WIP : Long short term memory prediction models for Time series Data

io.openems.edge.application/EdgeApp.bndrun

@@ -167,6 +167,7 @@
 	bnd.identity;id='io.openems.edge.meter.weidmueller',\
 	bnd.identity;id='io.openems.edge.meter.ziehl',\
 	bnd.identity;id='io.openems.edge.onewire.thermometer',\
+	bnd.identity;id='io.openems.edge.predictor.lstmmodel',\
 	bnd.identity;id='io.openems.edge.predictor.persistencemodel',\
 	bnd.identity;id='io.openems.edge.predictor.similardaymodel',\
 	bnd.identity;id='io.openems.edge.pvinverter.cluster',\
@@ -345,6 +346,7 @@
 	io.openems.edge.meter.ziehl;version=snapshot,\
 	io.openems.edge.onewire.thermometer;version=snapshot,\
 	io.openems.edge.predictor.api;version=snapshot,\
+	io.openems.edge.predictor.lstmmodel;version=snapshot,\
 	io.openems.edge.predictor.persistencemodel;version=snapshot,\
 	io.openems.edge.predictor.similardaymodel;version=snapshot,\
 	io.openems.edge.pvinverter.api;version=snapshot,\

io.openems.edge.predictor.lstmmodel/.classpath

@@ -0,0 +1,12 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<classpath>
+	<classpathentry kind="con" path="aQute.bnd.classpath.container"/>
+	<classpathentry kind="con" path="org.eclipse.jdt.launching.JRE_CONTAINER/org.eclipse.jdt.internal.debug.ui.launcher.StandardVMType/JavaSE-17"/>
+	<classpathentry kind="src" output="bin" path="src"/>
+	<classpathentry kind="src" output="bin_test" path="test">
+		<attributes>
+			<attribute name="test" value="true"/>
+		</attributes>
+	</classpathentry>
+	<classpathentry kind="output" path="bin"/>
+</classpath>

io.openems.edge.predictor.lstmmodel/.gitignore

@@ -0,0 +1,2 @@
+/bin_test/
+/generated/

io.openems.edge.predictor.lstmmodel/.project

@@ -0,0 +1,23 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<projectDescription>
+	<name>io.openems.edge.predictor.lstmmodel</name>
+	<comment></comment>
+	<projects>
+	</projects>
+	<buildSpec>
+		<buildCommand>
+			<name>org.eclipse.jdt.core.javabuilder</name>
+			<arguments>
+			</arguments>
+		</buildCommand>
+		<buildCommand>
+			<name>bndtools.core.bndbuilder</name>
+			<arguments>
+			</arguments>
+		</buildCommand>
+	</buildSpec>
+	<natures>
+		<nature>org.eclipse.jdt.core.javanature</nature>
+		<nature>bndtools.core.bndnature</nature>
+	</natures>
+</projectDescription>

io.openems.edge.predictor.lstmmodel/.settings/org.eclipse.core.resources.prefs

@@ -0,0 +1,2 @@
+eclipse.preferences.version=1
+encoding/<project>=UTF-8

io.openems.edge.predictor.lstmmodel/bnd.bnd

@@ -0,0 +1,16 @@
+Bundle-Name: OpenEMS Edge Predictor Lstm-Model
+Bundle-Vendor: OpenEMS Association e.V.
+Bundle-License: https://opensource.org/licenses/EPL-2.0
+Bundle-Version: 1.0.0.${tstamp}
+
+-buildpath: \
+	${buildpath},\
+	io.openems.common,\
+	io.openems.edge.common,\
+	io.openems.edge.controller.api,\
+	io.openems.edge.predictor.api,\
+	io.openems.edge.timedata.api,\
+	org.apache.commons.math3,\
+
+-testpath: \
+	${testpath}

