in the data processing system 300.”)
As to claim 16, Lingras as modified by Dempster and Lu teaches the method of claim 1,
wherein the FNN is configured to perform forward propagation and backward propagation to update the bivariate parameter functions until a stopping criterion is reached. (see Lu paragraph [0024] "The technical details of FDNN are summarized in the Appendix B (Forward Propagation) and Appendix C (Back Propagation) below.", and see Lu paragraph [0031] "The solution of this model depends on back-propagation method, which is discussed in Appendix C below.", and see Lu paragraph [0066] "Therefore, the forward propagation algorithm can be written as the following:", and see Lu paragraph [0067] "The back propagation algorithm can be derived from the traditional neural network method.", and see Lu paragraph [0068] "Gradient decent is applied to estimate the weight function coefficients. The recursive process stops when R.sup.(r) converges.")
As to claim 17, Lingras as modified by Dempster and Lu teaches the method of claim 1, wherein the RCK transforms the time-series data using kernels having randomly selected lengths from a set of predetermined values, random weights, random bias, random dilation, and random padding. (see Dempster section [3] "Rocket transforms time series using a large number of random convolutional kernels, i.e., kernels with random length, weights, bias, dilation, and padding.", and see Dempster section [3.1] "Essentially all aspects of the kernels are random: length, weights, bias, dilation, and padding. For each kernel, these values are set as follows (as determined by experimentation to produce the highest classification accuracy on the ‘development’ datasets): – Length. Length is selected randomly from {7, 9, 11} with equal probability, making kernels considerably shorter than input time series in most cases. – Weights. The weights are sampled from a normal distribution, ∀w ∈ W, w ∼ N(0, 1), and are mean centered after being set, ω = W − W. As such, most weights are relatively small, but can take on larger magnitudes. – Bias. Bias is sampled from a uniform distribution, b ∼ U(−1, 1).", and see Dempster section [3.1] "– Dilation. Dilation is sampled on an exponential scale d = ⌊2^x⌋, x ∼ U(0, A), where A = log2 ((l_input − 1)/(l_kernel − 1)), which ensures that the effective length of the kernel, including dilation, is up to the length of the input time series, l_input.", and see Dempster section [3.1] "– Padding. When each kernel is generated, a decision is made (at random, with equal probability) whether or not padding will be used when applying the kernel. If padding is used, an amount of zero padding is appended to the start and end of each time series when applying the kernel, such that the ‘middle’ element of the kernel is centered on every point in the time series")
As to claim 19, Lingras as modified by Dempster and Lu teaches the method of claim 1,
wherein the FNN is configured to incorporate temperature time-series data as a functional feature to improve the short-term energy consumption forecast. (see Lingras paragraph [0080] "Step K2: Extract average, minimum, and maximum hourly profiles for all the twelve months for important weather parameters such as temperature, solar radiation, wind speed, and humidity.", and see Lingras paragraph [0093] "(a) The prediction models will be typically developed using weather parameters and setpoints as input; (b) Separate modules may be developed for different time periods; and, (c) There may be a further categorization of prediction models depending on external temperature.")
As to claim 20, Lingras as modified by Dempster and Lu teaches the method of claim 1,
further comprising maintaining a plurality of specialized FNN models, wherein each specialized FNN model of the plurality is trained using historical time-series data from buildings within a respective classification group defined by a combination of the type of building and the climatic zone, and wherein the selecting the specialized trained FNN model comprises selecting from the plurality of specialized FNN models based on the classification group generated by the RCK. (see Lingras paragraph [0063] "For each element of the matrix, the discrepancy calculation is performed, and the element with the lowest value in the matrix is indicative of the best cluster match.", and see Lingras paragraph [0103] "Step C5: Identify the buildings from the knowledge cloud that best match the weather and energy profiles of the new building.", and see Lingras paragraph [0104] "Step C6: Create a decision module that combines the decision trees of the buildings from the knowledge cloud identified in step C5.", and see Lingras paragraph [0111] "At step 804, a physical building model 740 is selected from a building model database 100 that corresponds to the energy profile 720 for the client building 860.")
Claim(s) 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Lingras et al. (US 20200151836 A1) in view of Dempster et al. (ROCKET: Exceptionally fast and accurate time series classification using random convolutional kernels, 29 Oct 2019) and Lu et al. (US 20210313065 A1) and Biswas et al. (US 20230333528 A1).