io.openems.edge.predictor.lstmmodel/readme.adoc

@@ -0,0 +1,28 @@
+= Long short term model predictor
+
+The Long Short-Term Memory (LSTM) model is a type of recurrent neural network (RNN) that is particularly well-suited for time series prediction tasks, including consumption and production power predictions, due to its ability to capture dependencies and patterns over time. https://en.wikipedia.org/wiki/Long_short-term_memory[More details of LSTM]
+
+This application is used for predicting power (consumption and production) values.
+Here, For power prediction, LSTM models can analyze historical power data to learn patterns and trends that occur over time, such as:
+
+* Daily and Seasonal Variations: example, Consumption power often follows cyclic patterns (e.g., higher usage during the day, lower at night). Production power often higher during the day and none during the nights.
+* External Factors: LSTM can incorporate external factors like weather, day of the week, or holidays to improve prediction accuracy.
+
+== Training LSTM for  Power Predictions:
+
+* Input Data (Channels address "_sum/ConsumptionActivePower"): Time series data of past consumption levels.
+* Pre-processing: Data needs to be scaled and sometimes transformed to remove seasonality or noise.
+* Training: The LSTM is trained on historical data using techniques like backpropagation through time (BPTT), where it learns to minimize the error between predicted and actual consumption.
+* Prediction: Once trained, the model can predict future power consumption for various time steps ahead (e.g., hours, days, or even weeks).
+
+In practice, LSTMs are favored for their ability to learn complex time-related patterns, making them effective in forecasting energy demand patterns that can inform Energy management system (EMS), energy distribution, and cost optimization strategies.
+
+== Note for activating the predictor
+
+To run this predictor, please create a folder named "models" in the OpenEMS data directory (openems/data/).
+
+Initially, a generic model will be used for predictions, which may not yield optimal results. However, a training process is scheduled to occur every 45 days, during which the models in this directory will be updated. The 45-day interval consists of 30 days for training and 15 days for validation.
+
+As a result of this process, a new model will be trained and will automatically replace the previous one.
+
+https://github.com/OpenEMS/openems/tree/develop/io.openems.edge.predictor.lstmmodel[Source Code icon:github[]]

io.openems.edge.predictor.lstmmodel/src/io/openems/edge/predictor/lstmmodel/Config.java

@@ -0,0 +1,31 @@
+package io.openems.edge.predictor.lstmmodel;
+
+import org.osgi.service.metatype.annotations.AttributeDefinition;
+import org.osgi.service.metatype.annotations.ObjectClassDefinition;
+
+import io.openems.edge.predictor.api.prediction.LogVerbosity;
+
+@ObjectClassDefinition(//
+		name = "Predictor Lstm-Model", //
+		description = "Implements Long Short-Term Memory (LSTM) model, which is a type of recurrent neural network (RNN) designed to capture long-range dependencies in sequential data, such as time series. "
+				+ "This makes LSTMs particularly effective for time series prediction, "
+				+ "as they can learn patterns and trends over time, handling long-term dependencies while filtering out irrelevant information.")
+@interface Config {
+
+	@AttributeDefinition(name = "Component-ID", description = "Unique ID of this Component")
+	String id() default "predictor0";
+
+	@AttributeDefinition(name = "Alias", description = "Human-readable name of this Component; defaults to Component-ID")
+	String alias() default "";
+
+	@AttributeDefinition(name = "Is enabled?", description = "Is this Component enabled?")
+	boolean enabled() default true;
+
+	@AttributeDefinition(name = "Channel-Address", description = "Channel-Address this Predictor is used for, e.g. '_sum/UnmanagedConsumptionActivePower'")
+	String channelAddress();
+
+	@AttributeDefinition(name = "Log-Verbosity", description = "The log verbosity.")
+	LogVerbosity logVerbosity() default LogVerbosity.NONE;
+
+	String webconsole_configurationFactory_nameHint() default "Predictor Lstm-Model [{id}]";
+}

io.openems.edge.predictor.lstmmodel/src/io/openems/edge/predictor/lstmmodel/LstmModel.java

@@ -0,0 +1,129 @@
+package io.openems.edge.predictor.lstmmodel;
+
+import io.openems.common.types.OpenemsType;
+import io.openems.edge.common.channel.BooleanReadChannel;
+import io.openems.edge.common.channel.Doc;
+import io.openems.edge.common.channel.DoubleReadChannel;
+import io.openems.edge.common.channel.StringReadChannel;
+import io.openems.edge.common.channel.value.Value;
+import io.openems.edge.common.component.OpenemsComponent;
+
+public interface LstmModel extends OpenemsComponent {
+
+	public enum ChannelId implements io.openems.edge.common.channel.ChannelId {
+		LAST_TRAINED_TIME(Doc.of(OpenemsType.STRING) //
+				.text("Last trained time in Unixstimestamp")), //
+		MODEL_ERROR(Doc.of(OpenemsType.DOUBLE) //
+				.text("Error in the Model")), //
+		CANNOT_TRAIN_CONDITON(Doc.of(OpenemsType.BOOLEAN) //
+				.text("When the data set is empty, entirely null, or contains 50% null values."));
+
+		private final Doc doc;
+
+		private ChannelId(Doc doc) {
+			this.doc = doc;
+		}
+
+		@Override
+		public Doc doc() {
+			return this.doc;
+		}
+	}
+
+	/**
+	 * Gets the Channel for {@link ChannelId#CANNOT_TRAIN_CONDITON}.
+	 *
+	 * @return the Channel
+	 */
+	public default BooleanReadChannel getCannotTrainConditionChannel() {
+		return this.channel(ChannelId.CANNOT_TRAIN_CONDITON);
+	}
+
+	/**
+	 * Gets the Cannot train condition in boolean. See
+	 * {@link ChannelId#CANNOT_TRAIN_CONDITON}.
+	 *
+	 * @return the Channel {@link Value}
+	 */
+	public default Value<Boolean> getCannotTrainCondition() {
+		return this.getCannotTrainConditionChannel().value();
+	}
+
+	/**
+	 * Internal method to set the 'nextValue' on
+	 * {@link ChannelId#CANNOT_TRAIN_CONDITON} Channel.
+	 *
+	 * @param value the next value
+	 */
+	public default void _setCannotTrainCondition(boolean value) {
+		this.getCannotTrainConditionChannel().setNextValue(value);
+	}
+
+	/**
+	 * Internal method to set the 'nextValue' on
+	 * {@link ChannelId#CANNOT_TRAIN_CONDITON} Channel.
+	 *
+	 * @param value the next value
+	 */
+	public default void _setCannotTrainCondition(Boolean value) {
+		this.getCannotTrainConditionChannel().setNextValue(value);
+	}
+
+	/**
+	 * Gets the Channel for {@link ChannelId#LAST_TRAINED_TIME}.
+	 *
+	 * @return the Channel
+	 */
+	public default StringReadChannel getLastTrainedTimeChannel() {
+		return this.channel(ChannelId.LAST_TRAINED_TIME);
+	}
+
+	/**
+	 * Gets the Last time trained time in Unix time stamp. See
+	 * {@link ChannelId#LAST_TRAINED_TIME}.
+	 *
+	 * @return the Channel {@link Value}
+	 */
+	public default Value<String> getLastTrainedTime() {
+		return this.getLastTrainedTimeChannel().value();
+	}
+
+	/**
+	 * Internal method to set the 'nextValue' on {@link ChannelId#LAST_TRAINED_TIME}
+	 * Channel.
+	 *
+	 * @param value the next value
+	 */
+	public default void _setLastTrainedTime(String value) {
+		this.getLastTrainedTimeChannel().setNextValue(value);
+	}
+
+	/**
+	 * Gets the Channel for {@link ChannelId#MODEL_ERROR}.
+	 *
+	 * @return the Channel
+	 */
+	public default DoubleReadChannel getModelErrorChannel() {
+		return this.channel(ChannelId.MODEL_ERROR);
+	}
+
+	/**
+	 * Gets the Model error. See {@link ChannelId#MODEL_ERROR}.
+	 *
+	 * @return the Channel {@link Value}
+	 */
+	public default Value<Double> getModelError() {
+		return this.getModelErrorChannel().value();
+	}
+
+	/**
+	 * Internal method to set the 'nextValue' on {@link ChannelId#LAST_TRAINED_TIME}
+	 * Channel.
+	 *
+	 * @param value the next value
+	 */
+	public default void _setModelError(Double value) {
+		this.getModelErrorChannel().setNextValue(value);
+	}
+
+}