As to claim 18, Lingras as modified by Dempster and Lu teaches the method of claim 1,
wherein the short-term energy consumption forecast obtained from the selected specialized FNN model (see Lingras paragraph [0072] "The AI-based building energy model 600 may be used to predict building state and energy consumption rapidly and in real-time.", and see Lingras paragraph [0076] "the AI-based building energy model 600 (or BRM) is a rapid-prediction model that has been trained on the datasets generated through a physical simulation software module 550 using the adjusted building model 760.", and see Lingras paragraph [0104] "Step C6: Create a decision module that combines the decision trees of the buildings from the knowledge cloud identified in step C5.")
based on the classification group corresponding to the type of building and the climatic zone. (see Lingras paragraph [0024] "an individual building is treated as belonging to a cluster determined by core variables including but not limited to building type (e.g., large office, post-1980 construction or secondary school, pre-1980 construction), climate zone specifications", and see Lingras paragraph [0056] "energy profiles are separated into separate groups based on known attributes about the buildings (such as climate zone specification 1020, building type 1010, etc.).", and see Lingras paragraph [0104] "Step C6: Create a decision module that combines the decision trees of the buildings from the knowledge cloud identified in step C5.")
Lingras as modified by Dempster and Lu does not explicitly teach "enables pre-incident planning of optimal power shut-off during a natural disaster"
However, Biswas teaches
enables pre-incident planning of optimal power shut-off during a natural disaster (see Biswas paragraph [0022] "The example embodiments are directed to a host application, such as used by an emergency management system, in which data-driven algorithms use non-power systems data (e.g., geospatial data of transmission assets, geospatial data and timing data of forecasted weather data, ground and aerial crew inspection inputs, load criticality factors, etc.) along with power systems data (e.g., information about transmission lines, substation equipment, etc.) in a scientific manner to automatically determine a sequence of outaging and/or restoring power system equipment.", and see Biswas paragraph [0036] "the system may determine that the power grid 130 is to be outed in response to an emergency situation that includes severe weather such as a wildfire, a hurricane, a storm, and the like. Rather than out the entire grid 130 at once, the example embodiments can identify a sequence among various smaller sub-sections 141, 142, 143, 144, and 145 that can reduce unnecessary downtime and ensure grid stability based on geospatial data and trajectory/timing data of the severe weather.", and see Biswas paragraph [0037] "an outage plan 140 can be generated that includes a sequence in which the sub-sub-sections 141, 142, 143, 144, and 145 should be outed, and the timing information at which such outages should occur.", and see Biswas paragraph [0038] "The system can generate an outage plan and/or restoration plan based on the geolocation and the timing of forecasted weather (such as severe weather) with respect to the geolocation of transmission assets such as transmission lines, transformers, and the like. By incorporating attributes of forecasted weather into the planning, the system can prevent over shedding of power (e.g., unnecessary power shut off, etc.) and at the same time prevent blackouts or brownouts from instability.", and see Biswas paragraph [0039] "knowing the proper timing and sequence of large-scale outage and restoration operations during the planning stage and then following the planned timing and sequence during the execution stage is vital in determining the security and stability of the grid under such challenging situations.", and see Biswas paragraph [0040] "The result is a sequence of instructions that can be displayed via a user interface and/or input into an EMS system where the EMS system is able to automatically shut down the transmission assets on the power grid based on the rankings/priorities generated by the system described herein.", and see Biswas paragraph [0082] "The overall priority output in 218 may include a sequence or transmission assets to be powered down as well as time points at which such transmission assets are to be powered down. This information may then be executed by the system which send signals to the transmission assets to power down based on the instructions output in 218.")
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to further modify Lingras as modified by Dempster and Lu to apply the energy consumption forecasts in the pre-incident outage planning of Biswas, because both operate within an energy management system and Biswas teaches that planning the timing and sequence of power shut-off before severe weather prevents unnecessary power shut-off while avoiding blackouts and grid instability, see Biswas paragraph [0038].
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ABDULLAH K ABOUD whose telephone number is (571)272-0025. The examiner can normally be reached Mon-Fri 8am-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Li B Zhen, can be reached at (571) 272-3768. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/ABDULLAH KHALED ABOUD/Examiner, Art Unit 2121
/Li B. Zhen/Supervisory Patent Examiner, Art Unit 2121
Prosecution Insights